Investigation of the molecular characteristics of bisindole inhibitors as HIV-1 glycoprotein-41 fusion inhibitors

Article abstract of DOI:10.1016/j.ejmech.2018.10.048

In previous work, we described 6-6‘-bisindole compounds targeting a hydrophobic pocket on the N-heptad repeat region of viral glycoprotein-41 as effective inhibitors of HIV-1 fusion. Two promising compounds with sub-micromolar IC₅₀'s contained a benzoic acid group and a benzoic acid ester attached at the two indole nitrogens. Here we have conducted a thorough structure-activity relationship (SAR) study evaluating the contribution of each of the ring systems and various substituents to compound potency. Hydrophobicity, polarity and charge were varied to produce 35 new compounds that were evaluated in binding, cell-cell fusion and viral infectivity assays. We found that (a) activity based solely on increasing hydrophobic content plateaued at ～ 200 nM; (b) the bisindole scaffold surpassed other heterocyclic ring systems in efficacy; (c) a polar interaction possibly involving Gln575 in the pocket could supplant less specific hydrophobic interactions; and (d) the benzoic acid ester moiety did not appear to form specific contacts with the pocket. The importance of this hydrophobic group to compound potency suggests a mechanism whereby it might interact with a tertiary component during fusion, such as membrane. A promising small molecule 10b with sub-μM activity was discovered with molecular weight <500 da and reduced logP compared to earlier compounds. The work provides insight into requirements for small molecule inhibition of HIV-1 fusion.

Full text of DOI:10.1016/j.ejmech.2018.10.048

European Journal of Medicinal Chemistry 161 (2019) 533e542
Contents lists available at ScienceDirect
European Journal of Medicinal Chemistry
journal homepage: http://www.elsevier.com/locate/ejmech
Research paper
Investigation of the molecular characteristics of bisindole inhibitors as
HIV-1 glycoprotein-41 fusion inhibitors
Guangyan Zhou a, Shidong Chu a, Ariana Nemati a, Chunsheng Huang c, Beth A. Snyder c,
, b, *
Roger G. Ptak c, Miriam Gochin a
a Department of Basic Sciences, Touro University-California, Vallejo, CA, 94592, USA
b Department of Pharmaceutical Chemistry, University of California San Francisco, CA, 94143, USA
c Southern Research Institute, 431 Aviation Way, Frederick, MD, 21701, USA
a r t i c l e i n f o
a b s t r a c t
Article history:
In previous work, we described 6-6‘-bisindole compounds targeting a hydrophobic pocket on the N-
heptad repeat region of viral glycoprotein-41 as effective inhibitors of HIV-1 fusion. Two promising
compounds with sub-micromolar IC50's contained a benzoic acid group and a benzoic acid ester attached
at the two indole nitrogens. Here we have conducted a thorough structure-activity relationship (SAR)
study evaluating the contribution of each of the ring systems and various substituents to compound
potency. Hydrophobicity, polarity and charge were varied to produce 35 new compounds that were
evaluated in binding, cell-cell fusion and viral infectivity assays. We found that (a) activity based solely
on increasing hydrophobic content plateaued at ~ 200 nM; (b) the bisindole scaffold surpassed other
heterocyclic ring systems in efficacy; (c) a polar interaction possibly involving Gln575 in the pocket could
supplant less specific hydrophobic interactions; and (d) the benzoic acid ester moiety did not appear to
form specific contacts with the pocket. The importance of this hydrophobic group to compound potency
suggests a mechanism whereby it might interact with a tertiary component during fusion, such as
Received 22 August 2018
Received in revised form
18 October 2018
Accepted 19 October 2018
Available online 23 October 2018
Keywords:
HIV-gp41
Fusion inhibitor
Bisindole
Suzuki coupling
Sub-mM activity
membrane. A promising small molecule 10b with sub-mM activity was discovered with molecular weight
<500 da and reduced logP compared to earlier compounds. The work provides insight into requirements
for small molecule inhibition of HIV-1 fusion.
© 2018 Elsevier Masson SAS. All rights reserved.
1. Introduction
binding inhibitors. These inhibitors block the interaction between
two protein domains of gp41, the NHR trimer containing the hy-
Previous studies by our lab and others on small molecule inhi-
bition of HIV-1 fusion have revealed the importance of hydrophobic
groups and negative charge for measurable bioactivity [1e4]. In-
hibitors bind to a hydrophobic cavity within the ectodomain of
transmembrane glycoprotein gp41 that is exposed during fusion
[5], and potency is generally correlated to pocket binding affinity
[2,6]. The cavity is defined by N-heptad repeat (NHR) residues
565e579 on two chains of the gp41 homotrimer (HXB2
numbering). It is lined by several non-polar branched chain amino
acids and a nearby tryptophan, but also includes polar residues Thr
and Gln, as well as positively charged Lys and Arg residues that
likely lead to the requirement for negative charge on pocket-
drophobic pocket and gp41 C-heptad repeat (CHR) helices that bind
in the pocket following a gp41 conformational change during
fusion [7]. The specific pocket binding domain of the CHR is Trp-x-
x-Trp-Asp-x-x-Ile where Trp and Ile residues intercalate within the
pocket and Asp forms a salt bridge with the pocket Lys [8].
Studies of protein-protein interaction inhibitors (PPI) have
generally revealed that they are more hydrophobic, more aromatic
and larger than drugs targeting enzymes [9e11]. Analysis of recent
small molecule PPI modulators currently in the clinic revealed a
broad range of molecular sizes and complexity, particularly with a
logP range of 1.6e10.5 [12]. Efficiency of binding to the hydrophobic
patch at the core of the protein eprotein interaction surface was
cited [11] as a key determinant of potency. In previous studies, we
identified a promising scaffold containing two 6-60 connected
indole rings B and C supporting substituted benzyl rings A and D at
positions 1 or 3 on the indole rings (Fig. 1) [2,6,13]. A structure-
* Corresponding author. Touro University - California, 1310 Club Drive, Mare Is-
land, Vallejo, CA, 94592, USA.
E-mail address: miriam.gochin@tu.edu (M. Gochin).
https://doi.org/10.1016/j.ejmech.2018.10.048
0223-5234/© 2018 Elsevier Masson SAS. All rights reserved.

534
G. Zhou et al. / European Journal of Medicinal Chemistry 161 (2019) 533e542
Fig. 1. Bisindole scaffold structure. Compounds 1e4 are shown, and their equivalents in the current study.
activity relationship (SAR) study upheld the contention that hy-
drophobicity played an essential role in potency, as did having a
negative charge on the ligand. The following key features were
observed: 1) A free carboxylic acid on ring A was indispensable for
activity. 2) Ring D conferred an order of magnitude improvement in
antiviral potency compared to three ring (A-B-C) compounds, with
the m-benzoic acid ester providing optimal performance (com-
points, or substitute polar moieties in order to sustain or increase
solubility and potentially introduce hydrogen bonds. Modification
strategies were based in part on docking studies predicting a narrow
S-shaped structure for the compounds, conforming to the shape of
the hydrophobic groove [2]. Fig. 2 shows the array of polar and non-
polar residues lining the pocket with an overlay of several docked
structures described in the current study. We note that the modeling
is limited by a lack of complete knowledge of HP structure [14]. We
examined 19F substitution on different rings, made additional sub-
stitutions on ring D, replaced indoles B and/or C with chemically
similar alternatives, and varied the linkage between rings C and D.
Overwhelmingly the results confirmed the bisindole scaffold as the
optimal choice, and showed that simply increasing hydrophobic
interactions did not yield a significant increase in potency. The SAR
indicated that indole B interactions with pocket residues contributed
more than those of indole C, suggesting that B is buried more deeply
within the pocket. The methylene linker between ring C and D could
be substituted with a sulfoxy or carbonyl group, suggesting that
polar contacts could occur with pocket residues. A three-ring com-
pounds 1, 2 Fig. 1, IC50 ¼ 0.3e0.4
also being favorable (compound 3, Fig. 1, IC50 ¼ 0.8
10 substitution was preferred over 3, 30 substitution (compound 4,
Fig. 1, IC50 ¼ 5.0 M), as was indole over benzimidazole [2], sug-
mM), and the m-methoxy group
mM). 3) Indole 1,
m
gesting that the hydrophobic surface presented by the outer
vertices of rings B and C was critical for activity. 3) Variation of non-
polar substituents on ring D had a mild effect on potency (~2-fold),
while a charged substituent on ring D such as carboxylic acid
yielded two orders of magnitude reduced antiviral potency, even
though binding affinity for the hydrophobic pocket (HP) did not
diminish. The in vivo effect suggested that membrane could play a
role in bioactivity. The deleterious effect caused by adding more
than one negative charge might be attributed to polar head groups
of phospholipids repelling a negatively charged ligand. In this
sense, the protein-protein domain interaction may be complicated
by proximity of the pocket to membrane. And, while increasing
hydrophobicity could increase interactions within the pocket, it
might also render the molecules less soluble and potentially non-
specific. Thus a delicate balance of hydrophobicity and charge is
required to maximize the inhibitory potential of the ligands.
Here we set out to explore this balance by making additional
changes to the scaffold to increase or decrease hydrophobic contact
pound 10b was identified as the first example of a sub-
mM inhibitor
with a molecular weight <500 da.
2. Results
2.1. Effect of fluorine substitution
Fluorine is a small atom with high electronegativity, slightly
larger than a hydrogen atom. It has been introduced into medicinal
candidates not only to improve metabolic stability but also binding
affinity to the target protein. Fluorine substitution could lead to a
slightly increased nonspecific lipophilicity [15,16]. The effect of
fluorine substitution on the intrinsic potency of the bisindole
compounds is shown in Table 1. Introduction of fluorine on rings A,
C and/or D reduced potency by up to 10-fold, compared to the
corresponding hydrogen counterparts. We observed two cases in
which a single fluorine substitution at any position had a significant
deleterious effect, while a subsequent second fluorine substituent
had little additional effect on potency. For example, in order of
potency, 11a > 11h z 11i; 9a [ 9o z 9p. However, a more varied
effect was observed for other compounds. A single fluorine at the 5
position of indole C or at the meta position of ring D on 9a left
activity slightly lower but still at sub-mM levels (15d, 9d). The same
5-F-indole substitution on compound 3, however, led to a 10-fold
reduction in potency (9q). KI for HP binding was correlated to
observed antiviral activity for the compounds in Table 1, (R2 ¼ 0.80
(CCF); R2 ¼ 0.58 (VCF)). The drop in potency upon fluorine substi-
tution may be a result of decreased solubility associated with
increased hydrophobicity. Among the five variants of 9a in Table 1, a
strong correlation (R2 ¼ 0.93) was observed between increase in
logP and loss of potency in cell-based assays. The effect was more
pronounced when starting with compounds having a higher logP
(9a and 9n vs. 11a). Therefore we observed that fluorination did not
improve potency in the case where compounds were already highly
hydrophobic.
Fig. 2. Predicted conformation of several ligands in the current study in the hydro-
phobic pocket of PDB structure 2xra. Residue numbers in the construct are shown
together with their corresponding HXB2 Env numbers in brackets. The two chains
making up the binding site are distinguished by E and A in the 2xra construct or by
using a prime to mark the second chain in HXB2 numbering. Ligands shown are 9g,
10b, 9c, 15b, 9j, 11g.

G. Zhou et al. / European Journal of Medicinal Chemistry 161 (2019) 533e542
535
Table 1
Testing fluorine substitutions on bisindole compounds.
logPa
log Sb
KI (HP)c
IC50 CCFc
IC50 VCF Ba-Lc
IC50 VCF IIIBc
CC50
c
X
Y
Z
MWt
11i
F
F
H
H
F
CH3
CH3
CH3
416.4
398.4
380.4
532.6
6.84
6.61
6.42
7.88
ꢀ8.03
ꢀ7.63
ꢀ7.42
ꢀ9.84
2
3
1.4
0.7
2
9.7
8.2
4.5
0.5
7.8
9.1
4.3
0.4
>20
>20
>50
>20
11h
11a
15d
H
H
F
3.6
1.3
0.4
d
d
9a
H
H
H
H
514.6
532.6
7.88
7.51
ꢀ10.12
ꢀ8.77
0.6
1
0.2
0.8
0.4
0.4
0.3
0.5
>20
>20
9d
9o
9p
9q
H
F
H
H
540.5
558.5
9.10
9.62
ꢀ10.13
ꢀ11.34
2.3
1.7
1.4
2.1
1.4
1.4
1.5
2.0
>20
>20
H
H
F
504.6
486.6
8.23
8.11
ꢀ9.30
ꢀ9.20
5
1
3.5
0.8
9.4
0.8
7.0
0.9
>20
>25
d
9n
H
a
Calculated water-octanol partition coefficient.
Calculated solubility.
b
c
All values in
m
M.
d
Previously published in Ref. [2].
2.2. Examination of substituents at group D
a series of amine substitutions. Compounds 9i, 9j, 9k, 9l were much
more active than 9g. Three of them had lower activity
Previous observation of the important contribution of benzyl
ring D to antiviral potency, as well as the loss of potency associated
with addition of a negatively charged substituent on D spurred us
to consider additional substituent changes at this location. The best
performing compounds contained a carboxylic acid ester substit-
uent, although this position is exposed to solvent in predicted
binding orientations (Fig. 2). We examined additional molecules to
confirm this effect. Table 2 shows the results. The t-Bu ester on ring
D (compound 9c) has increased hydrophobicity compared to 9a
(methyl ester) and 9b (ethyl ester) and showed marginally
(IC50 ¼ 1.2e4.7
(IC50 ¼ 0.5 M), marginally lower than that of 9a. Switching D to a
saturated six-membered piperidine ring did not significantly
M). The sub- M activity
of 9l and 9r may in part be attributed to the hydrophobic t-Bu
carboxylate ester associated with added benefit in 9c.
For the compounds in Table 2, the KI for solution binding did not
correlate well to IC50, driven by the discrepancy observed for the
eCOOH e substituted and to some extent the eNH- substituted
compounds Therefore the binding assay did not fully recapitulate
the in situ mechanism, implying that other biological components
may be involved. It is possible that polarity due to an NeH or C]O
group, together with hydrophobic esterification, supports insertion
of ring D into anionic phospholipid membrane or an anionic lipo-
philic protein domain. This would explain the detrimental effect of
a negatively charged substituent on D and the failure of the solution
binding experiment to correctly predict activity of 9g and 9h.
mM) than 9a. One of them, 9l, had sub-mM activity
m
impact activity (compound 9r, IC50 ¼ 0.6
m
m
improved antiviral activity (IC50 ¼ 0.2
mM). Increasing hydrophobic
bulk by adding a second m’-carbomethoxy group to form 9e slightly
VCF
reduced antiviral activity (IC
¼ 0.6
m
M), as did adding a m’-F
M). Adding the m-carbomethoxy
group to a p-OMe substituted ring D had an even more pronounced
50
VCF
substituent in 9d, (IC
¼ 0.5
m
50
VCF
negative effect (9f, IC
¼ 3e4
mM). Thus added hydrophobic bulk
50
at ring D was not tolerated. We previously observed that a nega-
tively charged m-carboxylate was highly deleterious (compound
VCF
9g, IC
¼ 60
mM) [2], and confirmed this effect with a new com-
50
2.3. Adjustment of the linker between groups C and D
pound having both a m-carboxylate and m’-methoxy group (com-
pound 9h). This compound had 15-fold lower activity against CCF
compared to 9n and no measurable antiviral activity up to 20
These two compounds both exhibited sub- M binding affinity to
the HP, similar to that of compounds 9a, 9b and 9n.
We further explored the polarity at ring D with a nitro group and
Due to the proximity of Gln 5750 to the predicted pose of the
compounds, the potential for a polar interaction was tested by
converting the methylene of the benzyl group of D into sulfonyl or
carbonyl. In the eCO- containing compound, the phenyl group was
replaced entirely by a t-Bu ester, which proved to be successful as a
mM.
m

536
G. Zhou et al. / European Journal of Medicinal Chemistry 161 (2019) 533e542
Table 2
Testing substituents at ring D.
logPa
log Sb
KI (HP)c
IC50 CCFc
IC50 VCF Ba-Lc
IC50 VCF IIIBc
CC50
c
X
Y
MWt
d
9a
9b
9c
9d
9e
9f
9g
9h
9i
-COOCH3
-COOCH2CH3
-COOC(CH3)3
-COOCH3
-COOCH3
-COOCH3
-COOH
-H
-H
-H
-F
514.6
528.6
556.7
532.6
572.6
544.6
500.6
548.6
501.5
471.6
513.6
571.7
563.7
7.88
8.13
8.36
7.51
7.45
7.82
7.40
7.44
7.33
7.11
7.46
8.53
8.24
ꢀ10.12
ꢀ9.90
ꢀ9.21
ꢀ8.77
ꢀ10.11
ꢀ9.27
ꢀ9.14
ꢀ9.09
ꢀ9.22
ꢀ8.81
ꢀ9.45
ꢀ10.79
ꢀ7.79
0.6
1.0
0.3
1.0
1.0
0.4
0.6
0.8
2.7
3.1
0.9
1.5
1.2
0.2
0.7
0.2
0.8
0.2
2
12
8
2
3
1
2
1.2
0.4
0.3
0.2
0.4
0.6
4.2
58
>20
2.8
1.1
4.7
0.4
0.6
0.3
0.5
0.2
0.5
0.6
3.1
67
>20
3.4
1.3
4.5
0.5
0.6
>20
>40
>10
>20
>20
~25
>100
>20
>20
>20
>20
>20
>20
d
-COOCH3
-OCH3 (p)e
d
-H
-COOCH3
-H
-H
-H
-H
-COOH
-NO2
9j
9k
9l
-NH2 (p)a
-NH-COCH3
-NH-COOC(CH3)3
9r
a
Calculated water-octanol partition coefficient.
Calculated solubility.
All values in mM.
Previously published in Ref. [2].
Para substituted.
b
c
d
e
substituent on the ring. The results are shown in Table 3. -SO2-
substitution for -CH2- by attachment of a p-methylphenylsulfonyl
(Glide docking, Schrodinger, LLC, New York, NY, 2018) revealed a
common binding pose, adopted by most ligands, that is shown in
Fig. 2. Fig. 3 shows the ligand interaction diagram for 10b. The core
of the inhibitor fit into a hydrophobic channel defined by Leu568’
(Leu A:26), Val570 (Val E:28), Trp571’ (Trp A:29) and Ile573 (Ile
E:31). Two putative H-bonds were present. The first, between the
carboxylate on ring A and Lys 574 (Lys E:32) was commonly
observed for all docked compounds in this study. For 10a and 10b,
an additional hydrogen bond formed between the sulfonyl or
carbonyl group and Gln 575’ (Gln A:33). The -O-tBu group of 10b is
solvent exposed in this model, although it is clearly important since
saponification reduced activity by a factor of 5 (data not shown).
Additionally, the KI of 10b did not accurately reflect its bioactivity,
implying that the t-Bu group may be involved in interactions other
than pocket binding. 10a docked in an identical binding mode to
10b with the added potential of ring stacking between ring D and
Trp 5710. The lower molecular weight and/or 20e30% reduction in
logP of 10a and 10b compared to 9 m and other 4-ring compounds
in Table 2 indicated an improvement in drug-like characteristics
group as ring D led to compound 10a with IC50 ¼ 0.9
mM, the same
potency as that of 9m, containing a p-trifluoromethyl benzyl group
as ring D. Although fluorine substitution is likely to have reduced
the bioactivity of 9m compared to a protonated counterpart (not
measured), based on results shown in Table 1, the sub-mM potency
of 10a is in line with that of other 4-ring compounds. The relative
reduction in logP of at least 1.6 units, meanwhile, suggests that a
polar interaction involving the -SO2- moiety is substituting for
hydrophobic contacts. eCO- as the linker, in the form of the Boc
group, yielded significant enhancement of potency compared to
previous compounds lacking a benzyl group at D (ref [2] and
Table 4), with 10b displaying 0.9e1.0
molecular weight and logP are substantially lower than for com-
M activity discussed in Tables 1 and 2.
mM antiviral activity. Both
pounds with ꢁ1.0
m
Although limited in scope, the results here suggest that a polar
interaction, likely with a Gln in the pocket, replaced previous hy-
drophobic interactions conferred by the CH2 linker. Virtual docking
Table 3
Testing added polarity in the C-D connector.
Group Dd
MWt
520.6
logPa
6.25
log Sb
KI (HP)c
1.2
IC50 CCFc
1.3
IC50 VCF Ba-Lc
0.8
IC50 VCF IIIBc
0.9
CC50
c
10a
ꢀ7.99
>50
,e
9ma
524.5
466.5
8.98
6.81
ꢀ9.69
ꢀ8.72
1
2
0.9
0.9
1.0
1.0
>50
>20
10b
1.9
0.8
a
Calculated water-octanol partition coefficient.
Calculated solubility.
All values in mM.
Groups A, B, C and linker position as in Table 2.
Previously published in Ref. [2].
b
c
d
e

G. Zhou et al. / European Journal of Medicinal Chemistry 161 (2019) 533e542
537
e
Table 4
Testing modifications at rings B, C.a
Group B
Group C
Group D
MWt
logPc
log Sd
KI (HP)e
IC50 CCFe
IC50 VCF Ba-Le
IC50 VCF IIIBe
CC50
f
11a
11b
Indole
Indazole
Indole
Indole
Indole
Indazole
Indazole
Indazole
indole
CH3
CH3
1-CH3
2-CH3
1-CH3
CH3
380.4
399.4
399.4
399.4
400.4
382.5
486.6
6.42
5.97
5.93
5.69
5.41
3.58
8.11
ꢀ7.42
ꢀ7.29
ꢀ7.31
ꢀ6.84
ꢀ7.10
ꢀ5.74
ꢀ9.20
1.4
2.1
2.5
6.4
19
10
1
1.3
4.6
7
4.5
4.3
>50
>20
>20
>20
~200
>50
>25
g
>20
10.1
14
>20
>25
0.8
>20
11.3
>20
>20
>25
0.9
g
g
11c1
11c2
Indole
28
g
11e
11f
Indazole
Indoline
Indole
>25
>10
0.8
f
9n
Indole
9s
3-Me-indole
indole
Indole
500.6
488.6
487.6
8.31
5.40
7.53
ꢀ8.50
ꢀ8.65
ꢀ9.13
1.2
3.6
1.0
4
10.5
1.9
12.9
2.4
~70
>25
>20
15e
15f
Indoline
indole
1
Indazoleb
n.d.
5.8
5.5
18d
indole
m-aniline
, H
462.5
7.20
ꢀ8.47
1.5
>10
>20
>20
>10
18a
18c
indole
indole
m-aniline
m-aniline
, CH3
476.6
562.7
7.78
8.60
ꢀ8.44
0.8
1.6
4.2
1.6
5.5
1.2
4.2
>20
>10
ꢀ10.56
0.45
11d
11g
Indole
indole
Indoline
468.6
480.6
6.88
7.24
ꢀ9.00
ꢀ9.60
0.9
4.7
2.9
2.6
2.3
1.2
2.8
1.7
>25
>20
3-Me-indole
18b
15b
indole
indole
m-aniline
, CH3
456.5
515.6
6.95
7.20
ꢀ8.95
ꢀ9.43
1.6
1.0
5.9
0.3
15.6
1.4
11.3
1.4
>20
>50
Indazole
15c
indazole
Indole
515.6
6.96
ꢀ8.82
3.2
2.8
3.1
2.5
>20
a
Group A is connected at N1 of group B except.
b
c
d
e
f
Connected at N2 of indazole; group D is connected at N1 (indole, indazole, indoline) or N (aniline) of group C.
Calculated water-octanol partition coefficient.
Calculated solubility.
All values in mM.
Previously published in Ref. [2].
g
m-F on ring A.
while activity was retained.
Without the methyl group on the aniline nitrogen, vertices 2 and 3
are both absent. We also tested indazole with nitrogen at position 2
as a potential substitute for indole, allowing for increased inhibitor
polarity while perhaps not having the negative effect of a benz-
imidazole. Indazole and indoline groups are regularly used in bio-
logical or pharmaceutical research [17e19]. The SAR, shown in
Table 4, was conducted concurrently with variation in group D and
we therefore identified four reference compounds, 9a, 9n, 10b, 11a
to which the non-bisindole containing compounds could be
compared. They have group D ¼ m-methylbenzoate, m-methox-
ybenzyl, Boc and methyl, respectively. The results of Table 4 are
summarized in Table 5 as an activity score relative to the reference
compounds. Substituting any of the heterocyclic variants for indole
at position B resulted in at least a six-fold loss of potency. For
example, substituting indole with indazole or 3-methylindole at
position B caused a 10-fold reduction in activity in either case (15c
vs 9a, 9s vs. 9n). 11f, with an indoline at position B, had no
2.4. Examination of the molecular core
The 6-60 bisindole scaffold influences the shape of the inhibitors,
and is predicted to contribute significant interactions with sur-
rounding amino acids when buried in the hydrophobic pocket. To
further understand the importance of the scaffold, alternative
heterocyclic rings were substituted for indoles to test specific in-
teractions. Thus indole groups were exchanged (Fig. 4) for indoline,
3-methyl-indole, aniline or indazole. In previous studies, we
determined that adding a polar nitrogen atom at vertex 3 (by
benzimidazole substitution) caused a 50-fold loss in potency [2].
Saturation of the 2e3 double bond in indoline is accompanied by
loss of full planarity. Adding a methyl group at position 3 increases
hydrophobic content and bulk, while an N-methyl substituted an-
iline “removes” vertex 3 completely, compared to an indole.

538
G. Zhou et al. / European Journal of Medicinal Chemistry 161 (2019) 533e542
Fig. 3. Interaction diagram for virtually docked ligand 10b in the pocket of 2xra (Glide docking). Negative and hydrophobic residues surrounding the ligand are shown in purple and
green respectively, while polar residues are shown in blue. Hydrogen bonds to carbonyl groups are shown as magenta colored arrows, and an additional salt bridge interaction with
the carboxylate group is represented by a line. Solvent exposure is indicated by grey circles, with the size reflecting the degree to which each vertex is exposed. Residue numbers
shown are those for the 2xra construct, with residues 26e35 corresponding to HXB2 residues 568e577 (see also Fig. 2). (For interpretation of the references to color in this figure
legend, the reader is referred to the Web version of this article.)
binding, making it difficult to fully interpret observed changes in
activity. Nevertheless it is apparent that aniline-containing com-
pounds performed poorly. As an example, the potency of 10b
containing an indole at C and Boc group replacing D was unique,
since the benefits of substitution of Boc for benzyl did not extend to
a compound with aniline at the C position, i.e. 18b was much worse
than 18a (Table 4). Increased bulk of the substituents at the aniline
nitrogen was also tested in series 18d, 18a and 18c, with an
improvement in KI for HP binding and mixed results in bioactivity.
Some of this could be due to decreased solubility associated with
the increase in hydrophobicity, but as observed for compounds in
Table 2 we again see that increasing bulk at this end of the molecule
was counterproductive.
Unlike indole or indoline, the indazole group can be linked to
group D and/or A at N2 as well as N1. N2-linked isomers (11c2 and
15f) proved to have lower activity compared to N1-linked coun-
terparts and were not explored further.
In summary, an indole at position B seemed to be essential for
good bioactivity. Additionally, none of the tested compounds was
consistently better than with an indole at position C. These results
speak strongly to the need for the bisindole shape and likely tight
fit within the groove of the HP. They also suggest that indole B is
more deeply buried in the pocket, in agreement with the predicted
binding pose (Fig. 2).
Fig. 4. Substitutions of alternative single ring or fused ring structures for indole at the
core of the molecular scaffold, showing the changes in vertex character at the equiv-
alent of indole positions 2 and 3.
observable activity, compared to 11a which had IC50 ¼ 4
mM.
3. Discussion
Interestingly, the calculated logP for 11f is relatively low (3.58),
which could be a factor in its lack of activity, and might suggest a
lower limit of logP for successful intervention in the HP. Changes of
ring C had a more moderate effect on bioactivity, typically a factor
of 2e3, except for substitution with an aniline which had varied
effects. Changing hydrophobicity at the 3-position of indole C (with
an indoline in 11d or with 3-methylindole in 11g) was detrimental,
but removing this vertex in 18b by substituting an N-methylated
aniline or removing both vertices 2 and 3 in 18d also had a negative
effect. Substitution with aniline, in addition to affecting local in-
teractions, increases molecular flexibility because of free rotation of
the carbon-nitrogen bond. This can potentially alter the in-
teractions of group D and introduce an entropic penalty upon
In this study, we designed, synthesized and tested a series of
compounds based on a bisindole scaffold as entry inhibitors against
HIV-1 gp41. The bisindole scaffold consisted of 4 rings A-B-C-D
attached linearly, with A ¼ benzoic acid, B and C ¼ indole and D ¼ a
substituted benzyl group. Variations included substituting B and C
indoles with alternative fused ring or single ring structures,
examining substitutions on or for ring D, and replacing hydrogen
with fluorine at different positions. The following findings were
observed:
Activity reached a plateau with increasing hydrophobicity Several
of the new compounds had sub-mM activity against viral infectivity
and cell-cell fusion, and the most active compound 9c exhibited

G. Zhou et al. / European Journal of Medicinal Chemistry 161 (2019) 533e542
539
Table 5
Exploring the character of B ꢀ C moieties.
VCFa
Group B
Group C
Fold increase in IC
Example compounds
50
indole
indole
indole
indole
indole
1
2
2
2
9a, 9n, 10b, 11ab
15e vs. 9n; 11d vs. 10b
11g vs 10b
15b vs 9a; 11c vs 11a
18b vs. 10b (15); 18d vs. 10b (>20); 18a vs. 9n (2)
11b vs. 11a; 15c.vs.9a
9s vs. 9n
11f vs 11a
indoline
3-Me-indole
indazole
aniline
indole
indole
indole
2e20
ꢃ6
10
>6
>5
indazole
3-Me-indole
indoline
indazole
indole
indazole
11e vs. 11a
a
Inhibition of virus e cell infectivity.
Reference compounds defined by substituent D, for comparison with non - bisindole variants.
b
VCF
binding affinity KI ¼ 0.3
mM and IC
¼ 0.2
m
M against fusion.
character, logP ~7 or higher, was a necessary requirement for sub-
mM inhibitory activity against HIV-1 fusion. Further enhancement
50
However, this compound has a high logP (8.36) and low predicted
solubility (logS ¼ ꢀ9.21), beyond the range of drug-like character-
istics and possibly incurring non-specific hydrophobic interactions
and/or toxicity at high concentrations [20]. Studies of fluorinated
derivatives revealed that increasing logP was detrimental to ac-
tivity, suggesting an upper limit beyond which bioavailability is
impacted for this class of compounds. Increased hydrophobicity
could mask the potency of the compounds in cell-based assays
because of the reduced solubility.
of hydrophobicity was insufficient as a means of improving potency
beyond a mid-nM level. Instead, potency could be promoted by
adding specific polar interactions, taking into account a nearby Gln
residue in predicted docked structures. SAR data suggested that the
most effective molecules could also interact with lipids. The best
compound to emerge from this study is 10b with logP ¼ 6.8 and
anti-fusion activity 0.8e1.0 mM. Hydrogen bonding between the
Boc carbonyl group and pocket Gln 5750 may underlie its improved
activity compared to other 3-ring compounds, providing the po-
tential for improved specificity and drug-likeness. The compound
has a molecular weight of 466.5da, on the small side for a PPI
(typically ranging from 241 to 974 da [21]), leaving sufficient
compound space for its optimization. While logP is still high, many
optimization examples have shown that reduction in logP is
possible using medicinal chemistry efforts [22].
6-60 bisindole was the preferred scaffold By substituting various
alternative fused ring or single ring moieties for indole, we found
that the 6-6‘-bisindole remained the preferred scaffold. It is likely
that this structure conforms well to the gp41 pocket shape as we
have previously surmised [2]. The group B indole was particularly
important and docking studies show it making 14 close contacts
(ꢁ4 Å) with atoms on Ile 573, Lys 574 and Gln 577, while indole C
made 6 close contacts with atoms on Leu 5680 and Trp 571‘.
Position D may contribute a ternary interaction with lipids We
discovered that the exact nature of substitutions at position D did
not significantly impact activity, other than the requirement that
they not involve negative charge or significantly added bulk. Yet
ring D or an equivalent hydrophobic group contributed ~10-fold to
antiviral potency. The positive charge on the HP would be expected
to encourage negative charge on ligands, but we have consistently
observed that negative charge on D is unfavorable. It is possible that
ring D may not be specific for pocket binding interactions but
instead may associate with lipid or other hydrophobic entity. Hy-
drophobicity, in addition to being a requirement to interact effec-
tively with certain residues in the pocket, may encourage
partitioning in areas near membrane where the HP is located.
5. Experimental section
5.1. Chemistry
The organic synthesis is described in Schemes 1 e 4 . Details of
the syntheses are provided in the Supplementary Data. In-
termediates 5a-e and 6a-n were obtained either from a commercial
source or prepared according to Scheme S1 in the Supplementary
Data. Scheme 1 shows the synthesis of bisindole compounds 9a-n.
9a, 9b and 9n are previously prepared compounds 1, 2 and 3,
respectively (Fig. 1). 9o-q are corresponding fluorinated derivatives
of 9n, prepared by the same method with available starting mate-
rials. 9ꢂ has trifluoromethoxy on ring D, 9p has both tri-
fluoromethoxy on ring D and m-fluoro on ring A, 9q has 5-fluoro on
ring B. 9r was prepared from 8a and 1-Boc-4-
bromomethylpiperidine, and 9s was prepared according to
Scheme 1 using 5d (3-methyl-6-bromoindole) and 6n as starting
materials. Similarly 10a and 10b were prepared from 8a using the
appropriate chemical group in place of 6.
A
polar substitution improved drug-likeness Introducing a
carbonyl or sulfonyl linker between C and D improved drug-like
characteristics of the ligands without reducing potency. 10a, con-
taining a sulfonyl linker, had the same activity (0.9
containing a methylene linker, while logP was reduced by 2.7 units.
The same 0.8e1.0
M activity was observed for 10b (D ¼ Boc),
mM) as 9 m,
m
which contained a carbonyl in the linker and an eO-tBu group
instead of substituted phenyl. This compound had logP reduced by
2 units and significantly lower molecular weight compared to 9 m.
In docking studies, the carboxylate group on ring A of all com-
pounds participated in a hydrogen bond with the εNH2 of Lys 574,
while the sulfonyl or carbonyl groups in 10a or 10b, respectively,
participated in an additional hydrogen bond with Gln 5750.
Scheme 2 describes the synthesis of 3-ring systems 11b-e,
including compounds substituted with an indazole or indoline in
place of indole. For indazole compounds, reactions resulted in a
mixture of 1 and 2-linked isomers, which were separated using
flash chromatography and identified by COSY NMR. 11c1 (shown in
Scheme 2B) and 11c2 are N1-methyl and N2-methyl isomers at ring
C. Similar reactions to those shown were used to obtain different
combinations of the ring systems. Scheme 2A was employed to
make 11a using 5a and 6a as starting materials [2], 11f using 5c and
6a, and 11h using 5a and 6d. 11i was made as in Scheme 2B using 5-
fluoro-6-bromoindole in place of 5b. 11g was made as in Scheme 2C
using 5d in place of 5c.
4. Conclusions
In conclusion, this study of small molecule intervention in the
hydrophobic pocket of gp41 has revealed that a bisindole core in
the molecular scaffold together with substantial hydrophobic
Scheme 3 describes synthesis of 4-ring compounds 15b and 15c

540
G. Zhou et al. / European Journal of Medicinal Chemistry 161 (2019) 533e542
Scheme 1. Synthesis of 4-ring compounds 9a-9n.
Scheme 2. Synthesis of 3-ring compounds 11b-11e.
with one indole and one alternative fused ring structure. It also
shows an improved method for the synthesis of 9a. Based on our
practical experience, it was difficult to separate 9a from 8a by col-
umn chromatography, and the ester of 6a was labile in strong basic
conditions. By employing a benzyl group to protect the carboxylic
acid on 6-bromoindol-1-ylmethylbenzoic acid, followed by bor-
ylation and Suzuki coupling with 7a, we successfully obtained
higher yield and easy purification of 9a. 15d is a fluorinated

G. Zhou et al. / European Journal of Medicinal Chemistry 161 (2019) 533e542
541
Scheme 3. Synthesis of 4-ring compounds containing a fused ring other than indole.
Scheme 4. Synthesis of aniline containing compounds.
derivative of 9a prepared by the same method. 15e was prepared
using Scheme 2C but replacing the first Boc protection step with
reaction of 5c with 6n. 15f, prepared using a procedure similar to
Scheme 2A, contains an N2-substituted indazole at ring B.
Scheme 4 describes synthesis of aniline containing compounds
18a-d.
York, NY, 2018). Inhibition constants for HP-binding (KI) were
measured in a competitive inhibition fluorescence experiment [23]
by concentration-dependent displacement of fluorescently labeled
HP-binding C-peptide from the NHR binding site. Concentrations
that yielded 50% inhibition of biological activity (IC50) were ob-
CCF
tained using cell-cell fusion (IC ) and viral infectivity (virus-cell
50
VCF
fusion; IC ) assays as previously described [2]. Antiviral assays
50
were run using lab-adapted viral strains Ba-L and IIIB. Details are in
the Supplementary Data.
5.2. Compound properties
Compound properties are reported in Tables 1e4. Structures
were optimized using LigPrep (LigPrep, Schrodinger, LLC, New York,
NY, 2018). Pharmaceutically relevant properties including logP
(water/octanol partition coefficient) and logS (aqueous solubility)
were calculated using QikProp (QikProp, Schrodinger, LLC, New
Acknowledgements
This work was supported by National Institutes of Health (NIH)
Grant AI122847 to MG and internal funds from Southern Research

542
G. Zhou et al. / European Journal of Medicinal Chemistry 161 (2019) 533e542
[11] M.A. Kuenemann, L.M. Bourbon, C.M. Labbe, B.O. Villoutreix, O. Sperandio,
Which three-dimensional characteristics make efficient inhibitors of protein-
protein interactions? J. Chem. Inf. Model. 54 (2014) 3067e3079.
[12] G. Fischer, M. Rossmann, M. Hyvonen, Alternative modulation of protein-
protein interactions by small molecules, Curr. Opin. Biotechnol. 35 (2015)
78e85.
Institute. Molecular graphics were produced using the Maestro
program, provided free to academic users. We thank Schrodinger
LLC, NY, NY for a free trial version of the ligand preparation and
docking software.
[13] G. Zhou, D. Wu, E. Hermel, E. Balogh, M. Gochin, Design, synthesis, and
evaluation of indole compounds as novel inhibitors targeting Gp41, Bioorg.
Med. Chem. Lett 20 (2010) 1500e1503.
Appendix A. Supplementary data
[14] B. Strockbine, R.C. Rizzo, Binding of antifusion peptides with HIVgp41 from
molecular dynamics simulations: quantitative correlation with experiment,
Proteins 67 (2007) 630e642.
Supplementary data to this article can be found online at
https://doi.org/10.1016/j.ejmech.2018.10.048.
[15] H.J. Bohm, D. Banner, S. Bendels, M. Kansy, B. Kuhn, K. Muller, U. Obst-Sander,
M. Stahl, Fluorine in medicinal chemistry, Chembiochem 5 (2004) 637e643.
[16] E.P. Gillis, K.J. Eastman, M.D. Hill, D.J. Donnelly, N.A. Meanwell, Applications of
fluorine in medicinal chemistry, J. Med. Chem. 58 (2015) 8315e8359.
[17] D.D. Gaikwad, A.D. Chapolikar, C.G. Devkate, K.D. Warad, A.P. Tayade,
R.P. Pawar, A.J. Domb, Synthesis of indazole motifs and their medicinal
importance: an overview, Eur. J. Med. Chem. 90 (2015) 707e731.
[18] J. Hu, Y. Wang, Y. Li, D. Cao, L. Xu, S. Song, M.S. Damaneh, J. Li, Y. Chen,
X. Wang, L. Chen, J. Shen, Z. Miao, B. Xiong, Structure-based optimization of a
series of selective BET inhibitors containing aniline or indoline groups, Eur. J.
Med. Chem. 150 (2018) 156e175.
References
[1] M. Gochin, G. Zhou, Amphipathic properties of HIV-1 gp41 fusion inhibitors,
Curr. Top. Med. Chem. 11 (2011) 3022e3032.
[2] G. Zhou, V. Sofiyev, H. Kaur, B.A. Snyder, M.K. Mankowski, P.A. Hogan,
R.G. Ptak, M. Gochin, Structure-activity relationship studies of indole-based
compounds as small molecule HIV-1 fusion inhibitors targeting glycoprotein
41, J. Med. Chem. 57 (2014) 5270e5281.
[3] L. Lu, F. Yu, L. Cai, A.K. Debnath, S. Jiang, Development of small-molecule HIV
entry inhibitors specifically targeting gp120 or gp41, Curr. Top. Med. Chem. 16
(2016) 1074e1090.
[4] L. Cai, S. Jiang, Development of peptide and small-molecule HIV-1 fusion in-
hibitors that target gp41, ChemMedChem 5 (2010) 1813e1824.
[5] D.C. Chan, C.T. Chutkowski, P.S. Kim, Evidence that a prominent cavity in the
coiled coil of HIV type 1 gp41 is an attractive drug target, Proc. Natl. Acad. Sci.
U.S.A. 95 (1998) 15613e15617.
[6] G. Zhou, D. Wu, B. Snyder, R.G. Ptak, H. Kaur, M. Gochin, Development of
indole compounds as small molecule inhibitors of HIV-1 gp41, J. Med. Chem.
54 (2011) 7220e7231.
[7] F. Naider, J. Anglister, Peptides in the treatment of AIDS, Curr. Opin. Struct.
Biol. 19 (2009) 473e482.
[8] Y. He, S. Liu, W. Jing, H. Lu, D. Cai, D.J. Chin, A.K. Debnath, F. Kirchhoff, S. Jiang,
Conserved residue Lys574 in the cavity of HIV-1 Gp41 coiled-coil domain is
critical for six-helix bundle stability and virus entry, J. Biol. Chem. 282 (2007)
25631e25639.
[9] X. Ran, J.E. Gestwicki, Inhibitors of protein-protein interactions (PPIs): an
analysis of scaffold choices and buried surface area, Curr. Opin. Chem. Biol. 44
(2018) 75e86.
[19] A. Aguilar, J. Lu, L. Liu, D. Du, D. Bernard, D. McEachern, S. Przybranowski, X. Li,
R. Luo, B. Wen, D. Sun, H. Wang, J. Wen, G. Wang, Y. Zhai, M. Guo, D. Yang,
S.
Wang,
Discovery
of
4-((3'R,4'S,5'R)-6''-Chloro-4'-(3-chloro-2-
cyclohexane-1,2'-pyrrolidine-3',3''-
fluorophenyl)-1'-ethyl-2''-oxodispiro[
indoline]-5'-carboxamido)bicyclo[2.2.2]octane -1-carboxylic acid (AA-115/
APG-115): a potent and orally active murine double minute 2 (MDM2) in-
hibitor in clinical development, J. Med. Chem. 60 (2017) 2819e2839.
[20] P.D. Leeson, B. Springthorpe, The influence of drug-like concepts on decision-
making in medicinal chemistry, Nat. Rev. Drug Discov. 6 (2007) 881e890.
[21] A.P. Higueruelo, A. Schreyer, G.R. Bickerton, W.R. Pitt, C.R. Groom,
T.L. Blundell, Atomic interactions and profile of small molecules disrupting
protein-protein interfaces: the TIMBAL database, Chem. Biol. Drug Des. 74
(2009) 457e467.
[22] A.L. Hopkins, G.M. Keseru, P.D. Leeson, D.C. Rees, C.H. Reynolds, The role of
ligand efficiency metrics in drug discovery, Nat. Rev. Drug Discov. 13 (2014)
105e121.
[23] L. Cai, M. Gochin, A novel fluorescence intensity screening assay identifies
new low molecular weight inhibitors of the gp41 coiled coil domain of HIV-1,
Antimicrob. Agents Chemother. 51 (2007) 2388e2395.
[10] J.A. Wells, C.L. McClendon, Reaching for high-hanging fruit in drug discovery
at protein-protein interfaces, Nature 450 (2007) 1001e1009.