Enhanced potency of bivalent small molecule gp41 inhibitors

Article abstract of DOI:10.1016/j.bmc.2016.11.010

Low molecular weight peptidomimetic inhibitors with hydrophobic pocket binding properties and moderate fusion inhibitory activity against HIV-1 gp41-mediated cell fusion were elaborated by increasing the available surface area for interacting with the heptad repeat-1 (HR1) coiled coil on gp41. Two types of modifications were tested: 1) increasing the overall hydrophobicity of the molecules with an extension that could interact in the HR1 groove, and 2) forming symmetrical dimers with two peptidomimetic motifs that could potentially interact simultaneously in two hydrophobic pockets on the HR1 trimer. The latter approach was more successful, yielding 40–60 times improved potency against HIV fusion over the monomers. Biophysical characterization, including equilibrium binding studies by fluorescence and kinetic analysis by Surface Plasmon Resonance, revealed that inhibitor potency was better correlated to off-rates than to binding affinity. Binding and kinetic data could be fit to a model of bidentate interaction of dimers with the HR1 trimer as an explanation for the slow off-rate, albeit with minimal cooperativity due to the highly flexible ligand structures. The strong cooperativity observed in fusion inhibitory activity of the dimers implied accentuated potency due to the transient nature of the targeted intermediate. Optimization of monomer, dimer or higher order structures has the potential to lead to highly potent non-peptide fusion inhibitors by targeting multiple hydrophobic pockets.

Full text of DOI:10.1016/j.bmc.2016.11.010

Bioorganic & Medicinal Chemistry 25 (2017) 408–420
Contents lists available at ScienceDirect
Bioorganic & Medicinal Chemistry
journal homepage: www.elsevier.com/locate/bmc
Enhanced potency of bivalent small molecule gp41 inhibitors
Vladimir Sofiyev a, Hardeep Kaur a, Beth A. Snyder c, Priscilla A. Hogan c, Roger G. Ptak c, Peter Hwang d,
Miriam Gochin a,b,
⇑
a Department of Basic Sciences, Touro University-California, Vallejo, CA 94592, United States
b Department of Pharmaceutical Chemistry, University of California San Francisco, CA 94143, United States
c Southern Research Institute, 431 Aviation Way, Frederick, MD 21701, United States
d Department of Biophysics and Biochemistry, University of California San Francisco, CA 94143, United States
a r t i c l e i n f o
a b s t r a c t
Article history:
Low molecular weight peptidomimetic inhibitors with hydrophobic pocket binding properties and mod-
erate fusion inhibitory activity against HIV-1 gp41-mediated cell fusion were elaborated by increasing
the available surface area for interacting with the heptad repeat-1 (HR1) coiled coil on gp41. Two types
of modifications were tested: 1) increasing the overall hydrophobicity of the molecules with an extension
that could interact in the HR1 groove, and 2) forming symmetrical dimers with two peptidomimetic
motifs that could potentially interact simultaneously in two hydrophobic pockets on the HR1 trimer.
The latter approach was more successful, yielding 40–60 times improved potency against HIV fusion over
the monomers. Biophysical characterization, including equilibrium binding studies by fluorescence and
kinetic analysis by Surface Plasmon Resonance, revealed that inhibitor potency was better correlated
to off-rates than to binding affinity. Binding and kinetic data could be fit to a model of bidentate inter-
action of dimers with the HR1 trimer as an explanation for the slow off-rate, albeit with minimal coop-
erativity due to the highly flexible ligand structures. The strong cooperativity observed in fusion
inhibitory activity of the dimers implied accentuated potency due to the transient nature of the targeted
intermediate. Optimization of monomer, dimer or higher order structures has the potential to lead to
highly potent non-peptide fusion inhibitors by targeting multiple hydrophobic pockets.
Received 20 September 2016
Revised 31 October 2016
Accepted 3 November 2016
Available online 19 November 2016
Keywords:
HIV-1 gp41
Hydrophobic pocket
Fusion inhibition
Peptidomimetic inhibitors
Bivalent inhibitors
Fluorescence
Surface plasmon resonance
Kinetics
Cooperativity
Ó 2016 Elsevier Ltd. All rights reserved.
1. Introduction
One approach to improving potency of small molecules is to
extend them to form additional interactions with HR1 beyond
The hydrophobic pocket (HP) on the surface of the gp41 tri-
meric HR1 coiled coil is an important target for antiretroviral drugs
against HIV fusion.1 The pocket sits at the interface between HR1
and helical HR2 segments which wrap down the sides of the coiled
coil to form a six-helix bundle structure upon successful virus–host
membrane fusion (Fig. 1A).2,3 The pocket is exposed to inhibitors
for 20–30 min during a conformational change of gp41 prior to
six-helix bundle formation.4 Small molecules that bind in this
the hydrophobic pocket. HR1 has multiple potential interactive
sites both within a single groove and between adjacent grooves
of the trimer. Bidentate ligands are generally 1–2 orders of magni-
tude more potent than their monomeric counterparts against tar-
gets with more than one binding site.12,13 To our knowledge, this
concept has not been tested with small molecule inhibitors target-
ing gp41. For C-peptides, the effect of multivalent interactions is
clear. 18 residue HP-binding C-peptides containing the motif
pocket have been developed into low to sub-
lM inhibitors of
WxxWDxxI are low l
M binders14,15 and fusion inhibitors,16, while
HIV entry,5–8 but so far efforts at small molecule discovery have
failed to achieve the potency afforded by C-peptides derived from
HR2.9,10 These are nM inhibitors, including T20 which is an FDA-
approved fusion inhibitor.11 Non-peptide inhibitors of HIV fusion
could have advantageous properties as drugs, such as longer half-
life and lower cost and the potential for oral bioavailability.
34 residue C-peptides have low nM activity.17 Dimerization of 22
residue helical C-peptides improved activity by a factor of 9.18 Sim-
ilarly, D-peptides designed to bind in the hydrophobic pocket19
have activity improved by 10–300 fold upon dimerization with a
long polyethylene glycol (PEG) linker, presumably by simultaneous
interactions in two HP’s of the trimer.20,21
An important consideration in designing extension or dimeriza-
tion of small molecule hydrophobic pocket binders is selection of a
tether that will not disturb the binding mode. Helical pep-
tidomimetics are suitable candidates for this approach because
⇑
Corresponding author at: Touro University – California, 1310 Club Drive, Mare
Island, Vallejo, CA 94592, United States.
E-mail address: miriam.gochin@tu.edu (M. Gochin).
http://dx.doi.org/10.1016/j.bmc.2016.11.010
0968-0896/Ó 2016 Elsevier Ltd. All rights reserved.

V. Sofiyev et al. / Bioorg. Med. Chem. 25 (2017) 408–420
409
Fig. 1. Experimental and model structures of the gp41 HR1–HR2 domain and interactions. A. Homology modeling of HIV-1 HXB2 ectodomain from the structure of the
corresponding SIV domain (pdb 2EZO). The HR1 trimer is shown as a space filling model in tan, and loop and HR2 residues are shown as a ribbon representation in yellow; B.
Expansion of the hydrophobic pocket region from A. C. Crystal structure (pdb 2R5D) of D-peptide PIE7 showing Trp and Leu residues filling the pocket; D, E. Modeled low
energy structures of two subunit and three subunit peptidomimetics with Trp and Leu residues filling the pocket. See Fig. 2 for the structures of the peptidomimetic
inhibitors.
they are designed to emulate the amphipathic C-peptide helix,
which has HP binding residues on one face of the helix (Fig. 1B).
Peptidomimetic inhibitors are expected to have side chains occu-
pying the pocket while the non-peptidic ‘‘backbone” is exposed
above the pocket and available for tethering.
We previously examined a combinatorial library of three-sub-
unit peptidomimetics containing a central aryl-alkoxy scaffold
replacing the peptide backbone, H2N-R3-[R2]-R1-OH, as well as
the precursor two-subunit library, O2N-[R2]-R1-OH, developed by
the Boger group.22 Screening with a binding assay for the gp41
hydrophobic pocket revealed a clear propensity for aromatic resi-
dues at R1 and R2, especially for Trp, Nap and Phe4Cl.23 Ile was also
well-tolerated at R1 in three subunit compounds with an aryl R3,
consistent with the inhibitors emulating the conserved WWI motif
(Fig. 1C, D). NMR paramagnetic relaxation enhancement measure-
ments on two inhibitors of this class, O2N-[HoPhe]-Ala-OH and
O2N-[Ala]-Nap-OH confirmed this binding mode.24 We selected
two subunit compounds with R1 = Trp or Phe4Cl and R2 = Trp to
explore attachment of a tether to the aryl alkoxy moiety, and to
evaluate the effect of extension or dimerization on binding affinity
and inhibition of virus–cell and cell–cell fusion.
2. Materials and methods
2.1. Synthetic chemistry
2.1.1. Sources
3-Fluoro-4-nitrobenzoic acid was purchased from Matrix Scien-
tific (Fisher), 3-(2-hydroxyethyl)indole, Zn nanopowder, was pur-
chased from Sigma-Aldrich.ꢁTHF was purchased from EMD
Millipore (VWR). 4-Chlorobenzylalcohol was purchased from Alfa
Aesar. L-Tryptophan tert-butyl ester hydrochloride was purchased
from Chem-Impex International.
of gp41 C-peptides. The compounds H2N-Asn-[Trp]-Trp-COOH
CCF
(KI = 0.8 lM, IC = 6 lM) and H2N-Trp-[Trp]-Leu-COOH (KI = 1.3 -
50
CCF
l
M, IC = 5 lM) were among the most potent compounds. In the
50
absence of R3 in the two subunit compounds, R2 = Trp or Nap and
R1 = Trp, Nap or Phe4Cl were the optimal choices. Bioactivity was
reduced by a factor of ꢀ4 compared to the three subunit com-
pounds listed.24
Most low energy simulated structures adopted the predicted
binding pose in which the side chains reached down into the
pocket and the aryl alkoxy ‘‘backbone” lay above the pocket
2.1.2. Preparation of the aryl subunits
The aryl subunits were synthesized following the general proce-
dure described previously.22

410
V. Sofiyev et al. / Bioorg. Med. Chem. 25 (2017) 408–420
1. R2 = 3-Ethyleneindole. Sodium hydride (60%, 0.5 g, 12.5 mmol)
4. Ester 4a (168 mg, 0.324 mmol) was taken up in dry HCl in
dioxane (4 M, 3.0 mL), stirred for 1 h, and concentrated. Purifica-
tion by silica gel chromatography (1:9 MeOH:DCM) yielded car-
boxylic acid 4 (93.2 mg, 62%). Data for 4: MS Calc. 463.91, found
(MꢃH)ꢃ 462.17; 1H NMR (acetone) d 7.68 (d, 1H, J = 7.8 Hz),
7.48–7.36 (m, 6H), 7.30–7.28 (m, 2H), 7.08 (t, 1H, J = 7.7 Hz), 7.01
(t, 1H, 7.4 Hz), 6.68 (d, 1H, J = 8.0 Hz), 5.04 (s, 2H), 5.00 (br, ꢀ1H),
3.5–3.4 (m, 2H).
and 3-(2-hydroxyethyl)indole (1.0 g, 5.4 mmol) were stirred in
10 mL THF at 0 °C for 15 min, after which 3-fluoro-4-nitrobenzoic
acid (1.04 g, 6.45 mmol) was added. The mixture was stirred at
0 °C overnight, and then quenched with aqueous saturated NH4Cl,
diluted with ethyl acetate, washed with 0.1 M HCl (50 mL, ꢂ2). The
organic layer was concentrated and purified by silica gel chro-
matography (10:1 DCM/MeOH, 1% acetic acid), affording 2.6 g pure
product (74%). Data for 1: 1H NMR (DMSO) d 7.87 (d, 1H, J = 8.4 Hz),
7.81 (s, 1H), 7.62 (dd, 1H, J = 8.3 Hz, 0.8 Hz), 7.58 (d, 1H, J = 7.8 Hz),
7.36 (d, 1H, J = 8.0 Hz), 7.23 (d, 1H, J = 1.9 Hz), 7.07 (t, 1H,
J = 7.3 Hz), 6.98 (t, 1H, J = 7.4 Hz), 4.43 (t, 2H, J = 6.9 Hz), 3.18 (t,
2H, J = 6.9 Hz).
2.1.4. Preparation of bivalent compounds
5. Solution of amine 3a (543 mg, 1.01 mmol) and succinic anhy-
dride (134 mg, 1.33 mmol) in THF (10 mL) was stirred for 4 h at
65 °C. Purification by silica gel chromatography (7:3 EtOAc/hex-
ane) afforded 5 (637 mg, 99%). Data for 5: Rf = 0.40 (10:1 DCM:
MeOH). 1H NMR (CDCl3) d 8.75 (s, 1H), 8.65 (s, 1H), 8.16 (d, 1H,
J = 8.2 Hz), 7.72 (s, 1H), 7.57 (d, 1H, J = 7.9 Hz), 7.53 (d, 1H,
J = 7.3 Hz), 7.28–7.22 (m, 3H), 7.12–6.99 (m, 5H), 6.96 (d, 1H,
J = 1.8 Hz), 6.89 (d, 1H, J = 7.6 Hz), 6.84 (d, 1H, J = 1.7 Hz), 5.00 (q,
1H, J = 7.4 Hz, 5.8 Hz), 3.99 (t, 2H, J = 5.5 Hz), 3.44–3.29 (m, 2H),
3.02 (t, 2H, J = 5.5 Hz), 2.54 (t, 2H, J = 6.1 Hz), 2.20 (t, 2H,
J = 5.8 Hz), 1.42 (s, 9H).
6. Solution of amine 4a (456 mg, 0.88 mmol) and succinic anhy-
dride (111 mg, 1.10 mmol) in THF (8 mL) was stirred overnight at
54 °C. Purification by silica gel chromatography (4:1 EtOAc/hex-
ane ? EtOAc) afforded 6 (306 mg, 56%). Data for 6: 1H NMR (CDCl3)
d 8.86 (s, 1H), 8.21 (d, 1H, J = 8.2 Hz), 8.08 (s, 1H), 7.57 (d, 1H,
J = 7.7 Hz), 7.30–7.27 (m, 4H); 7.22–7.20 (m, 2H), 7.11–7.07 (m,
2H), 7.04–7.00 (m, 2H), 6.95 (d, 1H, J = 7.6 Hz), 4.98 (q, 1H,
J = 7.3 Hz, 5.7 Hz), 4.82 (s, 2H), 3.42–3.29 (m, 2H), 2.69–2.58 (m,
>4H), 1.41 (s, 9H).
2. R2 = 4-chlorobenzyl. Sodium hydride (60%, 0.68 g, 17 mmol)
and 4-chlorobenzylalcohol (1.21 g, 8.5 mmol) were stirred in
15 mL THF at 0 °C for 15 min, after which 3-fluoro-4-nitrobenzoic
acid (1.5 g, 8.1 mmol) was added. The mixture was stirred at 0 °C
overnight, and then quenched with aqueous saturated NH4Cl,
diluted with ethyl acetate, washed with 0.1 M HCl (75 mL, ꢂ2)
and brine. The organic layer was concentrated and separated by sil-
ica gel chromatography (3:2:0.1 hexanes/ether/acetic acid ? 10:1
DCM/MeOH) to yield 2.2 g product 5 (89%). Data for 2: 1H NMR
(DMSO) d 8.00 (d, 1H, J = 8.3 Hz), 7.86 (d, 1H, J = 1.4 Hz), 7.67 (dd,
1H, J = 8.3 Hz, 1.5 Hz), 7.49 (s, 4H), 5.39 (s, 2H).
2.1.3. Preparation of 2-unit peptidomimetics
The preparation followed procedures similar to those described
previously.22
3a. (O-tBu) Compound 1 (605 mg, 1.85 mmol) and L-Trp t-butyl
ester (550 mg, 1.85 mmol) were combined in 12 mL dry DMF to
which EDCI (426 mg, 2.22 mmol), HOBt (300 mg, 2.22 mmol) and
DIPEA (0.97 mL, 5.57 mmol) were added. The solution was stirred
overnight, then diluted with EtOAc, washed with sodium bicarbon-
ate, 10% NaCl (ꢂ2), brine, and then concentrated. 850 mg of pro-
duct was obtained on a flash column in 40% EtOAc/hexane,
Rf = 0.35 and used immediately for the next reaction. The product
(850 mg, 1.49 mmol) was reduced by dissolving it in 24 mL of
5:1 acetone: water and adding Zn nanopowder (978 mg,
15.0 mmol), and NH4Cl (1.2 g, 22.4 mmol). An instantaneous
exothermic reaction occurred, after which the mixture was filtered,
diluted with 1:1 EtOAc/ether, washed with bicarbonate and brine
and evaporated. The solid was triturated and purified by silica
gel chromatography (50% EtOAc/hexane) to afford 3 in quantitative
yield. Data for 3a: Rf = 0.30 (50% EtOAc/hexane); MS calc. 538.64,
found (MꢃH)ꢃ 537.12; 1H NMR (CDCl3) d 8.16 (br, 1H), 8.05 (br,
1H), 7.65 (d, 1H, J = 7.9 Hz), 7.60 (d, 1H, J = 7.9 Hz), 7.36 (d, 1H,
J = 8.0 Hz), 7.25 (m, 1H), 7.20 (m, 2H), 7.16–6.99 (m, 6H), 6.54 (d,
1H, J = 8.1 Hz), 6.50 (d, br, 1H, J = 7.52), 5.02 (m, 1H), 4.19 (t, 2H,
J = 6.7 Hz), 3.37 (m, 2H), 3.22 (t, 2H, J = 6.7 Hz), 1.40 (s, 9H).
3. Ester 3a (213 mg, 0.396 mmol) was taken up in dry HCl in
dioxane (4 M, 6.0 mL), stirred for 1.5 h, and concentrated. Purifica-
tion by silica gel chromatography (1:4 MeOH:DCM) and filtration
through a plug of Celite using acetone as the solvent yielded car-
boxylic acid 3 (180 mg, 94%). Data for 3: MS calculated 482.53,
found (MꢃH)ꢃ 481.30; 1H NMR (acetone) d 7.68 (m, 2H), 7.41–
7.25 (m, 6H), 7.13–6.96 (m, 4H), 6.65 (d, 1H, J = 7.7 Hz), 5.00 (m,
1H), 4.19 (t, 2H, J = 6.6 Hz), 3.47–3.35 (m, 2H), 3.24 (t, 2H,
J = 6.6 Hz).
7. Solution of triethylene glycol diamine (H2N-(CH2CH2O)3–
CH2CH2NH2) (24.6 mg, 0.128 mmol), carboxylic acid 5 (130. mg,
0.204 mmol), EDCIꢁHCl (54.0 mg, 0.282 mmol), HOAt (37.2 mg,
0.273 mmol) and DIPEA (0.11 mL, 0.63 mmol) in DMF (2.0 mL)
was stirred for 18 h. Purification by silica gel chromatography
(10:1 DCM:MeOH) afforded the intermediate tBu ester-protected
dimer (143 mg). Data for tBu ester-protected dimer: 1H NMR
(CDCl3) d 9.32 (s, br, 1H), 8.95 (s, br, 1H), 8.13 (d, 1H, J = 8.5 Hz),
7.99 (s, 1H), 7.86 (s, 1H), 7.55 (m, 2H), 7.32 (d, 1H, J = 7.9 Hz),
7.26 (d, 1H, J = 8.5 Hz), 7.22 (s, br, 1H), 7.08 (m, 3H), 7.00 (m,
2H), 6.95 (d, 1H, J = 1.9 Hz), 6.91 (d, 1H, J = 7.5 Hz), 6.86 (t, 1H,
J = 5.1 Hz), 4.98 (q, 1H, J = 7.0 Hz), 4.01 (t, 2H, J = 5.9 Hz), 3.52 (br,
4H), 3.46 (t, 2H, J = 4.6 Hz), 3.38 (m, 4H), 3.06 (t, 2H, J = 5.4 Hz),
2.36 (m, 4H), 1.41 (s, 9H).
This intermediate (143 mg) was taken up in dry HCl in dioxane
(4 M, 2.0 mL), stirred for 70 min, and concentrated under a stream
of Ar. Purification by silica gel chromatography (10:1 DCM:
MeOH ? 4:1 [90:10:0.6:0.6 CH2Cl2:MeOH:H2O:NH4OH]:MeOH)
afforded dimer 7 (55.4 mg, 41% over 2 steps). Data for 7: Rf = 0.20
(10:1 DCM:MeOH); MS calculated: 1335.43 (ammonium salt),
found: (MꢃH)ꢃ 1333.83; 1H NMR (DMSO) d 9.09 (s, 0.5H), 9.02
(s, 0.5H), 8.73 (d, 0.6H, J = 7.5 Hz), 8.11 (d, 1H, J = 8.2 Hz), 8.05
(m, 2H), 7.64 (m, 1H), 7.55 (t, 1H, J = 7.5 Hz), 7.45 (m, 2H), 7.33
(m, 4H), 7.25 (d, 1H, J = 8.1 Hz), 7.18 (d, 1H, J = 2.0 Hz), 7.11–6.82
(m, 4H), 4.68 (m, 1H), 4.26 (m, 2H), 3.41 (m, 4H), 3.24 (m, 6H),
2.62 (q, 2H, J = 7.6 Hz), 2.43 (m, 4H).
8. Solution of triethylene glycol diamine (H2N-(CH2CH2O)3–
CH2CH2NH2) (47.4 mg, 0.246 mmol), carboxylic acid 6 (306 mg,
0.493 mmol), EDCIꢁHCl (117 mg, 0.610 mmol), HOAt (81.4 mg,
0.598 mmol) and DIPEA (0.26 mL, 1.49 mmol) in DMF (4.0 mL)
was stirred for 18 h and then concentrated. Purification by silica
gel chromatography (10:1 DCM:MeOH) afforded the intermediate
tBu ester-protected dimer (107 mg) along with 6 (159 mg). Data
for tBu-ester-protected dimer: 1H NMR (CDCl3) d 8.91 (s, br, 1H),
7.58 (d, 1H, J = 7.8 Hz), 7.31 (s, 1H), 7.25 (m, 5H), 7.16–7.03 (m,
6H), 6.94 (d, 1H, J = 2.1 Hz), 6.85 (d, 1H, J = 7.5 Hz), 4.99 (1H),
4a. (O-tBu) Compound 4 was prepared from compound 2 using
the same method as for preparation of compound 3. Purification by
silica gel chromatography (40% EtOAc/hexane) yielded 4 (781 mg,
92%). Data for 4: Rf = 0.18 (40% EtOAc/hexane); 1H NMR (CDCl3) d
9.14 (br, 1H), 8.32 (d, 1H, J = 8.4 Hz), 7.94 (s, 1H), 7.58 (d, 1H,
J = 7.9 Hz), 7.30–7.24 (m, 6H), 7.13–7.07 (m, 2H), 7.02 (m, 2H),
6.92 (d, 1H, J = 7.4 Hz), 4.97 (q, 1H, J = 7.0 Hz), 4.81 (br, 2H), 3.36
(m, 4H).

V. Sofiyev et al. / Bioorg. Med. Chem. 25 (2017) 408–420
411
4.98 (s, 2H), 3.59 (m, 4H), 3.51 (m, 2H), 3.41 (m, 4H), 2.64 (m, 4H),
1.41 (s, 9H).
gel chromatography (9:1 DCM:MeOH) afforded 11 (45.7 mg, 63%).
Data for 11: Rf = 0.18 (9:1 DCM:MeOH); MS calc. 748.27, found
(MꢃH)ꢃ 746.44. 1H NMR (acetone) d 10.01 (s, br, 0.3H), 9.87 (s,
br, 0.6H), 8.51 (s, 0.7H), 8.32 (d, 1H, J = 8.0 Hz), 7.67 (d, 1H,
J = 7.7 Hz), 7.55 (m, 2H), 7.46 (m, 3H), 7.43–7.31 (m, 6H), 7.27 (s,
1H), 7.10–6.95 (m, 5H), 5.02 (m, 1H), 3.50 (dd, 2H, J = 14.4,
4.8 Hz), 3.37 (m, 2H), (m, 2H), 3.04 (t, 2H, J = 7.2 Hz), 2.52 (t, 2H,
J = 7.8 Hz), 2.37 (t, 2H, J = 7.4 Hz), 2.22 (q, 2H, J = 5.8 Hz), 1.62 (t,
2H, J = 7.4 Hz), 1.44 (t, 2H, J = 7.4 Hz), 1.29 (m, 2H).
This intermediate (107 mg) was taken up in dry HCl in dioxane
(4 M, 1.5 mL), stirred for 65 min, and concentrated under a stream
of Ar. Purification by silica gel chromatography (90:10:0.6:0.6 CH2-
Cl2:MeOH:H2O:NH4OH ? 9:1 [90:10:0.6:0.6 CH2Cl2:MeOH:H2O:
NH4OH]:MeOH) afforded dimer 8 (17.8 mg, 6% over 2 steps). Data
for 8: MS calc 1296.4 (ammonium salt), found (MꢃH)ꢃ 1295.65; 1H
NMR (DMSO) d 9.40 (s, br, 0.3H), 9.29 (s, br. 0.6H), 7.98 (m, 2H),
7.57 (dd, 2H, J = 6.9, 2.1 Hz), 7.52 (d, 1H, J = 6.9 Hz), 7.45 (dm 2H,
J = 8.6 Hz), 7.40 (s, 1H), 7.35 (d, 1H, J = 8.1 Hz), 7.27 (d, 1H,
J = 8.1 Hz), 7.22 (m, 1H), 7.07 (m, 1H), 6.99 (t, 1H, J = 7.7 Hz), 6.85
(t, 1H, J = 7.2 Hz), 4.89 (1H, d, J = 5.0 Hz), 4.68 (m, 2H), 4.13 (q,
4H, J = 5.1 Hz), 2,66 (m, 2H), 2.35 (m, 4H). (Note: A residual water
peak at 3.33 ppm masks peaks in that region).
2.2. Assays
2.2.1. Fluorescence binding assay
Inhibition constants Ki for binding in the hydrophobic pocket
were determined using a fluorescence intensity assay as previously
described.14 FeII(env2.0)3 was used to mimic the hydrophobic
pocket in the gp41 NHR coiled coil. Env2.0 has the sequence bpy-
2.1.5. Preparation of extended monomers
9. To a solution of indole-3-propionic acid (0.62 g, 3.3 mmol)
and 6-aminohexanoic acid methyl ester⁄ (0.60 g, 3.3 mmol) in
20 mL DMF were added EDCIꢁHCl (0.76 g, 4.0 mmol), HOBt
(0.54 g, 4.0 mmol) and DIPEA (2.0 mL, 12 mmol) and the resulting
mixture was stirred overnight. It was then diluted with EtOAc
(120 mL), washed with conc. aq. NaHCO3 (30 mL), 10% NaCl
(30 mL), brine (30 mL), and concentrated. The resulting methyl
ester intermediate was taken up in 4:1 THF:MeOH (30 mL) and
aq. NaOH (25%, 5 mL) was added. After the reaction mixture was
stirred overnight, it was quenched with aq. HCl (2.4 M, 50 mL)
and extracted with EtOAc (2 ꢂ 100 mL). Combined organic extracts
were washed with brine (30 mL) and concentrated affording 9
(1.22 g, 100%). Data for 9: MS calc (MꢃH)ꢃ 301.1552, found
301.15; NMR (CDCl3) d 8.37 (s, 0.9H), 7.57 (d, 1H, J = 7.8 Hz), 7.33
(d, 1H, J = 7.3 Hz), 7.16 (m, 1H), 7.09 (m, 1H), 6.97 (s (br), 1H),
5.59 (s, 0.9H), 3.11 (br, 4H), 2.57 (br, 2H), 2.26 (t, 2H, J = 7.4 Hz),
1.53 (br, 2H), 1.31 (br, 2H), 1.14 (br, 2H).
10. To a solution of carboxylic acid 9 (75.4 mg, 0.249 mmol) and
amine 3a (149 mg, 0.276 mmol) in 2.0 mL dry DMF were added
EDCIꢁHCl (56.5 mg, 0.29 mmol), HOAt (41.8 mg, 0.31 mmol) and
DIPEA (0.13 mL, 0.75 mmol) and the resulting mixture was stirred
for 48 h. It was then diluted with EtOAc (30 mL), washed with conc.
aq. NaHCO3 (15 mL), 10% NaCl (15 mL), brine (15 mL), and concen-
trated. Purification by silica gel chromatography (4:1 EtOAc:hex-
anes ? EtOAc) afforded tBu ester-protected intermediate
(205 mg, 100%). Rf = 0.30 (4:1 EtOAc:hexanes).
This intermediate (153 mg) was taken up in dry HCl in dioxane
(4 M, 2.5 mL), stirred for 2 h, and concentrated. Purification by sil-
ica gel chromatography (10:1 DCM:MeOH) afforded 10 (31.8 mg,
22%). Data for 10: Rf = 0.18 (9:1 DCM:MeOH); MS calc: 766.88,
found (MꢃH)ꢃ 765.48; 1H NMR (acetone) d 8.34 (s, 0.4H), 8.32 (s,
0.5H), 8.15 (s, 0.7H), 7.67 (m,2H), 7.56 (d, 1H, J = 7.8 Hz), 7.47 (s
(br), 1H), 7.41 (m, 2H), 7.35 (d, 2H, J = 8.2 Hz), 7.27 (br, 2H), 7.18
(t, 1H, J = 5.5 Hz), 7.12 – 6.96 (m, 7H), 4.99 (m, 1H), 4.29 (t, 2H,
J = 6.2 Hz), 3.52–3.35 (m, 2H), 3.24 (t, 2H, J = 6.4 Hz), 3.18 (q, 2H,
J = 6.4 Hz), 3.07 (t, 2H, J = 7.6 Hz), 2.56 (t, 2H, J = 7.6 Hz), 2.14 (t,
2H, J = 7.6 Hz), 1.53 (t, 2H, J = 7.6 Hz), 1.43 (t, 2H, J = 7.6 Hz), 1.25
(m, 2H).
11. To a solution of carboxylic acid 9 (60.0 mg, 0.198 mmol) and
amine 4a (102 mg, 0.2 mmol) in 2.0 mL dry DMF were added
EDCIꢁHCl (58.4 mg, 0.305 mmol), HOAt (33.4 mg, 0.245 mmol)
and DIPEA (0.10 mL, 0.57 mmol) and the resulting mixture was
stirred overnight. It was then diluted with EtOAc (30 mL), washed
with conc. aq. NaHCO3 (15 mL), 10% NaCl (15 mL), brine (15 mL),
and concentrated. Purification by silica gel chromatography (19:1
DCM:MeOH) afforded tBu ester-protected intermediate (141 mg,
89%). Rf = 0.24 (10:1 DCM:MeOH).
GQAVEAQQHLLQLTVWGIKQLQARILAVEKK-amide, where bpy is
2,2-bipyridine-5-carboxylate, a bidentate ferrous iron chelator that
assures the trimeric structure of the NHR upon metal binding.
Underlined residues occur in WT HXB2 gp41. C-peptide C18e2.0-
FL (Ac-MTWBEWDREIBNYTSLIC, B = a-aminoisobutyric acid, WT
residues underlined) labeled with fluorescein at the cysteine resi-
due, was used to probe inhibitor binding. Quenching of C18e2.0-
FL occurred upon addition of FeII(env2.0)3, with a KD = 0.8
lM
and a minimal fractional fluorescence for the complex = 0.072. KI
was determined by measuring the dose dependent fluorescence
recovery in the presence of a competitive inhibitor. 7.2 lM Fe
(env2.0)3 (measured as concentration of HP binding sites, with
three equivalent sites per trimer) and 30 nM C18e2.0FL were used
in the competitive inhibition experiments, with serial dilution of
the inhibitors in the range 0.39–400 lM, in Tris-acetate buffer at
pH 7.0 and 4% DMSO.25 Concentration of dimeric inhibitors was
calculated as one molecule per dimeric unit. 12 points per binding
curve were measured in triplicate for each compound. Fractional
fluorescence data were fit to Eqs. (3) and (4) for 1:1 ligand:HP
binding (Section 3.3). For 1:2 ligand:HP binding, data were fit by
numerical solution (Mathcad, Mathsoft) of the equation
KD
½Rꢄ½Lꢄ
½RLꢄ
R þ L $ RL; KD
¼
; Lt ¼ ½Lꢄ þ ½RLꢄ
KI1
½Rꢄ½Iꢄ
R þ I $ RI; KI1
¼
; Rt ¼ ½Rꢄ þ ½RLꢄ þ ½RIꢄ þ 2½R2Iꢄ
ð1Þ
½RIꢄ
KI2
½RIꢄ½Rꢄ
RI þ R $ R2I; KI2
¼
; It ¼ ½Iꢄ þ ½RIꢄ þ ½R2Iꢄ
½R2Iꢄ
where Lt, Rt and It are the concentrations of probe, HP binding sites
in receptor Fe(env2.0)3, and inhibitor, respectively, KI1 and KI2 are
the inhibition constants for the first and second binding interac-
tions. The observed fractional fluorescence was calculated from
Eq. (3).
The fit of calculated to observed data was determined from the
sum of squares (Eq. (2)):
12
X
2
SSQ ¼
fmaxððjFobs;i ꢃ Fcalc;ij ꢃ riÞ; 0Þg
ð2Þ
i¼1
where Fobs,i and Fcalc,i are the fractional fluorescence and ri is the
standard deviation of each data point.
2.2.2. Surface plasmon resonance experiments
Experiments were carried out on a Biacore T100 instrument. A
CM5 sensor chip surface was activated by injecting 360 lL EDC/
NHS (freshly prepared by mixing 5 mM EDC and 5 mM NHS (1:1
This intermediate (78.5 mg) was taken up in dry HCl in dioxane
(4 M, 1.5 mL), stirred for 2 h, and concentrated. Purification by silica
v/v) in biacore HBS-EP buffer), followed by introduction of
150 lL PDEA (80 mM in 0.1 M sodium borate buffer, pH 8.5) at a

412
V. Sofiyev et al. / Bioorg. Med. Chem. 25 (2017) 408–420
flow rate of 20
l
L/min. The bioreceptor FeII(envC)3 (envC = bpy-
replication or cell-cell fusion, but measuring cell viability, using 3-
(4,5-dimethylthiazol-2-yl)-5-(3-carboxymethoxyphenyl)-2-(4-sul-
fophenyl)-2H-tetrazolium (MTS, CellTiter 96 reagent; Promega;
used for MAGI assays) or using a resazurin cell viability reagent
(Alamar Blue or Presto Blue, Life Technologies; used for cell-cell
fusion assays) following the manufacturers’ protocols.
GQAVEAQQHLLQLTVWGIKQLQARI(d)C-amide, WT HXB2-Env resi-
dues underlined) was then bound to the sensor chip surface by
injecting 150
pH 7.0) for 2–9 min at 10
was deactivated with 50
acetate buffer, 1 M NaCl, pH 4.5) followed by a wash with 200
of freshly prepared 1 mM ferrous ammonium sulfate at 20
min, to ensure that any ferrous ion leached at low pH was restored.
Any non-specifically bound material was removed with 50 L of
lL (5 l
M FeII(envC)3 in 25 mM Tris-acetate buffer,
lL/min. Unreacted disulfide surface
l
L L-cysteine (50 mM in 0.1 M sodium
lL
lL/
2.3. Computational docking
l
Ligand charges and low energy conformations were obtained
using OpenEye software35,36 (fixpka, molcharge, omega2-2.5.1.4
and szybki, OpenEye Scientific Software, Santa Fe, NM. http://
www.eyesopen.com). Compounds were docked into pdb structure
3P7K using Autodock Vina. A model of bivalent binding of the
dimer 7 (Fig. 3) was obtained using low energy conformers of
two partial structures, split at the edge of the PEG3 linker. Struc-
tures were docked and linker dihedrals were manually adjusted
to reconnect the two parts. The local geometry was then optimized
and clashes removed using the default minimization protocol in
Chimera (UCSF) (Potential energy of complex ꢃ12.4 ꢂ 103, RMSD
0.0244 Å over 2460 atoms).
500 mM NaCl at 50 lL/min. An equivalent reference surface was
generated using the same procedure as above excluding the FeII(-
envC)3 coupling step. Analytes were detected at serially increasing
concentrations from 0.098 to 50 lM in Hepes buffered saline
(20 mM Hepes, 150 mM NaCl) at pH 7.0 containing 1% DMSO, using
a 60 s association time and 300 s dissociation time. DMSO bracket-
ing was used to correct the refractive index. Analysis of sensorgram
data was carried out with Biacore Biaevaluation software.
2.2.3. Cell-cell fusion assay
Cell-cell fusion was measured using cell lines obtained through
the NIH AIDS Research and Reference Reagent Program, Division of
AIDS, NIAID, NIH. Target cells were TZM-bl cells (#8129, con-
tributed by J.C. Kappes, X. Wu and Tranzyme Inc.) expressing
CD4, CCR5 and CXCR4,26 and containing an integrated reporter
gene for firefly luciferase under control of HIV-1 LTR.27 Effector
cells were HL2/3 (#1294, contributed by B.K. Felber and G.N. Pav-
lakis) which produce HXB2 Env, Tat and Rev.28 Serially diluted
inhibitors were added to 96 well plates containing 25,000 TZM-
bl cells per well cultured overnight. 50,000 HL2/3 cells were added
per well, and fusion allowed to proceed for 6 h in reduced serum
medium (Gibco) with a final concentration of 1% DMSO. Luciferase
expression was measured using Luciferase Assay Reagent (Pro-
mega) according to the manufacturer’s instructions. Controls con-
taining 1% DMSO with and without HL2/3 cells were measured
for each compound, and experiments were performed in triplicate.
3. Results
3.1. Synthesis and description of compounds
Six compounds were prepared as shown in Fig. 2 and Schemes
1–3. There were two series, based on the two-subunit pep-
tidomimetic compounds
3 (H2N-[Trp]-Trp-OH) and 4 (H2N-
[Phe4Cl]-Trp-OH), designated ‘‘monomers”.
Dimers 7 and 8 had two peptidomimetic monomers connected
by a 22.5 Å succinate-PEG3-succinate spacer (PEG3 = triethylene
glycol), calculated to be sufficiently long to permit simultaneous
binding into two hydrophobic pockets of the HR1 trimer. A model
of 7 produced by docking into the HR1 trimer and energy mini-
mization (see ‘Materials and methods’) is shown in Fig. 3. The
two monomers reach into two HP’s of the trimer with the PEG lin-
ker units in mostly all trans conformation. Extended monomers 10
and 11 were substituted at the aniline with a hydrophobic alkyl tail
terminated by a Trp, with the goal of increasing interactions within
the HR1 groove beyond the HP. LC–MS and NMR figures of the final
compounds are shown in Supplementary Data Figs. S1 and S2.
2.2.4. Viral replication and attachment assays
Inhibition of HIV-1 replication was determined in CCR5- and
CXCR4-tropic MAGI antiviral assays and inhibition of HIV-1 attach-
ment/entry was determined in CCR5-tropic MAGI attachment
assays as previously described.29,30 HIV-1 isolates and cells were
obtained from the NIH AIDS Research and Reference Reagent
Program, Division of AIDS, NIAID, NIH, as follows: HIV-1 Ba-L from
Suzanne Gartner, Mikulas Popovic, and Robert Gallo.30,31 HIV-1 IIIB
from Robert C. Gallo.30,32 MAGI-CCR5 cells from Dr. Julie
Overbaugh.33,34 For MAGI antiviral assays, MAGI-CCR5 cells were
grown overnight in 96 well plates in DMEM supplemented with
10% FBS, using 10,000 cells per well. The following day the medium
was removed and compounds diluted in medium were added
(6 dilutions in triplicate at each dilution), followed by the addition
of either HIV-1 Ba-L (CCR5-tropic assay) or HIV-1 IIIB (CXCR4-tropic
assay) at approximately ten 50% tissue culture infective doses per
well (ꢀ10TCID50/well). Assay plates were incubated for 48 h, after
which medium was removed and HIV-1 Tat-induced b-Gal enzyme
expression was determined by chemiluminescence using Tropix
Gal-Screen (Applied Biosystems) according to the manufacturer’s
instructions. MAGI attachment assays were performed similarly,
but with a washout of unbound virus and compounds three (3)
hours post-infection. Assays were conducted at a serum concentra-
tion of 2%.
2.2.5. Cytotoxicity assay
The cytotoxic effect of the compounds was determined using
the identical cell culture procedure to that described above for viral
Fig. 2. Structures of peptidomimetic inhibitors used in this study.

V. Sofiyev et al. / Bioorg. Med. Chem. 25 (2017) 408–420
413
Monomers H2N-[Trp]-Trp-OH and H2N-[Phe4Cl]-Trp-OH were
weak CCF inhibitors (22–24 M) and were ineffective against
VCF. Activity improved to 1–2 M against CCF and 13–17
l
l
lM
against VCF with addition of the hydrophobic tail in compounds
10 and 11. Many studies indicate that increased hydrophobic char-
acter is consistent with increased fusion inhibitory activity,37
including the preference for aromatic side chains at positions R1,
R2 (and R3) of this peptidomimetic library (Fig. 2). The added
hydrophobic tail with the terminal Trp may interact in the HR1
groove of the hydrophobic gp41 protein, or may associate with
the lipid bilayer, orienting and concentrating the ligand in the bio-
logical milieu of the fusion reaction.
Dimerization of the inhibitors with a 22 Å PEG linker connecting
two monomers resulted in the best performing member of each
series, with compounds 7 and 8 having IC50’s < 1
lM against CCF
and 3–5 M against VCF. In this case, it is not expected that the
l
PEG linker would interact extensively with the membrane,20 sug-
gesting that additional HR1 interactions could form the basis of
the improved activity. We noted that the antiviral activity of 7
reached a plateau at ꢀ20% residual fusion activity of Ba-L virus
and ꢀ6% residual fusion activity of IIIB virus. The faster kinetics
of Ba-L fusion may limit the potency of the largest molecular
weight inhibitor 7.
3.3. Solution binding studies confirmed hydrophobic pocket binding
Fig. 3. Model showing the ability of dimer 7 to reach around the HR1 coiled coil
with the terminal -NH-[Trp]-Trp-OH units interacting in two hydrophobic pockets.
The surface representation of the HR1 (pdb 3P7K) has been made partly transparent
to reveal the docked aromatic rings of 7.
Next we wished to ascertain whether the 50-fold increase in
potency was a cooperative effect of dimers binding into two HP’s
of one HR1 trimer, as designed. We therefore examined whether
binding and kinetic data could discriminate between a monovalent
and bivalent binding mechanism. Affinity of the compounds for the
gp41 HP was determined by fluorescence using a competitive inhi-
bition assay that has been previously described.14,25 Briefly, the
hydrophobic pocket (receptor R) is formed by metal ion assisted
association between three HR1 peptides env2.0 spanning the
pocket sequence (HXB2 residues 560–584) that are covalently
linked to N-terminal 2,20-bipyridine (bpy). Ferrous ion binding in
1/3 stoichiometry to peptide results in spontaneous formation of
an HR1 trimeric coiled coil as a magenta-colored complex Fe
(env2.0)3. The fluorescence of HR2 probe peptide (HXB2 residues
3.2. Anti-fusion activity improved with compound elaboration
Activity of the two series of compounds was investigated in
cell–cell fusion (CCF) assays using effector cells expressing
CXCR4-tropic HXB2-Env, and in virus–cell fusion (VCF) assays
using laboratory adapted strains Ba-L (CCR5-tropic) and IIIB
(CXCR4-tropic). The results are shown in Fig. 4 and Table 1. CCF
was inhibited more readily than VCF, and augmenting the basic
monomer significantly improved anti-fusion activity. None of the
compounds exhibited any toxicity at the concentrations tested.
Scheme 1. Synthesis of two-subunit monomers 3 and 4.

414
V. Sofiyev et al. / Bioorg. Med. Chem. 25 (2017) 408–420
Scheme 2. Synthesis of dimers 7 and 8 from monomers 3 and 4.
Scheme 3. Addition of a hydrophobic tail to monomers 3 and 4, forming 10 and 11.
626–642) labeled with fluorescein at the C-terminus (L = C18-
e2.0FL) is quenched upon binding to Fe(env2.0)3, forming the basis
of the fluorescence assay. Compounds binding to the HP displace
the probe peptide with a concomitant increase in fluorescence.
Inhibition constants (KI) are determined by dose response mea-
surements. Data were fit to the equation
3.3.1. Avidity for the HP did not predict the order of observed biological
activity
In previous studies, we have shown a correlation between low
l
M binding to Fe(env2.0)3 and inhibition of HIV-fusion.5,6,23 Com-
petitive inhibition dose response curves for 3, 4, 7, 8, 10 and 11,
conducted using 7.2 Rt and 30 nM Lt, are shown in
Fig. 5A and B. Here too, we observed that low M inhibition in
l
M
l
KD
KD þ ½Rꢄ
the binding assay was commensurate with activity against HIV-
Env mediated fusion; however it was not completely correlated.
A higher concentration was required for 50% activity of dimer com-
pared to the corresponding extended monomer in the fluorescence
assay, while dimers were clearly more potent in biological assays.
Additionally, the slope of the binding curves was lower for the
bivalent inhibitors. Parameters FL, Rt and KI were determined from
dose response curves by fitting to Eqs. (1) and (2), and/or by
numerical simulation. Results are given in Table 2.
Fobs ¼ FRL þ ðFLÞ
ð3Þ
where KD is the dissociation constant of the Fe(env2.0)3-C18-e2.0FL
complex, and FL and FRL are the fluorescence of free and bound C18-
e2.0FL. Known parameters are
RL = 0.072.14
For 1:1 binding of ligand with HP’s on the trimer Fe(env2.0)3,
K
D = 0.8
lM
and fractional
F
the concentration of free HP, [R], was calculated analytically
according to the equation
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
3.3.2. Monomers and extended monomers could be fit to a 1:1 binding
model
Fluorescence data for monomers 3 and 4 were simulated by a 1
ligand:1 HP binding model, with the expected concentration of HP,
½Rꢄ ¼ fRt ꢃ It ꢃ KI þ ðRt ꢃ It ꢃ KIg=2
ð4Þ
Rt, It and Lt are the total concentrations of added receptor binding
sites, inhibitor and probe peptide respectively, and KI is the inhibi-
tion constant. Eq. (4) is an analytical approximation valid when Rt -
ꢅ Lt which is fulfilled by the experimental conditions. Eq. (3) was
also solved numerically for both 1:1 and 1:2 binding models,
requiring no approximations (‘Materials and methods’). The analyt-
ical approximation of 1:2 binding of ligand to two HP’s required
solution of a cubic equation and was not used.
Rt = 7.2 lM. Fits to the data for extended monomers 10 and 11
yielded Rt values that were 14–20% lower than those of the corre-
sponding monomer, which we interpreted as an outcome of the
low solubility of these hydrophobic inhibitors, causing aggregation
of the receptor–compound complex. Consequently, the aggregates
caused light scattering in the fluorescence assay, leading to values

V. Sofiyev et al. / Bioorg. Med. Chem. 25 (2017) 408–420
415
Fig. 4. Dose-response curves for each compound demonstrating percent inhibition of cell-cell fusion (A, B) and Ba-L viral infectivity (C, D) measured as relative luminescence
units (RLU) using a luciferase reporter. A, C. Compounds 3 (blue), 7 (black), 10 (red); B, D. Compounds 4 (blue), 8 (black) and 11 (red). Cell viability is shown in corresponding
open circles. Error bars are standard deviations of three measurements.
Table 1
Activity data against HIV-Env mediated fusion.*
Compounds
Cell–cell fusion
HXB2
Virus-cell fusion
CC50
Ba-L
IC50
IIIB
IC50
IC90
IC90
IC50
IC90
3
10
7
4
11
8
22
80
>50
13.7
3.6
>50
13.0
4.7
>50
>50
17.3
3.8
>50
14.3
5.1
>50
>50
1.3
0.6
24
2.2
0.4
15.1
6.3
49
5.6
4.4
31.1
>100
>50
26.5
18.3
30.5
70.0
>50
27.1
20.8
>100
>100
>50
>100
>100
*
3–6 repeats of each measurement, error is 10–20%, in lM.
for FL < 1 (Table 2). [Trp]-Trp containing peptidomimetic com-
pounds 3 and 10 were measurably more potent than [Phe4Cl]-
Trp compounds 4 and 11, consistent with their higher hydrophobic
the total number of dimer molecules. An excellent fit to the data
was in fact achieved using a 1:1 binding model with ꢀ40% reduc-
tion in Rt. These fits yielded KI’s of 2.3 lM and 8.5 lM for 7 and 8,
character. Observed inhibition constants were 7
l
M for 3, 15
l
M
respectively, 2–3-fold improved over the respective monomers.
The observed KI’s are assumed to be an average over monodentate
and bidentate binding. Best fit Rt values at ꢀ60% implicate some
aggregation as part of this model fitting, supported by the observa-
tion of FL values <1.
for 4. Adding the hydrophobic tail improved HP binding affinity
by a factor of 5–6, to 1.2 for 10 and 3.1 for 11, mirroring the results
obtained in fusion inhibition studies and further consolidating the
role that hydrophobic character and HP binding play in compound
activity.
The data could also be fit with the full complement of
Rt = 7.2
HP’s. Notably the binding constant for the first interaction,
KI = 20–21 M, was higher than the KI for the monomers 3 or 4,
lM using a model of 1:2 binding of one bivalent ligand:2
3.3.3. The data are compatible with a bivalent interaction of dimers
with Fe(env2.0)3
l
and likely reflects an entropic penalty due to the flexible PEG lin-
ker, while KI2 for the second interaction at the other end of the
PEG linker was more than an order of magnitude lower than for
the monomers, implying a cooperative binding effect or a local
concentrating effect due to the first interaction. There was no evi-
dence of aggregation from the FL values. The mathematical model-
ing does not specify whether the HP’s are on the same or different
trimers, since they are each considered independent and identical.
Using the experimental value of Rt = 7.2 lM, fluorescence data
for the bivalent inhibitors 7 and 8 could not be fit to a 1:1 binding
model. Instead, data fitting supported a model in which simultane-
ous binding of one bivalent ligand into two hydrophobic pockets
occurred (Table 2, Figs. 5C, D and 6). Since there are three HP’s
per trimer, the simplest model for full occupancy of the HP’s would
include a bidentate bound dimer and a monodentate bound dimer
(Fig. 6). This leaves only 67% of receptor binding sites available for

416
V. Sofiyev et al. / Bioorg. Med. Chem. 25 (2017) 408–420
Fig. 5. Fluorescence competitive inhibition dose response curves performed with 7.2
D show data fitting for dimers 7 and 8 to a 1:1 binding model (dashed line) and 1:2 binding model (bold solid line) assuming 7.2
l
M binding sites. A. 3 (blue), 7 (black), 10 (red) and B. 4 (blue), 8 (black), 11 (red). C and
M binding sites. The narrow solid lines show
l
an alternative fit to a 1:1 binding model with 40% reduced receptor concentration. RFU = relative fluorescence units, as a fraction of the fluorescence of probe plus compound
in the absence of receptor (FL).
Table 2
Results of fluorescence equilibrium binding assay for the hydrophobic pocket.y
Compound
Binding modelà
FL
Rt
(l
M)
Inhibition constant (
lM)
SSQ/10ꢃ3
3 (monomer)
10 (extended monomer)
7 (dimer)
1:1
1:1
1:1
1:1
1:2
1.08
0.85
0.79
0.80
0.95
7.2
6.2
7.2
4.3
7.2
KI = 7.2 1.1
KI = 1.2 0.2
KI = 1.0 0.3
2.13
2.07
10.11
0.014
0.928
§
KI = 2.3 0.3
KI1 = 20, KI2 = 0.2
4 (monomer)
11 (extended monomer)
8 (dimer)
1:1
1:1
1:1
1:1
1:2
1.00
0.85
0.99
0.90
1.05
7.2
5.5
7.2
4.5
7.2
KI = 15 2.6
KI = 3.4 0.18
KI = 5.3 1.5
KI = 8.5 1.3
0.258
0.066
18.7
0.596
0.168
§
KI1 = 21, KI2 = 1.3
The total HP concentration in the assay, Rt, was 7.2
l
M. Maximum fractional fluorescence FL < 1 could be due to light scattering.
y
Data were fit to Eqs. (3) and (4) or (1), and the sum of squares SSQ was calculated using Eq. (2), see text.
The binding model is defined as the ratio of ligand:HP.
Poor fit.
à
§
It is unlikely that a ligand with the limited potency of 7 or 8 could
bility, given that the inhibitors are quite hydrophobic. We there-
fore turned to surface binding studies to evaluate inhibitor
interactions.
bridge two trimers in solution. Since the data fitting required every
available HP, it is likely that a second bivalent inhibitor in the third
site of each trimer engages in an equilibrium exchange for biden-
tate binding with the first bivalent inhibitor, as illustrated in
Fig. 6. Data resolution does not permit us to distinguish this
dynamic model from the static scenario described above.
A third option for data fitting involves monodentate binding for
all of the dimers, with a significant degree of receptor aggregation
occurring during the assay. We cannot formally rule out this possi-
3.4. Surface binding studies yielded kinetics of binding
Hydrophobic pocket binding was evaluated using surface plas-
mon resonance (SPR) to measure association and dissociation
kinetics. Analysis of the fluorescence data and previous literature
observations12 suggested that bivalent binding should display an

V. Sofiyev et al. / Bioorg. Med. Chem. 25 (2017) 408–420
417
Fig. 6. A model of bidentate binding of dimers to the HR1 trimer. The HP’s, shown in orange, are arranged on a threefold axis of symmetry. Each green diamond shape
represents an aromatic side chain of the 2-subunit peptidomimetic monomer. Two equivalent dimers are depicted in green and light green.
altered kinetic profile compared to 1:1 binding, permitting us to
verify our interpretation of the solution equilibrium binding stud-
ies. Furthermore, we were curious to know whether kinetics, espe-
cially off-rate could be correlated more effectively to observed
biological potency than KI, given that the target is a transient inter-
mediate structure. Previous studies have shown that on-rate can
limit the effectiveness of a large protein inhibitor,38 and high
potency is accompanied by a slow off-rate.21,39
is likely due to the increased size and flexibility of the dimers,
while the slow off-rate is consistent with dual binding interactions
with Fe(envC)3 where there is a higher probability of rebinding
before the partner can dissociate. The extended monomers 10
and 11 have a large hydrophobic binding capacity and a flexible
hydrophobic tail, which translated into intermediate values for ka
and kd compared to the monomers and dimers, as well as lower
KI’s observed by fluorescence. Changes of the same order of magni-
tude in both ka and kd resulted in little discrimination in KD over
the dataset. Values for KD obtained by the SPR experiment are of
limited accuracy due to low precision in ka and the inability to
reach saturating concentrations of analyte.
SPR experiments were conducted on a Biacore CM5 chip using a
modified version of the receptor peptide Fe(envC)3. The peptide
envC is terminated by a D-cysteine residue, which orients the thiol
group along the axis of the peptide and permits vertical anchoring
of the receptor to the surface groups. We have previously used this
receptor in electrochemical and SPR binding studies of peptide and
small molecule–gp41 HP interactions.40 The design creates a
homogeneous ligand surface with open binding sites on the three
faces of the threefold symmetrical receptor, such that dimers could
simultaneously bind into two pockets, if that is indeed their
mechanism.
3.4.1. A bivalent model gave an improved fit to the SPR data for the
dimers
We observed improved fits to the data for the dimers using a
bivalent analyte binding model provided by the Biacore software,
(Figs. 7D and S4D). The results of the data fitting are shown in
Table 4.
We achieved immobilization responses in the range 2460–5670
units, corresponding to approximately one Fe(envC)3 molecule per
32.6–8.3 nm2, respectively, or effective concentrations between 2.6
and 6 mM.41,42 A reference channel was prepared without receptor,
and signals from this channel were subtracted out to yield SPR sen-
sorgrams that reflected interactions with Fe(envC)3. Specificity of
analyte binding was confirmed by observing both analyte and Fe
(envC)3 concentration dependent responses (Supplementary Data,
The observation that kd1 > kd2 suggested that bivalent interac-
tions make an important contribution to slowing the off-rate,
and therefore to potency against fusion.
3.4.2. Kinetic rates correlated with inhibitor potency
The relatively small change in KI between monomers and
dimers can be explained by the high entropic penalty due to con-
straining the flexible structures. 7 and 8 have 35 and 33 rotatable
bonds, respectively, 19 of which involve the flexible succinimidyl-
PEG3 linker.43 For the extended monomers, observed KI were lower
and ka’s higher than for the dimers, confirming that the additional
hydrophobic contribution to enthalpy more than offset the entro-
pic penalty due the 18–19 added rotatable bonds. The biophysical
properties of the dimers provided evidence that slow kinetics was
a more important quality for inhibiting fusion than the KI, since 7
and 8 were more than an order of magnitude more potent as fusion
inhibitors than 3 and 4, and several times more potent than 10 and
11. They also explain the lower correlation observed between
fusion inhibition and KI compared to kd, as shown in Fig. 8. We note
that ka also correlated with inhibitor potency (not shown).
Fig. S3). Typically, below ꢀ10
lM, analyte responses were well
below 100 RU, suitable for kinetic analysis. At higher concentra-
tions, most compounds gave disproportionately high analyte
responses, likely due to non-specific binding or aggregation. These
sensorgrams were not included in the analysis.
We measured the association rate (ka), dissociation rate (kd) and
rate constant (KD = kd/ka) for each compound at various concentra-
tions, by analyzing sensorgrams using Langmuir 1:1 binding and
1:2 binding kinetic analysis. The maximum analyte response,
RUmax, occurring at saturating concentrations of analyte, was not
obtained experimentally due to non-specific aggregation, and
was determined from a good fit to SPR data for compound 7.
Remaining sensorgrams were fit by fixing RUmax proportionately
according to the molecular weight of the analyte. Uncertainty in
RUmax led to ꢀ30% uncertainty in ka, but still allowed us to com-
pare kinetics within the compound set. Off-rate kd was less depen-
dent on RUmax. The results are reported in Table 3, shown for the
[Phe4Cl]-Trp series in Fig. 7 and for the [Trp]-Trp series in Supple-
mentary Data Fig. S4. The data clearly reveal significantly slower
kinetics for bivalent compounds 7 and 8 compared to monovalent
compounds 3 and 4, with intermediate rates observed for the
extended monomers 10 and 11. Monomers 3 and 4 had fast on-
4. Discussion
In this study, we evaluated the binding affinity, kinetics and
biological activity of peptidomimetic inhibitors that were modified
to extend interactions with gp41 beyond the HP. The core units
were arylalkoxy-amino acid helical mimetics supporting side
chains at the equivalent of i and i + 3 positions of a helix. These
were 3 (H2N-[Trp]-Trp-OH) and 4 (H2N-[Phe4Cl]-Trp-OH). Previ-
ous studies have identified these molecules as moderate
hydrophobic pocket binders and fusion inhibitors.23,24 Onto these
monomeric units, we grafted a largely flexible hydrophobic tail
that could extend the interaction of the inhibitor into the long
and off-rates, on the order of ka ꢀ 6 * 104 Mꢃ1 sꢃ1 and kd ꢀ 0.8 sꢃ1
.
Both on-rate and off-rate for dimers were an order of magnitude
slower, with ka ꢀ 3000 Mꢃ1 sꢃ1, and kd ꢀ 0.05 sꢃ1. The slow on-rate

418
V. Sofiyev et al. / Bioorg. Med. Chem. 25 (2017) 408–420
Table 3
Results of SPR 1:1 binding kinetic analysis (Biacore).
*
v2
ka (Mꢃ1 sꢃ1
)
kd/(sꢃ1
)
kd/ka (M)
RUmax
(RU2)
3 (monomer)
10 (ext. monomer)
7 (dimer)
4 (monomer)
11 (ext. monomer)
8 (dimer)
5.10e4
1.10e4
0.37e4
6.70e4
1.37e4
0.18e4
0.834
0.093
0.060
0.820
0.186
0.037
1.6eꢃ5
1.4eꢃ5
1.6eꢃ5
1.2eꢃ5
1.4eꢃ5
2.1eꢃ5
59
95
160
56
93
4.54
5.58
2.62
5.45
2.90
1.68
157
*
RUmax was fixed during global curve fitting of sensorgrams at 3, 4 or 5 concentrations for monomers, extended monomers or dimers, respectively.
Fig. 7. SPR sensorgrams obtained with immobilized Fe(envC)3. The chip was exposed to compounds for 60 s, followed by 300 s dissociation time. Data were collected from
ꢃ50 to 360 s and are shown expanded in the figure between ꢃ10 and 150 s. A. Compound 4, at concentrations 6.25, 3.13, 1.56 M; B. Compound 11, at concentrations 6.25,
3.13, 1.56, 0.78 M; C, D. Compound 8, at concentrations 6.25, 3.13, 1.56, 0.78, 0.39 M. Data were fit to a 1:1 binding kinetic model (A–C) or assuming bivalent analyte (D).
l
l
l
Residuals between observed and calculated data are shown below each sensorgram.
Table 4
Biacore fitting of dimer data to a bivalent binding model.
v
2
ka1 (Mꢃ1 sꢃ1
)
kd1 (sꢃ1
)
KD1 (M)
ka2 (RUꢃ1 sꢃ1
)
kd2 (sꢃ1
)
RUmax (RU)
(RU2)
7
8
2691
1489
0.0869
0.0638
3.2eꢃ5
4.3eꢃ5
4.54eꢃ5
2.27eꢃ5
0.0145
0.0033
160
157
2.95
0.44
hydrophobic groove (compounds 10 and 11). Alternative struc-
tures were created with a flexible PEG linker supporting two
monomeric units at either end, which could purportedly interact
simultaneously with hydrophobic pockets on two faces of the
coiled coil. These were compounds 7 (dimer of -HN-[Trp]-Trp-
OH) and 8 (dimer of -HN-[Phe4Cl]-Trp-OH). Fusion inhibition stud-
ies revealed that monomers were poor to moderate inhibitors,
monomers extended with a hydrophobic tail were more potent,
and the bivalent compounds with a monomer unit at each end of
the long 22.5 Å spacer were the most potent fusion inhibitors.
KI’s were obtained by equilibrium binding studies using fluores-
cence, and kinetic data were obtained by SPR. Data for 3, 4, 10,
and 11 conformed to a 1:1 binding model, while data for the
dimers 7 and 8 were commensurate with bidentate binding. Opti-
mal fits of the dimer data were obtained with a 1:2 ligand:HP bind-
ing model (Figs. 5 and 7), yielding two inhibition constants KI1 and

V. Sofiyev et al. / Bioorg. Med. Chem. 25 (2017) 408–420
419
Fig. 8. Correlations between kinetic data and fusion inhibition. A. Dissociation rate kd and B. Inhibition constant KI plotted against cell-cell fusion IC50. Data were obtained
from Tables 1–3.
KI2 for the first and second binding event, or by proportionate
reduction in binding site concentration in a 1:1 model in which a
maximum of two dimer molecules associate with each HR1 trimer,
yielding a single average <KI>. Significantly reduced dissociation
constants were obtained compared to monomers, presumably
due to rebinding events. The data could be interpreted as mon-
odentate binding of dimer molecules, although this required the
assumptions that 40% of the receptor aggregated during the bind-
ing assay, and that slow dissociation was simply a consequence of
the large number of degrees of freedom in the dimers.44
The variation of KI’s across the data set was lower than the dif-
ferences in their biological activity. The 6-fold reduction in KI
caused by adding a flexible hydrophobic tail in 10 and 11 improved
biological activity by a factor of 10–15, while dimers 7 and 8 were
40–60-fold more active than the monomers against fusion, despite
only 2–3-fold changes to <KI>. Examination of the kinetic data
revealed that dissociation rate kd was better correlated to biologi-
cal activity than KI. Dimers dissociated ꢀ20 times more slowly
than monomers.
ka also decreased with increasing size of the inhibitors, a conse-
quence of both size and the loss of conformational entropy upon
binding.44 Consequently, the range of KI = kd/ka remained small.
This is the enthalpy–entropy compensation effect, whereby struc-
tural mobility provided for enthalpic contributions through
enhanced conformational search space and number of interactions,
but simultaneously resulted in a large entropic penalty from inhi-
bitor rigidification upon binding.43 In this context, even the mono-
meric units with 8–9 rotatable bonds have poor avidity. The
number of rotatable bonds increased to 18–19 for 11 and 10, and
33–35 for 8 and 7, clearly a limiting factor in the potency of the
resulting molecules. It also affected the degree of cooperativity of
bidentate binding in the dimers. Slight cooperativity was apparent
from the lower average inhibition constant of dimers compared to
monomers and the observation of KI2 << KI1, and kd2 < kd1 in biva-
lent model fitting, but it was significantly dampened by the length
and flexibility of the spacer between the two monomers, resulting
cause an off-pathway effect.39 Another possibility is that the local
environment of gp41 near the membrane alters the behavior of
the inhibitors due to association of hydrophobic moieties with
lipids. Addition of a lipid-binding moiety to a peptide is known
to enhance peptide potency by orienting and concentrating the
peptide at the site of action.45–47 Yet a third possible explanation
is that the cooperativity does not arise from bidentate interactions
with one HR1 trimer, but rather by the bridging of two trimers on
the cell or virion surface. This was suggested to be the mechanism
of 20–100 fold increased potency of bivalent sialosides against
Influenza Virus mediated hemagglutination, where, similar to our
observations, a minimal increase in binding affinity to an isolated
trimer occurred for the dimers.48 This mechanism appears some-
what less likely due to the low receptor density on the surface of
HIV, where there are estimated to be ꢀ14 Env trimers per virion,
compared to ordered arrays of 500–1000 HA trimers per Influenza
virion.48,49 However, if more than one HIV Env trimer congregates
to form the fusion pore, it could provide a window of susceptibility
for a dimer bridge. The number is not known and has been vari-
ously thought to be one or a few.50
5. Conclusion
In conclusion, we have observed ꢀ50-fold improvement in
fusion inhibitory activity upon dimerization of small pep-
tidomimetic two-subunit compounds containing Trp and Phe4Cl
residues. Off-rate was found to be a more sensitive predictor of
fusion inhibitory activity than KI. Our data are consistent with
bidentate binding of the dimer molecules to the gp41 trimer, with
slow dissociation rates due to rebinding events, notwithstanding
the low cooperativity of dimer binding. The failure of solution
binding to recapitulate the cooperative effect observed in fusion
inhibition is likely due to linker flexibility and to the effect of kinet-
ics on the inhibitory potency of the dimers. By way of comparison,
a range of 10–300 fold improvement in fusion inhibitory activity
was found by dimerizing D-peptide inhibitors, a variation that
was related to the viral strain-dependent rate of fusion.20,21
Trimerization of D-peptides improved activity by a factor of 1000
or more.20,21 The mechanism was assumed to be due to tridentate
binding to one HR1 trimer, but no specific experiments confirmed
this binding mechanism. We have shown that multimerization is
an applicable strategy for improving non-peptide fusion inhibitors.
Local modifications to the structures to improve their thermody-
namic parameters, as well as globally increasing multivalency,
could yield orders of magnitude increase in potency. Importantly,
these inhibitors utilize specific HP interactions rather than non-
in a large KI1 and low ka1
.
By contrast, there was significant cooperativity observed in
fusion inhibitory activity of the dimers, suggesting that the biolog-
ical milieu is not emulated by the biophysical setup. A factor of 2–3
in <KI> became a factor of 40–60 in IC50. One possible explanation
is that the transient nature of the biological target accentuates the
role of inhibitor kinetics in potency. Previous studies have shown
that potency of HP binding inhibitors is highly dependent on the
rate of the fusion reaction.21 The low off-rate for the dimers could
result in gp41 molecules trapped in the inhibitor-bound state or

420
V. Sofiyev et al. / Bioorg. Med. Chem. 25 (2017) 408–420
specific hydrophobic or electrostatic interactions51 which can yield
low M binding to gp41 HR1 but are difficult to optimize. In this
regard, it is likely that non-specific hydrophobic interactions will
limit our ability to optimize inhibitors such as 10 and 11. Our
results demonstrate proof of principle of a method to enhance
avidity of non-peptide fusion inhibitors by focusing on multiple
HP interactions.
13. Choi S-K. Synthetic Multivalent Molecules: Concepts and Biomedical
Applications. John Wiley & Sons. 2004.
l
14. Cai L, Gochin M. Antimicrob Agents Chemother. 2007;51:2388.
15. Cole JL, Garsky VM. Biochemistry (Mosc). 2001;40:5633.
16. Oishi S, Ito S, Nishikawa H, et al. J Med Chem. 2008;51:388.
17. Liu S, Jing W, Cheung B, et al. J Biol Chem. 2007;282:9612.
18. Ling Y, Xue H, Jiang X, Cai L, Liu K. Bioorg Med Chem Lett. 2013;23:4770.
19. Eckert DM, Malashkevich VN, Hong LH, Carr PA, Kim PS. Cell. 1999;99:103.
20. Welch BD, VanDemark AP, Heroux A, Hill CP, Kay MS. Proc Natl Acad Sci USA.
2007;104:16828.
21. Welch BD, Francis JN, Redman JS, et al. J Virol. 2010;84:11235.
22. Shaginian A, Whitby LR, Hong S, et al. J Am Chem Soc. 2009;131:5564.
23. Whitby LR, Boyle KE, Cai L, Yu X, Gochin M, Boger DL. Bioorg Med Chem Lett.
2012;22:2861.
Acknowledgements
This work was supported by NIH – United States grants
GM087998 and AI122847 to MG and internal funds from Touro
University California and Southern Research Institute. Molecular
graphics images were produced using the UCSF Chimera package
from the Resource for Biocomputing, Visualization, and Informatics
at the University of California, San Francisco (supported by NIH P41
RR-01081). The authors are grateful to OpenEye Scientific Software
for the academic license.
24. Gochin M, Whitby LR, Phillips AH, Boger DL.
2013;27:569.
J Comput Aided Mol Des.
25. Gochin M. Assay Drug Dev Technol. 2012;10:407.
26. Derdeyn CA, Decker JM, Sfakianos JN, et al. J Virol. 2000;74:8358.
27. Wei X, Decker JM, Liu H, et al. Antimicrob Agents Chemother. 2002;46:1896.
28. Ciminale V, Felber BK, Campbell M, Pavlakis GN. AIDS Res Hum Retroviruses.
1990;6:1281.
29. Lackman-Smith C, Osterling C, Luckenbaugh K, et al. Antimicrob Agents
Chemother. 2008;52:1768.
30. Kagiampakis I, Gharibi A, Mankowski MK, et al. Antimicrob Agents Chemother.
2011;55:264.
31. Popovic M, Read-Connole E, Gallo RC. Lancet. 1984;2:1472.
32. Popovic MS, Gartner E, Read-Connole B, Beaver M, Reitz M. In: Valette L, Girard
M, eds. Retroviruses of Human AIDS and Related Animal Diseases. Marnes-La-
Coquette, France: Pasteur Vaccins; 1989:219.
33. Chackerian B, Long EM, Luciw PA, Overbaugh J. J Virol. 1997;71:3932.
34. Ratner L, Haseltine W, Patarca R, et al. Nature. 1985;313:277.
35. Hawkins PC, Nicholls A. J Chem Inf Model. 2012;52:2919.
36. Hawkins PC, Skillman AG, Warren GL, Ellingson BA, Stahl MT. J Chem Inf Model.
2010;50:572.
A. Supplementary data
Supplementary data associated with this article can be found, in
the online version, at http://dx.doi.org/10.1016/j.bmc.2016.11.010.
References
37. Gochin M, Zhou G. Curr Top Med Chem. 2011;11:3022.
38. Steger HK, Root MJ. J Biol Chem. 2006;281:25813.
39. Kahle KM, Steger HK, Root MJ. PLoS Pathog. 2009;5:e1000674.
40. Hou Y, Gochin M. Anal Chem. 2008;80:5924.
1. Chan DC, Chutkowski CT, Kim PS. Proc Natl Acad Sci USA. 1998;95:15613.
2. Buzon V, Natrajan G, Schibli D, Campelo F, Kozlov MM, Weissenhorn W. PLoS
Pathog. 2010;6:e1000880.
41. Muller KM, Arndt KM, Pluckthun A. Anal Biochem. 1998;261:149.
42. Quinn JG, O’Neill S, Doyle A, et al. Anal Biochem. 2000;281:135.
43. Ferrante A, Gorski J. J Mol Biol. 2012;417:454.
44. Pan AC, Borhani DW, Dror RO, Shaw DE. Drug Discov Today. 2013;18:667.
45. Wexler-Cohen Y, Shai Y. PLoS Pathog. 2009;5:e1000509.
46. Wexler-Cohen Y, Shai Y. FASEB J. 2007;21:3677.
3. Weissenhorn W, Dessen A, Calder LJ, Harrison SC, Skehel JJ, Wiley DC. Mol
Membr Biol. 1999;16:3.
4. Gallo SA, Finnegan CM, Viard M, et al. Biochim Biophys Acta. 2003;1614:36.
5. Zhou G, Sofiyev V, Kaur H, et al. J Med Chem. 2014;57:5270.
6. Zhou G, Wu D, Snyder B, Ptak RG, Kaur H, Gochin M. J Med Chem. 2011;54:7220.
7. Liu S, Wu S, Jiang S. Curr Pharm Des. 2007;13:143.
8. Wang Y, Lu H, Zhu Q, Jiang S, Liao Y. Bioorg Med Chem Lett. 2010;20:189.
9. Naider F, Anglister J. Curr Opin Struct Biol. 2009;19:473.
10. He Y, Xiao Y, Song H, et al. J Biol Chem. 2008;283:11126.
11. Matthews T, Salgo M, Greenberg M, Chung J, DeMasi R, Bolognesi D. Nat Rev
Drug Discov. 2004;3:215.
47. Ingallinella P, Bianchi E, Ladwa NA, et al. Proc Natl Acad Sci USA. 2009;106:5801.
48. Glick GD, Toogood PL, Wiley DC, Skehel JJ, Knowles JR.
1991;266:23660.
J Biol Chem.
49. Schiller J, Chackerian B. PLoS Pathog. 2014;10:e1004254.
50. Wilen CB, Tilton JC, Doms RW. Cold Spring Harb Perspect Med. 2012;2.
51. Gochin M, Cai L. J Med Chem. 2009;52:4338.
12. Smith JA, Amagasu SM, Hembrador J, et al. Mol Pharmacol. 2006;69:921.