Binding and Proton Blockage by Amantadine Variants of the Influenza M2WT and M2S31N Explained

Article abstract of DOI:10.1021/acs.jmedchem.6b01115

While aminoadamantanes are well-established inhibitors of the influenza A M2 proton channel, the mechanisms by which they are rendered ineffective against M2_S31N are unclear. Solid state NMR, isothermal titration calorimetry, electrophysiology,

Full text of DOI:10.1021/acs.jmedchem.6b01115

Subscriber access provided by University of Newcastle, Australia
Article
Binding and Proton Blockage by Amantadine Variants
of the Influenza M2WT and M2S31N Explained
Christina Tzitzoglaki, Anna K. Wright, Kathrin-Maria Freudenberger, Anja
Hoffmann, Ian Tietjen, Ioannis Stylianakis, Felix Kolarov, David Fedida, Michaela
Schmidtke, Guenter Gauglitz, Timothy A. Cross, and Antonios Kolocouris
J. Med. Chem., Just Accepted Manuscript • DOI: 10.1021/acs.jmedchem.6b01115 • Publication Date (Web): 20 Jan 2017
Downloaded from http://pubs.acs.org on February 4, 2017
Just Accepted
“Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted
online prior to technical editing, formatting for publication and author proofing. The American Chemical
Society provides “Just Accepted” as a free service to the research community to expedite the
dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts
appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been
fully peer reviewed, but should not be considered the official version of record. They are accessible to all
readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered
to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published
in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just
Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor
changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers
and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors
or consequences arising from the use of information contained in these “Just Accepted” manuscripts.
Journal of Medicinal Chemistry is published by the American Chemical Society. 1155
Sixteenth Street N.W., Washington, DC 20036
Published by American Chemical Society. Copyright © American Chemical Society.
However, no copyright claim is made to original U.S. Government works, or works
produced by employees of any Commonwealth realm Crown government in the course
of their duties.

Page 1 of 66
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
Binding and Proton Blockage by Amantadine Variants of the
Influenza M2 and M2
Explained
WT
S31N
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
1
2
3
4
5
3
Christina Tzitzoglaki, Anna Wright, Kathrin Freudenberger, Anja Hoffmann, Ian Tietjen,
1
3
5
4
Ioannis Stylianakis, Felix Kolarov, David Fedida, Michaela Schmidtke, Günter Gauglitz,
2,6
1,*
Timothy A. Cross, Antonios Kolocouris
1
Section of Pharmaceutical Chemistry, Department of Pharmacy, National and Kapodistrian
University of Athens, Athens, Greece.
2
Institute of Molecular Biophysics and National High Magnetic Field Laboratory, Florida State
University, Tallahassee, FL 32306.
3
Institut für Physikalische und Theoretische Chemie, EberhardꢀKarls Universität, Auf der
Morgenstelle 18, Dꢀ72076 Tübingen, Germany.
4
Jena University Hospital, Department of Virology and Antiviral Therapy, Hans Knoell Str. 2, Dꢀ
0
7745 Jena, Germany.
5
Department of Anesthesiology, Pharmacology and Therapeutics, University of British Columbia,
Vancouver, British Columbia, V6T 1Z3, Canada.
6
Department of Chemistry and Biochemistry, Florida State University, Tallahassee, FL 32306
2
0/01/201718/01/2017
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 2 of 66
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
ABSTRACT
While aminoadamantanes are wellꢀestablished inhibitors of the influenza A M2 proton channel, the
mechanisms by which they are rendered ineffective against M2S31N are unclear. Solid state NMR,
isothermal titration calorimetry, electrophysiology, antiꢀviral assays and molecular dynamics
simulations suggest stronger binding interactions for aminoadamantanes to M2WT compared to
negligible or weak binding to M2S31N. This is due to reshaping of the M2pore when N31 is present,
which in contrast to wild type (WT), leads: A) to the loss of the V27 pocket for the adamantyl cage
and to a predominant orientation of the ligand's ammonium group toward the Nꢀterminus and, B) to
the lack of a helical kink upon ligand binding. The kink, which reduces the tilt of the Cꢀterminal
helical domain relative to the bilayer normal includes the W41 primary gate for proton conductance
and may prevent the gate from opening, representing an alternative view for how these drugs
prevent proton conductance.
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
ACS Paragon Plus Environment

Page 3 of 66
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
INTRODUCTION
A proven vulnerability of influenza A infections is the blockage of the viral M2 proton channel. M2
is required for acidification of the virion interior during infection and neutralization of the transꢀ
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
1
Golgi network during viral egress. Amantadine ((Amt), 1) (Scheme 1) is an established inhibitor of
2
3
the influenza A/M2ꢀmediated proton currents and a licensed influenza A infection therapy. The
primary binding site of 1 is located within the pore of the tetrameric M2 transmembrane (M2TM)
4,5
6
domain that forms the transmembrane proton transit path. However, since 2005, the 1ꢀinsensitive
7
SerꢀtoꢀAsn mutation at position 31 in M2 (S31N) has become globally prevalent, abrogating the
8
9
clinical usefulness of 1 and possibly other previously reported M2 inhibitors. Thus, new agents
are needed to combat drugꢀresistant forms of influenza.
4
a,5a,b
Based on the experimental solid state NMR (ssNMR) structure of the complex M2TMꢀ1
,
the replacement of S31 residues at the drug binding site with the more polar and bulky Asn
sidechains may induce the adamantyl ring to move deeper into the M2 pore, toward the Cꢀterminus
close to G34, where the pore of the helical bundle has its largest diameter, and drugs larger than 1
might be effective. We and others have searched for such larger drugs and found, for example
compounds 5ꢀ7, which are active against A/Calif/07/2009 (H1N1) virus encoding S31N but not
10
against WSN/33 also bearing the same mutation. However, while a preliminary ssNMR
experiment of the S31N M2TM domain complexed with compound 5 showed evidence of drug
binding, electrophysiology (EP) studies showed no blockage for the full length S31N M2
10a
protein.
These studies therefore indicate that more detailed experimental investigation and
modelling are needed to explain the mechanisms by which aminoadamantanes are rendered
ineffective against M2S31N in order to design effective drugs.
Recent molecular dynamics (MD) simulations showed that the ammonium group of 1 has a
prevalent orientation toward the Cꢀterminus of M2TMWT (Udorn), with the adamantane
11ꢀ18
hydrocarbon cage shifted slightly toward Nꢀterminus of the A30/S31 Cα sites
, producing a
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 4 of 66
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
1
7,18
region of the pore devoid of water
consistent with blockage of proton transport and in
5
c,20,21
agreement with ssNMR findings.
In contrast, MD simulations of 1 in M2TMS31N from Gleed
17b
16
et al. and rimantadine in M2TMS31N from Alhadeff et al. suggested that these compounds have a
1
7b,22
variable orientation, but with a propensity to have the ammonium group toward the Nꢀterminus
of the M2TMS31N pore (Figure 1). In this orientation water occupancy is not widely occluded in any
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
17
16,17b
configuration. These studies
presented a description of the Potential of Mean Force (PMF)
curves of amantadine and rimantadine in interaction with the M2TMS31N pore but the main
structural/interaction changes due to S31N and leading to aminoadamantane resistance were not
defined. In addition, no experimental evidence is presented to date for the binding of
aminoadamantane ligands in M2TMS31N pore. Nevertheless, the propensity of 1’s amino group to
orient towards the Nꢀterminus was used for the successful design of 1ꢀpolar head conjugates that
23
were shown to be potent inhibitors of WSN/33 virus and proton conductance by M2S31N . A
ssNMR analysis of a system including one of these compounds, a phenylisoxazole derivative linked
with amantadine through a methylene bridge, with M2TM in membrane bilayers was realized. The
results showed that the compound's heterocyclic ring system may be trapped by the V27 side chains
24
at Nꢀterminus of the M2TMꢀpore with the isoxazole group forming hydrogen bonds with the N31
amide side chains.
We are interested to investigate how subtle changes in amantadine structure are related with
the binding affinity against M2TM variants. We previously used molecular mechanics Poissonꢀ
Boltzmann surface area (MM/PBSA) to interpret thermodynamic profiles measured using
isothermal titration calorimetry (ITC) for aminoadamantanes binding to the avian M2TMWeybridge
(
M2TMWeybridge has different two amino acids that do not line into the pore, i.e., V28→I28 and
L38→F38 compared to M2TMWT) in order to successfully prioritize aminoadamantane
25
derivatives. We also applied rigorous free energy binding calculations by the Bennett acceptance
ratio (BAR) approach to accurately predict relative binding affinities of aminoadamantane
ACS Paragon Plus Environment

Page 5 of 66
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
derivatives towards M2TMWT or M2TMWeybridge measured by ITC or other methods
2
6,27
respectively.
Here we investigate the interaction of aminoadamantane derivatives (Scheme 1) with
M2TMWT and M2TMS31N both experimentally and by MD simulations, with the aim of determining
how the S31N mutation leads to the inability of 1 and other aminoadamantanes to block S31N M2
proton transport. First, we measured binding constants of selected aminoadamantane derivatives
with ITC against M2TMWT and M2TMS31N in dodecylphosphocholine (DPC) micelles at alkaline
pH. Second, we applied a series of MD simulations (80 ns trajectories) of the M2TMWT and
M2TMS31N complexes with 1ꢀ10 (Scheme 1) to analyze in some detail how the polar N31 amide
side chains affect binding of an aminoadamantane ligand inside the M2TM pore. To this end, we
suggest that repulsive forces between N31 and the adamantane ring may influence the orientation
and binding interactions of the ligand, the overall stability of the complexes, and the shape of
M2TM. Third, we performed Oriented Sample (OS) and Magic Angle Spinning (MAS) ssNMR
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
28ꢀ30
experiments
on representative M2TMꢀligand complexes at alkaline pH using both M2TMWT
and M2TMS31N sequences to characterize the structural influence of compounds on the channel.
Fourth, we selected and tested new synthetic derivatives with slightly larger or similarlyꢀsized
adducts, like 8 and 10, than previously synthesized compounds 5ꢀ7, 9 for their abilities to bind
M2TMS31N. We realized EP experiments for selected compounds against both WT and S31N M2
proteins to test M2 channel blockage and additionally measured the antiviral potency of the
compounds against the naturally amantadineꢀresistant H1N1 influenza strain A/WSN/1933
(WSN/33 with M2N31) and its reverse geneticsꢀgenerated amantadineꢀsensitive variant with the
amino acid substitution N31S in M2 (WSN/33ꢀM2ꢀN31S) using cytopathic effect inhibition (CPE)
assays.
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 6 of 66
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
RESULTS
Binding affinities of aminoadamantanes to M2TM by ITC
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
3
1
ITC measurements were determined for both M2TMWT and M2TMS31N tetramers at pH 8
32
corresponding to the closed state of the M2TM pore (Table 1). Binding data for compounds 1, 5,
2
6
6
ꢀ8 to M2TMWT measured in our previous work were included in Table 1. Thus, as has been
2
6
published recently (see Tables 1 and S1) compounds 1ꢀ8 bind M2TM . Compound 9 and the
WT
newly synthesized analogue 10, which has a cyclohexane ring instead of the cyclopentane in 9, bind
d
M2TMWT with at least an order of magnitude lower dissociation constant (K ) than those for 1ꢀ8,
possibly because their size do not enable them to fit well inside M2TMWT. We note that techniques
for measuring affinity against M2TM such as ITC correspond to models of molecular recognition
for ligand binding to M2TM and do not necessarily reflect functional inhibition of M2 proton
currents or in vitro inhibition of influenza A virus. For example, 9 inhibits M2WTꢀdependent
currents as measured by electrophysiology with an IC50 of 10 ± 2 µM (Table 4) and inhibited
influenza A virus replication in infected cells with an IC50 of 3.92 ± 2.22 µM (Table 5) while
compound 10 was inactive. Nevertheless, ITC and related techniques like Surface Plasmon
Resonance (SPR) and structural techniques like ssNMR can provide useful insights for M2TM
binding and trends in structureꢀactivity relationships for the M2ꢀaminoadamantane system. For
example, compounds 1ꢀ4 did not bind efficiently to M2TMS31N according to ITC and previous SPR
3
3
measurements for 1 while 5, 6 with larger adducts connected to adamantane bind weakly to
M2TMS31N compared to M2TMWT according to ITC (Table 1) and ssNMR. The K values of low
d
affinity binders (e.g. 5, 6, 9, 10) against M2TMS31N, and 9 against M2TM , possess relatively
WT
large errors due to the limitations of the ITC method (as explained in the Supporting Information).
d
Subsequently comparison of the relative K values of compounds 5, 6, 9, 10 do not reflect relative
ACS Paragon Plus Environment

Page 7 of 66
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
binding affinity strength against M2TMS31N. Nevertheless, although the quantitation of the ITC
results against M2TMS31N is limited for this method, the measurements suggest that the K
d
values of
5
and 6 against M2TMS31N are much smaller compared to M2TMWT suggest weaker binding to
M2TMS31N (Table 1). Narrower linewidths and larger chemical shifts changes for 5 bound to
M2TMWT compared to M2TMS31N were observed in the ssNMR spectra described below suggesting
a reduction in dynamics for binding to M2TMWT and a more specific binding site, which in turn
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
31
causes a significant reduction in hydration due to drug induced desolvation of the binding pocket.
Solid state NMR of M2TM ꢀ aminoadamantanes complexes
OS ssNMR spectra
Drug binding to the channel pore of M2TMWT and M2TMS31N was evaluated using OS ssNMR
15
experiments. Spectral correlation observed in these data sets, between anisotropic N chemical
1
15
15
shifts and Hꢀ N dipolar coupling values for selectively Nꢀlabeled backbone amide sites, gives
34,35
rise to resonance patterns known as PISA (polarity index slant angle) wheels.
The shape, size
and position of the PISA wheel for αꢀhelical membrane proteins in uniformly oriented lipid bilayer
preparations is determined by the helical tilt relative to the bilayer normal, and is sensitive to drugꢀ
induced structural perturbations. Previous investigations by Cross and coꢀworkers have reported
dramatic changes in anisotropic chemical shifts and dipolar coupling values for M2TMWT in the
36
36
presence of 1. PISA wheel analysis of M2TMWT spectra in the apo state and in the presence of 1
or 5 correlates the shifts with a substantial reduction in the tilt angle for the Cꢀterminal region of the
M2TMWT helix, following ligand binding. In contrast, M2TMS31N in the presence of 5 results in a
1
0a
uniform 5 ± 2° decrease in the helical tilt imparting a change to the entire helixꢀhelix interface.
Here, to monitor changes in the resonance frequencies and to the resulting PISA wheel in the
15
presence of other variants of 1, a NꢀV28,A30,I42 (VAIꢀM2TM) labeling scheme was used.
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 8 of 66
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
Residues 28 and 30 are in the Nꢀterminal domain of the M2TM helix and I42 in the Cꢀterminal
region, and hence the labels effectively sample both segments. Resonance frequencies for these
sites are well resolved allowing for a PISA wheel analysis of drug binding in the channel pore that
induces small structural perturbations. The signals of these three pertinent and isotopically labeled
backbone amides were shifted in nearly an identical fashion when 1 or 5 bind to the M2TMWT
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
(Figure 2). As before, changes suggest only a slight perturbation to the Nꢀterminus and a much
more significant structural perturbation to the Cꢀterminus with a change in the tilt angle from 32° to
4b,36
just 22°, while the Nꢀterminal tilt changes by only a degree or two (Table 2).
These results
imply similar drug induced structural perturbations to the TM configuration of WT M2 protein in
the presence of 1 or 5.
For the binding studies to M2TMS31N, no observable effect was detected for binding of 1,
while 5 induces smaller changes compared to those induced by the binding of 1 or 5 to M2TMWT
based on the signals from the same three backbone amides (Figure 3). In addition, the linewidths are
narrower when 5 is bound to M2TMWT compared to M2TMS31N, suggesting less dynamics in the
WT complex implying a tighter complex.
There is a small difference in the structure of M2TMS31N versus M2TMWT without drugs in
lipid bilayers (black resonances in Figure 2 versus Figure 3). The M2TMS31N data suggests a helical
tilt of approximately 38°, somewhat greater than the helical tilt observed in the WT structure (32°)
4
,36
consistent with prior characterizations.
While the binding of 1 to M2TMWT produced a 10° kink
4b,36
near G34 in each helix of the tetramer,
ꢀresistant M2TMS31N. The shifts in the anisotropic resonance frequencies which resulted when 5
binding of 1 produced no such structural change for the
1
binds to M2TMS31N demonstrates a significant change in helical tilt from 38° to 33° in the
tetrameric complex (Table 3). Based on these resonances that sample the Nꢀ and Cꢀ regions of the
TM helix, the structural change appears to be a uniform change in tilt. Consequently, the residues
facing the pore remain the same, but the interactions between helices that provide the tetrameric
stability may change substantially. Upon binding of 1 or 5 to M2TMWT a kinked helix at G34 is
ACS Paragon Plus Environment

Page 9 of 66
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
produced but when adding these compounds to M2TMS31N helices are not kinked at G34. Instead,
when 5 binds to M2TMS31N, the entire helixꢀhelix interface changes with the ~5° reduction in tilt for
each of the four helices.
The structurally similar aminoadamantanes 6ꢀ8 produced similar ssNMR results when added
to M2TMS31N (Figure 4) in that the change in structure always reflected a uniform change in helical
tilt. 8 is a new derivative, which was synthesized to test the effect of Cꢀmethylation to binding
compared to Nꢀmethylation in 6. The measured helical tilt angles are 28° for 6, 31° for 7, and 33°
for 8, with the latter compound yielding results that are similar to those induced by 5 (Table 3). For
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
5
ꢀ8, there is no evidence for partial binding, since the unbound states are not observed in the spectra
with drugs present. With 5, V28 appears to have multiple resonances (Figure 3) clustered on an
anisotropic chemical shift of 160 ppm and a dipolar interaction of 4 kHz. The structural differences
would be small, i.e. no more than 1° change in the orientation of the V27ꢀV28 peptide plane with
respect to the bilayer normal. This suggests that even with 5 on the timescale of 10ꢀ100µs there is
no significant evidence for asymmetry in this tetrameric structure, i.e. there is no evidence from
these three labeled sites that M2TMS31N bound with different compounds 1, 5ꢀ8 forms a dimer of
37,38
dimer, as has been suggested for the S31N structure.
If such structures are present they must
interconvert on a timescale faster than the difference in spectral frequencies in Hz. Furthermore, if
this was occurring with significantly different structures there would be a reduction in the width and
height of the PISA wheel, which was not observed. The intensity of the resonances varies, again
especially for 5, suggesting that the compound may have competing binding interactions with the
pore leading to heterogeneity in the frequencies, the weakening of the A30 resonance and the small
dispersion of the V28 resonances as described above. These dynamics are in sharp contrast to the
intense resonance for I42 for M2TMS31N bound with 5. The uniformity of the resonance intensities
that result from the interaction with heterocyclic compounds 6ꢀ8 may suggest less dynamics and
more potent binders than 5 to M2TMS31N
.
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 10 of 66
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
MAS spectra
To evaluate further the effect of 6 binding to M2TMS31N, 2D NꢀCα correlation magic angle spinning
1
3
15
(MAS) ssNMR experiments were performed using uniformly C, Nꢀlabeled V7, A30, S31 and
13
15
G34 M2TMS31N ( C, NꢀVANG labeled M2TMS31N). The labeling used aims to explore the binding
interactions of amantadine analogs into M2TMS31N pore through measuring the effect of binding to
the chemical shifts of V27, A30, G34 and N31. In the M2TMWT the binding area includes the
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
4
a,5b
expanded area including V27, A30 and also G34, S31.
Spectra were obtained (Figure 5) for both
the apo protein (blue) and for 6 bound (red) to the labeled M2TMS31N. Addition of the drug to the
M2TMS31N sample resulted in chemical shift changes for N31 and G34 of 1.2 and 2.1 ppm,
respectively, V27 and A30 resonances remain unchanged, suggesting that they are not involved in
binding, while the isotropic chemical shift perturbation of the N31 and G34 residues suggest the
drug binding site for M2TMS31N suggesting either direct hydrogen bonding with the backbone or
2
1
15
13
indirect hydrogen bonding through water molecules. Significant N and/or Cα chemical shift
20,21
changes at V27, S31, G34 have been reported when rimantadine is bound to M2TMWT
with S31
21
experiencing a dramatic 7 ppm shift relative to the apo state. Despite the observed chemical shift
changes at residues N31 and G34 when 6 bound to the M2TMS31N, cross peak intensity was not
21
increased as was observed for the WT full length M2 in complex with rimantadine. This suggests
that the conformational heterogeneity of M2TMS31N was not increased for these residues in the
presence of the drug. A reduction in dynamics which is in agreement with narrower linewidths
would be anticipated with a specific binding site or a significant reduction in hydration due to drug
induced desolvation of the binding pocket. The narrower linewidths and increased anisotropic
chemical shifts for the M2TMWT and M2TMS31N bound to 5 (Figures 2, 3) illustrate reduced
dynamics and weaker or less specific binding of the drug to the M2TMS31N
.
ACS Paragon Plus Environment

Page 11 of 66
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
MD simulations of M2TM ꢀ aminoadamantanes complexes
Starting structure of protein ligandꢀcomplex
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
MD simulations of the complexes between 1, 2ꢀalkylꢀ2ꢀaminoadamantanes 2ꢀ5 and cyclic
derivatives 6ꢀ10 with both M2TMWT and M2TMS31N provide insights for the binding interactions
and possible structural changes in the binding area. All of the ligands' amino groups were
considered to be protonated according to model calculations preformed previously for ligands 1 and
2
5,26
4a,5b
6.
The structure of the complex M2TMWTꢀ1 (PDB ID 2KQT ) was used as a starting structure
for the simulations of M2TMWTꢀligand complexes and of M2TMS31Nꢀligand complexes. The PDB
4a,5b
ID 2KQT
structure was chosen since it was determined at pH 7.5 and DMPC planar bilayers and
4
a,5b
vesicles , i.e., similar conditions to those used in our simulations. Subsequently only small
structural changes are to be expected in M2TM structure during MD simulations of its complexes
with ligands and the equilibration phase should be short. It has been shown that the stability of the
binding region for 1 in the M2TM tetramer is increased considerably when using DMPC compared
3
9ꢀ42
to other lipids.
In addition, 2KQT (ref. 5b) should be considered as the best structure of the
amantadine bound state of M2TM since it utilized both isotropic chemical shift restraints and all of
the orientational restraints of an earlier structure characterized by Cross and coworkers (see ref. 4a).
The structure of the M2TMS31Nꢀligand (1ꢀ10) complexes were generated from M2TMWTꢀcomplexes
by mutating S31 to N (see also Experimental Part).
Complexes of ligands with M2TMWT
The MD simulations of the M2TM ꢀligand complexes reached equilibration in less than 40 ns
WT
with the protein system having full flexibility. The RMSDs values (Table S2) were ≤ 1.6 Å for
4a,5b
M2TM Cαꢀcarbons with respect to the initial structure of the production period
suggesting that
M2TMWTꢀligand (1ꢀ8) complexes were very stable and the M2TM tetramer structure was
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 12 of 66
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
considerably unchanged in the course of the simulation. The mean values of the Nꢀterminal and the
Cꢀterminal helical tilt of M2TMWT for the complexes with 1, 5 from the corresponding MD
trajectories were measured to be 31°, 18° and 31°, 19° respectively and are in very good agreement
with the values of 32°, 22° and 32°, 22° determined by OS ssNMR (Table 2). As noted above the
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
4
a,5b
starting structure 2KQT
having helical domain tilt angles of 31°, 19° were determined in similar
4a
conditions to those used in our MD simulations by Cross and coworkers. Even so these findings
suggest that the ligandꢀM2TMWT complex was successfully equilibrated and the conditions were
5
b,c
well adjusted to produce consistent results. Consistent with the experimental findings
and
25,26
previous observations,
the ammonium group of the aminoadamantane compounds was oriented
towards the Cꢀterminus with angles of < 77° between the CꢀN bond of the ligand and the
membrane's normal (Table S2). The complexes of compounds 1ꢀ8 inside the M2TM lumen are
stabilized through: a) formation of hydrogen bonds between their ammonium group and water
molecules in the region between H37 residues and the ligand and b) van der Waals interactions
between adamantyl group and a lipophilic pocket formed by V27 and A30 side chains (see Figure
1
5
6
). This binding area is in accordance (a) with the N chemical shifts perturbations for V27, A30
and S31 in complexes between 1 or 3ꢀazaspiro[5,5]undecane and M2TMWT compared to the apo
39
M2TMWT. and, (b) with the measured distance of 4.5 Å between G34Ca and the methyl group of
21
rimantadine from REDOR ssNMR experiments in studies of the full length protein. In addition,
for complexes of M2TMWT with 1ꢀ8, the distance between the adamantyl ring of the ligand and the
center of mass of the four V27 (V27ꢀAd) varies between 4 and 4.9 Å, and the distance between the
adamantyl ring of the ligand and the center of mass of the four A30 (A30ꢀAd) varies between 0.5
and 1.4 Å (Table S2). The position of the adamantyl ring inside the lumen was similar for ligands 2ꢀ
8 differing from 1 only by 0.2ꢀ0.6 Å. To account for the position of the ligands towards the Cꢀend
we measured the distance between the center of mass of the four V27 and the ammonium nitrogen
+
of the ligand (V27ꢀN ) which varied between 7.1 and 7.8 Å (Table S2). According to the measures
in Table S2 hydrogen bonding ability is higher for the subset of ligands 1ꢀ5, 9, 10 than for the
ACS Paragon Plus Environment

Page 13 of 66
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
subset 6, 8 or 7. Thus 1ꢀ5, 9, 10 can form on average 3 hydrogen bonds with neighboring water
molecules, 6, 8 can form 2 hydrogen bonds, and 7 one hydrogen bond respectively (Table S2). This
is what was expected since 1ꢀ5, 9, 10 have a primary ammonium group, 6, 8 have a secondary
ammonium group and 7 have a tertiary ammonium group. For ligands 3ꢀ8 which include a carbon
substituent at Cꢀ2 adamantane position the molecule is rotated in order to avoid repulsive van der
Waals forces of the alkyl group with the symmetric M2TM pore. The average angle between the
pore axis and CꢀN bond vector was increased progressively according to the alkyl group size, i.e.,
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
o
o
from 13.5 for 1 to 25ꢀ35 for 2ꢀ8 (Table S2). Although all ligands 1ꢀ8 have similar binding mode
as shown from the OS ssNMR spectra of complexes including 1, 5 (Figure 2), a subtle compromise
between hydrogen bonding and hydrophobic interactions with key pore residues, such as V27, A30,
affects the ligand tilt inside the pore and its different binding strength as shown by the ITC results
(Table 1, Table S1). The adamantyl ring is slightly toward the Nꢀterminus compared to the A30/S31
Cα, producing a region without water molecules. Snapshots of the simulation complexes with 5, 6
are shown in Figure 6.
The MD simulations suggest that the M2TMWTꢀligand complex stability is reduced when a
sizeable adduct is attached to adamantane, i.e., the M2TM pore does not efficiently accommodate
the large adducts in compounds 9 and 10. This is in accordance with the weak or no binding
respectively of these drugs according to the ITC results (Table 1). For example, the MD simulations
showed that the stability of the complex is considerably reduced in the case of 10 where after 80 ns
the ligand is shifted toward the Nꢀterminus; the mean distances of the trajectory V27ꢀAd, A30ꢀAd
+
and V27ꢀN are 3.5 Å, 4.7 Å and 2.1 Å, respectively, suggesting a loosening of the M2TMꢀcomplex
integrity at the Nꢀterminus (a snapshot of the simulation complex with 10 is depicted in Figure
S1.a).
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 14 of 66
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
Complexes of ligands with M2TMS31N
In the starting configuration of M2TMS31Nꢀaminoadamantane ligand complex the ammonium group
points toward the Cꢀterminus forming Hꢀbonds with waters between the ligand and H37 residues as
in the M2TMWT pore. The MD simulations of the complexes between M2TMS31N and 1ꢀ8 reach
equilibration in less than 60 ns with the protein system having full flexibility. In complexes with
M2TMS31N the ligand is more mobile inside the pore having a propensity to orient its ammonium
group toward the Nꢀterminus, in contrast to M2TMWTꢀligand complexes where the ligand forms a
strong complex with ammonium group oriented toward the Cꢀterminus. Thus, after a few ns of
unrestrained dynamics the ligand moves toward the Cꢀterminus by ~2 Å, probably because the
adamantyl group is repelled by the polar N31 side chains resulting in the loss of the V27 lipophilic
pocket, and the molecule rotates 180º through an attraction to the polar environment around N31.
This finding is consistent with the narrower linewidths and larger chemical shifts changes for
M2TMWT when bound with compound 5 compared to M2TMS31N as observed in the ssNMR
spectra suggesting a stronger binding interaction with M2TMWT. In addition waters are transferred
from the region between the ligand and H37 to the area around N31 where the drug's amine can
interact with the carbonyls of the N31 amide groups and waters around the amide side chains.
Snapshots of the MD simulation complexes of M2TMS31N with 1, 5, 6 and 10 are depicted in Figure
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
7
. The MD runs showed that by progressively increasing the size of the adduct connected to the
adamantyl moiety in complexes with 2ꢀ8 the ammonium group orientation turns toward the Nꢀ
terminus, and the drug keeps this orientation during the entire simulation period. While this change
was observed for 1 after ~ 20 ns of production, the time needed for the aminoadamantane ligand to
rotate towards the Nꢀterminus may be increased for some larger adducts. For example, in the case of
5
the ligand's ammonium group keeps its orientation toward the Cꢀterminus during the first 40 ns
and then the molecule turns towards the Nꢀterminus until the end of the 80 ns production time.
While no significant conformational change was observed for M2TM31N tetramer in its complexes
with 1ꢀ8 during the production period, the RMSDs for M2TMS31N Cαꢀcarbons were ≤ 2.8 Å from
ACS Paragon Plus Environment

Page 15 of 66
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
4
a,5b
the initial structure , i.e., 1 to 1.5 Å higher than the RMSDs for M2TMWT Cαꢀcarbons suggesting
less dynamics for M2TMWT complexes (Table S3). The mean values of the helical tilt angles for the
complexes of 1, 5ꢀ8 with M2TMS31N from the corresponding MD trajectories were measured to be ~
3
4°, 31°, 28°, 33°, 31° respectively and are close to 38°, 33°, 28°, 31° and 33° determined by OS
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
ssNMR (Table 3). The uniformly tilted helical structure is a conversion form the initial kinked state
4
a,5b
in the 80ꢀns MD, where the starting structure reflects a homology model of 2KQT.
Here the
major finding is the lack of kinked helices in M2TMS31N compared to the OS ssNMR structure
when 1 binds M2TMWT suggesting than the MD simulations describe well the ligandꢀM2TMS31N
structure.
For most of the ligands 1ꢀ8 the mean angle between the CꢀN bond vector and the normal of
the membrane is > 140° for more than 40 ns of the simulation reflecting the propensity of
ammonium group to orient towards the Nꢀterminus (Table S3). The distance between the center of
mass of the four V27 and the adamantyl ring of the ligands 1ꢀ8 (V27ꢀAd) was longer than in
complexes with the M2TM , which is consistent with an orientation of the ammonium group
WT
towards the Nꢀterminus. On average, the adamantyl ring in the M2TMS31N was found to be ~ 1ꢀ2 Å
toward the Cꢀterminus compared to the M2TMWT complexes. Accordingly, the distance between
the center of mass of the four V27 and the ligand’s amine nitrogen in M2TMS31N complexes was
shorter by 3ꢀ3.5 Å compared to M2TMWT complexes. In addition especially for the ligands with
large adducts the adamantyl ring is positioned toward the Cꢀterminus close to G34 (see distances
A30ꢀAd and G34ꢀAd in Table S3) whereas in the M2TMWT complexes, adamantyl ring is shifted
toward the Nꢀterminus embraced by V27 and A30. The sum of the Hꢀbonds between the ammonium
groups of the ligands and waters and between the ammonium groups and the N31 side chain amides
was on average three hydrogen bonds for the primary ammonium groups (ligands 1ꢀ5), two
hydrogen bonds for secondary (ligands 6 and 8) and one hydrogen bond for tertiary ones (ligand 7)
(Table S3). Compounds 9 and 10 form three hydrogen bonds only with water molecules in the
vicinity of N31 close to the wall of the pore, possibly because the ammonium group is sterically
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 16 of 66
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
crowded between the adamantyl and cycloalkane rings and cannot reach the side chains carbonyls
of N31 (Table S3). In compound 10, due to the large hydrocarbon framework, the adamantyl ring
moves more than in any other compound toward the Cꢀterminus, i.e., the distance between the
center of mass of the four V27 to the adamantyl cage was 3.7ꢀ2.3 Å longer than 1ꢀ8 to fit inside the
pore close to G34 and the distance between the center of mass of the four G34 to the adamantyl
cage was 2.7 Å. A snapshot of the simulation complex with ligand 10 is depicted in Figure 7.
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Electrophysiology results of aminoadamantanes blockage using fullꢀ
length M2
1
, its isomer 2, and three compounds with sizeable adducts (5, 6, and 9) were assessed for their
ability to inhibit low pHꢀdependent proton currents induced by fullꢀlength M2 protein
(A/California/07/2009 with M2N31) in transientlyꢀtransfected, voltage clamped HEK cells (Table
10a,43
10a
4
).
Data for compounds 1, 2, 5 were reported previously. It was found that none of these
compounds blocked inward proton currents better than 1, either following a standard 3ꢀminute
exposure to compound or after prolonged exposure for 30 minutes. These EPꢀbased results
indicated that the compounds were unable to block the fullꢀlength M2 channel encoding N31.
However, when this M2 protein was modified to encode the 1ꢀsensitive S31 sequence (through an
N31S mutation), all compounds inhibited proton currents with IC50s of 10 µM or less (Table 4); by
increasing adduct size the blocking effect of inward proton currents is reduced.
In vitro testing of aminoadamantanes against influenza A virus
The antiviral potency of the compounds 1ꢀ10 was measured against WSN/33 (H1N1) bearing the
44
26
M2 S31N mutation and its amantadineꢀsensitive variant WSN/33ꢀM2ꢀN31S in MDCK cells with
ACS Paragon Plus Environment

Page 17 of 66
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
4
5,46
CPE inhibitory assay.
Compounds 1ꢀ9 showed subꢀmicromolar IC50 values against WSN/33ꢀ
M2ꢀN31S but were inactive against WSN/33 (Table 5). Compound 10 was inactive against both
strains. The selectivity index (CC50/IC50) of compounds 4ꢀ7 was comparable to 1. The inhibition
efficiency showed stringent headꢀammonium group requirements for inhibition of amantadineꢀ
sensitive influenza A viruses ꢀ a result consistent with ITC and EP results.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
DISCUSSION
Unraveling the binding differences for aminoadamantanes to M2TMS31N and
M2TMWT
Based on the MD simulations results and experimental findings, the molecular basis for weak
binding and the inability of aminoadamantanes to effectively block M2S31N is described. The results
of the experimental data from ssNMR and ITC experiments directly correlate with the MD
simulation results showing that aminoadamantane derivatives are weaker binders in the pore of
M2TMS31N compared to M2TMWT and that 5ꢀ8 are stronger binders compared to 1ꢀ4 against
M2TMS31N
.
The S31N mutation of M2TM results in a shift of the hydrophobic adamantyl ring toward the
Cꢀterminus thereby losing the stabilizing hydrophobic interactions of the V27 isopropyl groups with
the adamantyl ring that is present in the M2TM . The bulky N31 side chains are oriented toward
WT
the Nꢀterminus and the V27 side chains and the ammonium group of the ligands is also turned
toward the Nꢀterminus to form significant hydrogen bonding interactions with the polar N31 side
chains and surrounding waters. This ammonium group orientational preference of amantadine and
17b
16
rimantadine has been previously noted by Gleed et al. and Alhadeff et al. . The hydrogen
bonding interactions with N31 is consistent with the MAS experimental data performed with
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 18 of 66
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
compound 6 showing a chemical shift perturbation for N31 and G34 compared to the apo
M2TMS31N. Distance measurements from ssNMR experiments showed the preference for the
5c
ammonium group of the aminoadamantane drugs orienting toward the Cꢀterminus in M2TMWT
and toward the Nꢀterminus in a complex of M2TMS31N with a conjugate of 1 having a
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
2
4
23bꢀd
phenylisoxazole polar head. In this class of conjugates
additional van der Waals interactions
with the Nꢀterminus stabilizes the ligand resulting in potential antiꢀinfluenza drugs. Here, where the
aminoadamantane ligands, such as 1ꢀ8, are hydrogenꢀbonded with the polar N31 environment,
favorable van der Waals and hydrophobic interactions as those in M2TMWT are missing. In the
M2TMWT the adamantyl ring is well accommodated by the V27 and A30 side chains and sizeable
adducts such as ligands 5ꢀ8 additionally fill the region between A30 and G34 (Figures 1 and 6) but
in M2TMS31N the adamantyl ring is close to A30 and in the vicinity of G34 (see Tables S2, S3)
lacking a favorable hydrophobic pocket. This is consistent with the absence of chemical shift
perturbations for V27 in the NCA MAS spectrum of 6 bound to M2TMS31N in comparison with the
apo M2TMS31N compared to the significant chemical shift changes at V27, S31, G34 which have
20,21
been reported when rimantadine is bound to M2TMWT relative to the apo state.
These structural
differences can be clearly observed in Figures 6 and 7. The lack of favorable van der Waals
interactions results to less stable complexes and a weaker binding for aminoadamantane ligands
d
consistent with the much smaller K values of 5 and 6 by ITC against M2TMS31N compared to
M2TMWT (Table 1). In addition, linewidths are narrower for the M2TMWTꢀcomplex with 5
compared to the M2TMS31N complex suggesting less dynamics consistent with stronger interactions
for the aminoadamantane derivatives in complexes with M2TMWT. The RMSDs for M2TM Cαꢀ
carbons are 1.0ꢀ1.5 Å higher for the trajectories of M2TMS31N compared to M2TMWT complexes
further suggesting more dynamics and weaker interactions in the S31N complexes.
In 1 and analogs with small adducts the adamantyl ring has a only a limited hydrophobic
contact with A30 and is close to polar N31 side chains which exert repulsive forces on adamantyl
ring. It should be noted that trajectories sampled in the MD simulations of ligands in the pore were
ACS Paragon Plus Environment

Page 19 of 66
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
of 80 ns length, i.e., they are much shorter than the microsecond to millsecond time scales sampled
by ssNMR. The MD runs of the complexes of 1ꢀ4 with M2TMS31N showed qualitatively that these
molecules can't bind M2TMS31N because significant favorable van der Waals interactions are
missing. This can be observed from snapshot for complex of 1 with M2TMS31N in Figure 7. In
molecules 5ꢀ8 with sizeable adducts of the adamantyl ring fill effectively the region between A30
and G34 and the interactions needed for binding are slightly improved resulting in weak binding to
M2TMS31N as compared to no binding for 1ꢀ4. This can be observed from snapshots for complexes
of 5, 6 with M2TMS31N in Figure 7. Indeed, 1 did not bind as documented by the OS ssNMR spectra
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
(Figure 3), but larger adducts as those present in compounds 5ꢀ8 appear to stabilize weak binding of
the drug in the region between A30 and G34 (see Figure 7) and this is in accordance with the results
from OS ssNMR spectra for weak binding of compounds 5ꢀ8 (Figures 3, 4) to M2TMS31N. In
addition, we were not able to detect any binding for 1ꢀ4 with M2TMS31N using ITC while we
obtained approximate binding constants for 5,6,9,10 (Table 1). NMR data were not obtained for 9
and 10 because they produced disordered lipid bilayers. Thus, the results from the combination of
simulations and experiments showed that 1 (and similarly 2ꢀ4) did not bind M2TMS31N contrary to
17b
the weak binding for 1 previously suggested from short simulations but 5ꢀ8 having sizeable
adducts display weak binding.
While binding strength of aminoadamantanes against M2TMWT was sensitive to modification
of the adduct, the M2TM ꢀaminoadamantane complex stability was found to be very sensitive to
WT
the adduct size, as shown by the ITC and MD simulation results when considering 9 and 10 that
produces unstable complexes. The MD simulations clearly show that 10 moves considerably toward
the Nꢀterminus of the pore losing specific binding interactions. Ligands 9 and 10 cause a dramatic
reduction in affinity for M2TMWT compared to 1ꢀ8 but did not affect M2TMS31N affinity, which
seems to be similar with that of 5 and 6. The forces that cause the rotation of the adamantyl ring in
the M2 pore appear to be inherent in the shared amine group and are not greatly perturbed by the
other ligand variations as characterized by MD and OS ssNMR for adducts 1ꢀ4, 5ꢀ8 against
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 20 of 66
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
M2TMS31N. Taken together, the results from the combination of MD simulations, ITC and OS
ssNMR showed no binding for 1 and similar in size analogues and only weak binding for sizeable
adducts. The binding is more specific as showed by the more stringent headꢀammonium group
requirements in the binding pocket of M2TMWT than in the M2TMS31N according to the ITC but
also as showed by the EP experiments and antiꢀviral assay results.
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Thus, EP experiments indicate that the aminoadamantanes block the S31 but not the N31 fullꢀ
11
length M2 protein. Notably, while the secondary gate formed by the V27 residues in the M2TMWT
has the potential to limit water access to the pore the hydrophilic asparagine sidechains make this
environment less hydrophobic and diminish the effectiveness of the V27 gate in the M2TMS31N
pore. In the mutant waters are observed above and below the ligand and in a few snapshots between
the ligand and the wall of the pore suggesting a relatively free passage through the M2TMS31N
12,17,18
lumen despite the presence of the ligand.
(snapshots from the simulation of the complexes of
M2TMS31N with 1, 6, 10 are depicted in Figure 7).
Changes in Cꢀterminus structureꢀfunction of M2TMS31N and M2TMWT induced
by V27 interactions with the adamantyl cage of amantadine variants
Furthermore it is suggested that the 10º helical kink that reduces the tilt of the Cꢀterminal portion of
the transmembrane helix in the WT is likely induced by the formation of a strong binding pocket for
the aminoadamantanes. This hydrophobic pocket formed by the V27 aliphatic sidechains coupled
with the aliphatic adamantyl cage of the aminoadamantanes may prevent the W41 gate from
opening further defeating proton conductance. In contrast the aminoadamantanes in M2TMS31N
included a loss of V27 lipophilic pocket and have their amino group drawn toward the Nꢀterminus
by the asparagine sidechains resulting in the presence of multiple water molecules in this region. In
addition the weak binding of the amino adamantanes results in no perturbation of the helical tilt and
the W41 gate can function normally.
ACS Paragon Plus Environment

Page 21 of 66
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
CONCLUSIONS
This work represents a study of the binding of amantadine variants against the proton channel
formed by the tetrameric structure of the influenza A M2 protein. Significantly, we focus on
aminoadamantane variants of 1 binding to M2TMS31N compared to M2TMWT aiming at
investigating why these variants are ineffective in blocking proton conductance of the M2S31N
channel. The results of this effort are based on a combination of experimental techniques and MD
simulations both performed in liquid crystalline lipid bilayer environments. Aminoadamantane
derivatives are known to be blockers of the M2 WT protein. They are known to bind in the pore and
are presumed to block proton access to the H37 tetrad that is known to shuttle protons through
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
47
aqueous pore into the viral interior. There are two gates that can inhibit conductance, the V27
tetrad, known as the secondary gate at the external entrance to the pore and the W41 tetrad near the
exit of the pore into the viral interior, known as the primary gate. These aminoadamantane
derivatives were weaker binders to M2TMS31N compared to M2TMWT as observed by a reduced
influence on the protein structure and by reduced amplitude of the channel dynamics as observed by
both the MD and experimental data. Moreover, these aminoadamantane derivatives were ineffective
against M2S31N while blocking M2WT protein. We suggest that 1 and the similar sized analogs 2ꢀ4
lack of binding affinity and the larger sized analogs 5ꢀ8 showed weak binding affinity to M2TMS31N
because of a lack of effective hydrophobic interactions as a result of reshaping the cavity when N31
is present which included loss of V27 lipophilic pocket. In contrast V27 interactions are present
with 1ꢀ8 in M2TMWT pore and these ligands are effective binders to M2TMWT. All ligands 1ꢀ8 have
a tight binding for M2TMWT and the binding is more specific as showed by the more stringent headꢀ
ammonium group requirements in the binding pocket of M2TMWT than in the M2TMS31N according
.
to the ITC, EP and CPE results The weak binding of 5ꢀ8 to M2TMS31N was significant enough to
induce observable changes in the helix tilt angles characterized by the experimental data and by the
MD simulations.
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 22 of 66
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
It is interesting to note that the blockage of M2TMWT by both 1 and 5 involved a 10° kink in
the TM helix and a very significant change in the helix orientation in the Cꢀterminal half of the TM.
This is likely due to the V27 side chains ꢀ adamantane hydrophobic interactions that are not possible
in M2TMS31N. While aminoadamantane ligand binding causes a kink in the Cꢀterminal half of
M2TMWT and a blockage for proton conductance by the M2 channel, aminoadamantane ligand
interactions with M2TMS31N did not result in a helix kink in the TM helix and proton conductance
was not blocked. While the helix kink has been associated with blockage and potential disabling the
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
4
8
opening of W41 gate in M2TMWT previously here, the explanation for the helix kink in the
M2TMWT and its absence in M2TMS31N has been suggested to be induced by the V27 interactions
with the adamantyl cage in the M2TMWT and the absence of such significant interactions in the
M2TMS31N
.
ACS Paragon Plus Environment

Page 23 of 66
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
EXPERIMENTAL METHODS
1 Synthesis of aminoadamantane ligands
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
The procedures applied for the synthesis of the new derivatives 8 and 10 are depicted in Schemes 2
and 3. The description of synthetic procedures for 8 and 10, experimental details and compound
characterization can be found in the Supporting Information. All compounds purity was ≥ 95% as
determined by elemental analysis (see Supporting Information).
2
M2TM peptide synthesis
M2TMWT peptides corresponding to residues 22ꢀ46 of the Udorn (A/Udorn/307/72) sequence of M2
Cꢀterminally amidated SSDPLVVAASIIGILHLILWILDRL) and of the S31N mutant peptide (Cꢀ
(
terminally amidated SSDPLVVAANIIGILHLILWILDRL) were synthesized by standard Fmoc (9ꢀ
fluorenylmethoxycarbonyl) solid phase peptide synthesis using an aminomethyl polystyrene resin
loaded with the amide linker and purified by reverse phase HPLC before used for the ITC
experiments. Additional quantities of these peptides needed were purchased from Centic Biotec,
15
Heildeberg, Germany. For ssNMR experiments M2TMWT and M2TMS31N (22ꢀ46) peptides with N
13
15
labeled at V28, A30 and I42 and M2TMS31N (22ꢀ46) peptides with C, N labeled at structurally
4
a,5b
13 15
important residues V27, A30
and N31, G34 ( C, NꢀVANG) were synthesized using Fmoc
15
15
15
13 15
chemistry. Fmocꢀ[ N]ꢀVal, Fmocꢀ[ N]ꢀAla, Fmocꢀ[ N]ꢀIle and Fmocꢀ[ C, N]ꢀVal, Fmocꢀ
13 15
13 15
13 15
[
C, N]ꢀAla, Fmocꢀ[ C, N]ꢀSer and Fmocꢀ[ C, N]ꢀGly were purchased from Cambridge
Isotope Laboratory (Andover, MA). Solidꢀphase syntheses of M2TM peptides (0.25 mmol) were
49,50
performed on an Applied Biosystems 430A peptide synthesizer as previously described.
The
peptide was cleaved from the resin by the treatment with ice cold 95 % TFA, 2.5 % H
2
O, 1.25 %
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 24 of 66
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
ethanedithiol, 1.25 % thioanisole and precipitated from TFA using ice cold ether. Following
centrifugation, the supernatant was discarded and the pellet was washed with cold ether again. The
51
precipitated peptide was dried under vacuum. A purification procedure previously described and
5
0
modified was used. Peptide purity and identity was confirmed using ESI mass spectrometry
(positive ion mode). The final peptide purity was 98 %.
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
3
ITC measurements of aminoadamantane ligands binding to M2TM
Binding affinities of the aminoadamantane derivatives 1ꢀ5, 6, 9, 10 for M2TMS31N were measured
52,53
by ITC experiments
in DPC micelles at pH 8. Compounds 7, 8 were not measured. M2TM
54
fragments form stable tetramers at this pH, in contrast to low pH (< 6.5) conditions. Furthermore,
experimental data indicate that 1 binds with higher affinity at alkaline pH to M2TM, where the pore
of the M2 channel is in the closed state, than at low pH, where the open state of M2TM is
5
4a,2a
prevalent.
All measurements were performed with a TAM 2277 (TA instrument) at pH 8 and 20 °C in a
buffer of 50 mM NaH
2
PO
4
and 100 mM NaCl. The peptide and the aminoadamantane derivative
ꢀ1
were dissolved in a freshly prepared DPC solution with a concentration of 13 mmol L .
Measurements against M2TMWT were conducted using 2 mL of 125 µM peptide (corresponding to
2
6
3
1.25 µM M2TMWT tetramer) as has been described previously. For the low affinity ligands 9 and
10 a 250 ꢁΜ peptide concentration was used. For M2TMS31N measurements the concentration of the
peptide used was 167 ꢁΜ, which was increased to 500 ꢁΜ when 5 was tested in order to get a curve
adequate for measurements. A concentration of 1.1 mM of the ligand was used for the titrant, of
which 7.6 µL (equivalent to 8.4 nmol) were dispensed in the peptide/DPC solution with each
injection. The time interval between two injections was set to at least 6 minutes. First injection
wasn’t used due to dilution effects.
ACS Paragon Plus Environment

Page 25 of 66
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
Synthetic M2TMS31N (residues 22ꢀ46) was reconstituted in DPC micelles at pH 8 at a 1:40
54
monomer/DPC ratio ꢀ which guarantees the quantitative formation of M2TM tetramers ꢀ by
dissolving and sonicating 334 nmol of M2TM with the 40ꢀfold amount of DPC in the
aforementioned buffer system (for M2TMWT a 1:26 ratio was applied for the measurements of
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
2
6
ligands 9, 10 compared to 1:57 applied previously ). Solutions of the ligands in the buffer were
titrated into the calorimetric cell at 20 °C. The released heat of binding was derived by subtracting
55,56
the heat of dilution from the heat of reaction.
Describing in more detail, to determine the heat of dilution ligand as well as receptor (M2TM)
were dissolved in buffer containing DPC. Two reference experiments were carried out, i.e., (a)
ligand was titrated into the buffer containing DPC and (b) receptor was titrated into buffer
containing DPC. The obtained signals showed, that the released heat of each injection is very small
and stays constant during the experiment. No interaction is detected between ligand and DPC
containing buffer or receptor and DPC containing buffer, respectively. The released heat is the heat
of dilution, which occurs when the two solutions are mixed. The heat of dilution was estimated for
each titration experiment individually. At the end of each measurement, the receptor is titrated with
the ligand being in an excess (the ratio for ligand/receptor after 19 injections is 2:1). If all binding
sites are saturated, ligand can't bind to the receptor anymore. The released heat is then equivalent to
the heat of dilution. As mentioned the heat of dilution is calculated as an average from the last five
titration steps, when the heat is constant. This heat is take into account as Q correction. It is
subtracted from all titrations steps in this measurement.
Data evaluation, including the integration of the peaks, was carried out with Digitam for
Windows v4.1. The measured heat per amount of substance against the molar ratio of titrant to
peptide tetramer was plotted and the affinity constants were calculated by nonꢀlinear regression of
57
the measured heat per injection using Origin 8.0 and are included in Table 1. Compounds 1ꢀ8 have
26
been measured against M2TMWT in a previous work (Table S1). For the calculation, the
concentration of the peptide was kept variable because the M2TM tetramer formation is not
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 26 of 66
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
complete. The fit function involves three parameters. One of them is a factor for the correction of
the peptide concentration (the difference between the concentration, which has been weighted in,
and the active concentration). The cell volume is fixed at 2 mL. The concentration of the ligand in
the solution is known, because pure substance was weighted in and the stoichiometry of the binding
of ligand to receptor was assumed to be 1:1. If these quantities are set, the concentration of active
receptor is obtained by fitting the measured data points. From the calculations, it can be seen, that
the concentration of active receptor is lower than the concentration of receptor that has been
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
2
6
weighted in. The binding affinity of 1 measured against M2TMWT in a previous work was 2.17 ±
0
.52 ꢁM and is comparable with the value of 12 ꢁM measured using analytical
54c
ultracentrifugation
and the value of 9 ± 2 ꢁM derived based on kinetic studies in
2a
electrophysiological experiments. For the M2TM peptide investigated in this study, the solubility
in the DPC micelles limits the concentration of M2TM that can be tested. Consequently, affinity
constants of low affinity binders, e.g., aminoadamantanes against M2TMS31N and 9, 10 against
M2TMWT possess relatively large errors.
4
Sample preparation for solid state NMR
1
5
NꢀV28
A
30
I42 M2TMWT or M2TMS31N was coꢀdissolved in trifluoroethanol (TFE) with DMPC in a
1
:30 molar ratio. (The molar ratio of 1 protein tetramer to 120 DMPC lipids was used. The
molecular weight of the M2TMWT peptide is MW = 2729 g/mole and the lipid is MW = 678
g/mole). The solvent was removed under a stream of nitrogen gas to yield a lipid film, and then
dried to remove residual organic solvent under vacuum for 12 hours. Thoroughly dried lipid film
was hydrated with 10 mM HEPES buffer at pH 7.5 to form multilamellar vesicles containing
M2TM in tetrameric state. This suspension was bath sonicated, dialyzed against 2L HEPES 10 mM
pH 7.5 buffer for 1 day and centrifuged at 196,000xg to harvest unilamellar proteoliposomes. The
pellet was reꢀsuspended in a 1 mL aliquot of the decanted supernatant containing the ligand,
ACS Paragon Plus Environment

Page 27 of 66
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
resulting in a 1:6 molar ratio of the M2TM tetramer to drug. Typically a preparation for solid state
NMR would include 5 mg of protein and 37 mg of DMPC. For samples that included drug the
w
molar ratio of drug to tetramer was 6:1. For 1 with a molecular weight of M = 187.5 g/mole that
would mean the addition 0.52 mg for a 5 mg protein sample (There is sufficient evidence in the
literature to demonstrate that M2 structure is not perturbed by drug concentration. The most obvious
example in this manuscript, is the absence of structural changes in the S31N spectra in the presence
of 1) Following overnight incubation at 37°C, the pellet was deposited on 5.7x10mm glass strips
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
(Matsunami Trading, Osaka, Japan). The bulk of the water from the sample was removed during a
two day period in a 98 % relative humidity environment at 298 K. Rehydration of the slides, before
stacking and sealing into a rectangular sample cell, increased the sample weight by 40ꢀ50 %.
Compounds 1, 5 were used for ssNMR experiments against M2TMWT; compounds 1, 5ꢀ8 were used
for ssNMR experiments against M2TMS31N; compounds 9 and 10 produce disordered lipid bilayers
31
according to the P spectra (not shown).
5
Solid state NMR experiments of M2TMꢀaminoadamananes
complexes
OS ssNMR spectra
2
8
29,30
1
15
PISEMA and SAMPI4
spectra were acquired at 720 MHz utilizing a lowꢀE H/ N double
58
resonance probe. Acquisition took place at 303 K, above the gel to liquid crystalline phase
transition temperature of DMPC lipids. Experimental parameters included a 90° pulse of 5 µs and
crossꢀpolarization contact time of 0.8ꢀ1 ms, a 4 s recycle delay and a SPINAL decoupling
5
8
15
sequence. 32 t
compounds 1, 5 and nine t
increments were obtained for the spectrum of NꢀV28
1
increments were obtained for the spectrum of NꢀV28
A
30
I
42 M2TMWT with
42 M2TMWT without drug. 16ꢀ28 t
30
A I42 M2TMS31N with compounds 1, 5ꢀ8 and
15
1
for the sample of NꢀV28
A
30
I
1
15
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 28 of 66
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
1
5
nine t
1
increments for the sample of NꢀV28
A
30
I
42 M2TMS31N without drug. Spectral processing
5
9
15
was done with NMRPIPE and plotting with SPARKY. N chemical shifts were referenced to a
2 8 4
concentrated solution of N H SO , defined as 26.8 ppm relative to liquid ammonia.
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
NCA MAS spectra
15
13
Nꢀ Cα correlation experiments were performed on a Bruker Avance 600 MHz NMR
60,61
13
Spectrometer with an NHMFL 3.2 mm lowꢀEꢀfield triple resonance probe.
The C chemical
6
2
15
shifts were referenced using the published chemical shifts of adamantane relative to DSS and N
13
chemical shifts were calculated with IUPAC relative frequency ratios between the DSS ( C) and
15
63,64
liquid ammonia ( N).
2 kHz and a calibrated sample temperature of ꢀ10 °C. 30 points were collected in the
dimension for an acquisition time of 5ꢀ6.25 ms, while in the direct dimension the acquisition time
Spectra were acquired at magic angle spinning (MAS) frequency of 10ꢀ
15
1
N
1
5
13
was 10.2 ms. 92 kHz of proton decoupling was used in all experiments. To get one bond Nꢀ C
correlation a mixing time of 5 ms was used. Spectra were processed with Topspin.
6 MD simulations of M2TMꢀaminoadamantane complexes
Docking calculations
65
The ligands in their ammonium forms were built by means of Maestro 8.5 and were then
66
minimized by means of Macromodel 9.6 and the MMFF94 force field implemented with
Macromodel 9.6 using the CG method and a distanceꢀdependent dielectric constant of 4.0 until a
ꢀ1
ꢀ1
convergence value of 0.0001 kJ Å mol was reached. The M2TM ꢀ1 complex structure (PDB ID
WT
4
a,5b
2
KQT ) served as a model structure for M2TMWT with bound ligands. Nꢀ and Cꢀtermini of the
M2TM model systems were capped by acetyl and methylamino groups. After applying the protein
preparation module of Maestro, all hydrogens of the protein complex were minimized with the
ACS Paragon Plus Environment

Page 29 of 66
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
AMBER* force field by means of Maestro/Macromodel 9.6 using a distanceꢀdependent dielectric
constant of 4.0. The molecular mechanics minimizations were performed with a conjugate gradient
ꢀ1
ꢀ1
(
CG) method and a threshold value of 0.0001 kJ Å mol as the convergence criterion. The
structures of the protein and ligand 1 were saved separately and were used for the subsequent
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
docking calculations. The ligands minimized in this manner were docked into the M2TMWT binding
site. Docking poses of the aminoadamantane derivatives 1ꢀ10 in the M2TMWT bound state were
67,68
generated with GOLD 5.2
considering five water molecules located between ammonium group
of 1 and H37 within the M2TMWT poreꢀbinding site and applying the ChemPLP implemented in the
69,70
software.
The option “toggle” was used to let the algorithm decide whether taking into account a
water molecule or neglecting it based on an empirical desolvation penalty. The region of interest
used by GOLD was defined to contain the atoms that were within ∼15 Å of 1 binding site in the
receptor structure. The “allow early termination” command was deactivated. For all the other
parameters, GOLD default values were used. Ligands were submitted to 30 genetic algorithm runs.
Ten docking poses were produced for each ligand which were visually inspected using the UCSF
71
Chimera package. The docking pose with the best ChemPLP score was used for the subsequent
MD simulations with M2TMWT and M2TMS31N structures created as described below.
MD simulations
Models of M2TMS31Nꢀaminoadamantane complexes were generated from M2TMWT
65
aminoadamantane complexes by mutating amino acids S31 to N31 with Maestro and preparing the
structure as described above i.e. Nꢀ and Cꢀtermini of the M2TM peptides were capped by acetylꢀ
1
and methylamino groups, respectively. For N31, the side chain rotamers may have χ angles ꢀ160°
or ꢀ80° corresponding to N31 side chains placed at the interface between helices or inside the
lumen respectively. Structures of M2TMS31N(18ꢀ60) in DPC micelles solved by solution NMR
7
2
spectroscopy show residue 31 in the helix−helix interface while MAS ssNMR studies in 1,2ꢀ
diphytanoylꢀsnꢀglyceroꢀ3ꢀphosphocholine (DPhPC) bilayers showed the side chain of the N31
ACS Paragon Plus Environment