JVI Accepted Manuscript Posted Online 6 April 2016 J. Virol. doi:10.1128/JVI.03246-15 Copyright © 2016, American Society for Microbiology. All Rights Reserved.
1 2 3 4 5
7
Optimization of the Solubility of HIV‐1‐Neutralizing Antibody 10E8 through Somatic
8
Variation and Structure-Based Design
9 10
Young D. Kwona, Ivelin S. Georgieva, Gilad Ofeka, Baoshan Zhanga,
11
Mangaiarkarasi Asokana, Robert T. Bailera, Amy Baoa, William Carusoa, Xuejun Chena, Misook
12
Choea, Aliaksandr Druza, Sung-Youl Koa, Mark K. Loudera, Krisha McKeea, Sijy O’Della,
13
Amarendra Pegua, Rebecca Rudicella, Wei Shia, Keyun Wanga, Yongping Yanga, Mandy Algera,
14
Michael F. Bendera, Kevin Carltona, Jonathan W. Coopera, Julie Blinnb,
15
Joshua Eudaileyb, Krissey Lloydb, Robert Parksb, S. Munir Alamb, Barton F. Haynesb, Neal N.
16
Padtec, Jian Yuc, David D. Hoc, Jinghe Huangd, Mark Connorsd, Richard M. Schwartza,
17
John R. Mascolaa*, and Peter D. Kwonga*
18 19 20 21 22 23 24 25 26
a
Vaccine Research Center, NIAID, National Institutes of Health, Bethesda, MD 20892, USA
b
Duke University Human Vaccine Institute, Departments of Medicine, Surgery, Pediatrics and Immunology, Duke University School of Medicine, and the Center for HIV/AIDS Vaccine Immunology-Immunogen Discovery at Duke University, Durham, North Carolina 27710, USA. c
The Aaron Diamond AIDS Research Center, Rockefeller University, New York, NY 10016
d
HIV-Specific Immunity Section, Laboratory of Immunoregulation, National Institute of Allergy and Infectious Diseases, and National Institutes of Health, Bethesda, Maryland 20892, USA.
27 28
*Correspondence should be addressed to John R. Mascola (
[email protected]) and Peter D.
29 30
Kwong (
[email protected])
Downloaded from http://jvi.asm.org/ on January 6, 2019 by guest
6
Extraordinary antibodies capable of near pan-neutralization of HIV-1 have been identified.
32
One of the broadest is antibody 10E8, which recognizes the membrane-proximal external
33
region (MPER) of HIV-1 and neutralizes >95% of circulating HIV-1 strains. If delivered
34
passively, 10E8 might serve to prevent or to treat HIV-1 infection. Antibody 10E8,
35
however, is markedly less soluble than other antibodies. Here, we describe the use of both
36
structural biology and somatic variation to develop optimized versions of 10E8 with
37
increased solubility. From the structure of 10E8, we identified a prominent hydrophobic
38
patch; reversion of four hydrophobic residues in this patch to their hydrophilic germline
39
counterparts resulted in a ~10-fold decrease in turbidity. We also used somatic variants of
40
10E8, identified previously by next-generation sequencing, to optimize heavy and light
41
chains; this process yielded several improved variants. Of these, variant 10E8v4 with 26
42
changes versus the parent 10E8 was the most soluble, with a paratope we showed
43
crystallographically to be virtually identical to that of 10E8, a potency on a panel of 200-
44
HIV-1 isolates also similar to that of 10E8, and a half-life in rhesus macaques of ~10 days.
45
An anomaly in 10E8v4 size-exclusion chromatography that appeared to be related to
46
conformational isomerization was resolved by engineering an inter-chain disulfide. Thus,
47
by combining a structure-based approach with natural variation in potency and solubility
48
from the 10E8 lineage, we successfully created variants of 10E8, which retained the potency
49
and extraordinary neutralization breadth of the parent 10E8, but with substantially
50
increased solubility.
51
2
Downloaded from http://jvi.asm.org/ on January 6, 2019 by guest
31
52
Importance
54
Antibody 10E8 could be used to prevent HIV-1 infection, if manufactured and delivered
55
economically. It suffers, however, from issues of solubility, which impede manufacturing. We
56
hypothesized that the physical characteristic of 10E8 could be improved through rational design,
57
without compromising breadth and potency. We used structural biology to identify hydrophobic
58
patches on 10E8, which did not appear to be involved in 10E8 function. Reversion of
59
hydrophobic residues in these patches to their hydrophilic germline counterparts increased
60
solubility. Next, clues from somatic variants of 10E8, identified by next generation sequencing,
61
were incorporated. A combination of structure-based design and somatic variant optimization led
62
to 10E8v4, with substantially improved solubility and similar potency versus the parent 10E8.
63
The co-crystal structure of antibody 10E8v4 with its HIV-1 epitope was highly similar to that
64
with the parent 10E8, despite 26 alterations in sequence and substantially improved solubility.
65
Antibody 10E8v4 may be suitable for manufacturing.
66
3
Downloaded from http://jvi.asm.org/ on January 6, 2019 by guest
53
67
Introduction Over the last 5 years, extraordinary antibodies have been identified capable of effectively
69
neutralizing HIV-1 (7, 8, 10, 13, 19, 34, 37, 38, 41). In addition to serving as potential templates
70
for an antibody-based HIV-1 vaccine, the passive delivery of these antibodies could be used to
71
prevent HIV-1 infection or to treat those infected with HIV-1 therapeutically (2, 5, 24, 36).
72
Such passive use of antibodies, however, would require their economical manufacturing
73
and delivery, and HIV-1-neutralizing antibodies often have characteristics which make their
74
manufacture less than optimal. Antibody 10E8, which targets the membrane-proximal external
75
region (MPER) of the gp41 subunit (14, 21), is one of these: it neutralizes 98% of a panel of 181
76
diverse HIV-1 isolates (14). Despite this extraordinary breadth, its poor solubility impedes
77
manufacturing. Other MPER antibodies, such as 2F5 and 4E10 (4, 38, 43, 44) , however, have
78
greater solubility, suggesting that the poor solubility of 10E8 is not intrinsic to its function and
79
could be improved.
80
Here we use a combination of structural biology and somatic variant optimization to
81
improve the solubility of antibody 10E8. We hypothesized that the reduced solubility of antibody
82
10E8 reflected the aggregation of hydrophobic surfaces. From the structure of 10E8, we
83
identified hydrophobic patches, which did not appear to be required for function, and reverted
84
residues in these patches to their hydrophilic germline counterparts. Similarly we identified
85
somatic variants which were more soluble, but less potent than the parent 10E8. In these variants,
86
we tested somatic alterations in the mature 10E8 that appeared to be of functional relevance by
87
altering residues in more soluble but less potent variants to their counterparts in the somatically
88
mature 10E8. Overall, we created several variants (Table S1), with increased solubility, all of
89
which showed no poly-reactivity and retained the breadth and potency of the parent 10E8. Of
4
Downloaded from http://jvi.asm.org/ on January 6, 2019 by guest
68
these, antibody 10E8v4 appeared to be the most soluble, and we determined its co-crystal
91
structure with its MPER epitope. We also characterized the poly-reactivity of 10E8v4, its
92
bioavailability in mice and rhesus macaques, and its behavior on size-exclusion chromatography
93
(SEC). An anomaly in SEC behavior appeared to be related to slow conformational
94
isomerization, so we engineered a disulfide linking the heavy and light chains of 10E8 to resolve
95
this issue. Our findings show how a combination of structural biology and somatic variant
96
optimization can be used to improve the manufacturing characteristics of an antibody, with
97
10E8v4 or its disulfide-locked variant potentially suitable for manufacturing.
98 99
MATERIALS AND METHODS
100
Antibody Expression and Purification. Mammalian codon-optimized genes encoding either an
101
antibody heavy chain or a light chain were synthesized and cloned into mammalian expression
102
vector pVRC8400 (VRC, NIAID, Bethesda, MD). For small scale preparation, 50 μg of antibody
103
heavy chain and 50 μg of light chain plasmid DNAs were combined in 5 ml of Opti-MEM
104
medium (Invitrogen, CA), then mixed with 5 ml of transfection medium containing 0.27 ml of
105
ExpiFectamine 293 transfection reagent (Invitrogen, CA) in Opti-MEM medium. The complex
106
of DNAs and ExpiFectamine 293 transfection reagent was incubated for 20 minutes at room
107
temperature before mixing with 80 ml of Expi 293S cell culture (2.5 x 106 cells/ml) in a 250-ml
108
shaking flask. The transfected cell culture was returned to suspension incubation for 24 hours at
109
37°C, 8% CO2 and 125 rpm, and then fed with 10 ml of the antibody expression enhancement
110
medium Ab Boost (ABI, VA). Six days post transfection, the supernatant was harvested by
111
centrifugation and filtered through 0.22 µm filters. The antibody IgG was captured by an affinity
112
column of protein A (Protein A Plus Agarose, Thermo Scientific, Rockford, IL), and further
5
Downloaded from http://jvi.asm.org/ on January 6, 2019 by guest
90
113
purified by a size exclusion column (Superdex 200, GE Healthcare). Purified antibodies were
114
dialyzed against 1X phosphate-buffered saline (PBS), and characterized with SDS-PAGE.
115
Structure-Based Engineering of Antibody 10E8 Variants:
117
Identification of hydrophobic patches. To identify hydrophobic patches, we used the DSSP
118
program (16) to calculate solvent accessible surface area (SASA) for each antibody residue.
119
Hydrophobic residues with SASA of more than 20 Å2 that were not part of the known paratope
120
and that were not deemed to be essential for the stability of the paratope, the heavy-light chain
121
interface, or other antibody structural elements, were selected for further analysis. Candidate
122
mutations were identified using the OSPREY protein design suite of programs (11), as well as
123
from next-generation sequencing (NGS) data.
124
Identification of functionally important somatically altered residues. To identify regions of
125
10E8, somatically altered and important for neutralization, we aligned sequences of more
126
potently neutralizing somatic variants with that of less potently neutralizing variants, selected
127
residues that were in close proximity to the MPER epitope, if different, swapped the
128
corresponding residues in combination (e.g. single, double, triple, or quadruple mutations) and
129
tested neutralization potency against a 9-virus panel. The 9-virus panel was selected to include
130
strains representing (i) the spectrum of neutralization sensitivity to wildtype 10E8, and (ii)
131
diverse HIV-1 clades.
132
Creation of a disulfide-locked CDR H3. To create a disulfide-locked CDR H3, we examined
133
the gp41 MPER peptide bound-10E8v4 Fab structure and identified the Cα of Tyr100e in the
134
CDR H3 region of the heavy chain and the Cα of Ser30 in the light chain as being separated by
6
Downloaded from http://jvi.asm.org/ on January 6, 2019 by guest
116
5.8 Å , close to the optimal Cα-Cα distance for a disulfide bond. Therefore we replaced these
136
residues with cysteines to form a disulfide bond.
137
Nomenclature of designs. Sequences for structure-based designs are shown in Table S1. The
138
nomenclature is as follows:
139
HC6-S74Y: somatic variant heavy chain HC6 with S74Y mutation, HC6-S74Y-DKTT: heavy
140
chain HC6-S74Y with L72D, I75K, F77T and M84T mutations, H6-DTKT: somatic variant
141
heavy chain H6 with L72D, S74T, I75K and F77T mutations, H6-DTKT-DNTY: heavy chain
142
H6-DTKT with N28D, D31N, S52T and H98Y mutations, H8-DYKT: somatic variant heavy
143
chain H8 with L72D, S74Y, I75K and F77T mutations, L3-ASPAKQ: somatic variant light chain
144
L3 with S1A, Y2S, T8P, G9A, G16K and R17Q mutations. 10E8v1: heavy chain HC6-S74Y-
145
DKTT + light chain L3, 10E8v4: heavy chain H6-DTKT-DNTY + light chain L3-ASPAKQ,
146
10E8v5: heavy chain HC6-S74Y-DKTT + light chain L3-ASPAKQ.
147 148
Assessment of Antibody-Mediated Neutralization of HIV-1. Neutralization was measured
149
using single-round-of-infection HIV-1 Env-pseudoviruses and TZM-bl target cells, as described
150
previously (33). Neutralization curves were fit by nonlinear regression using a 5-parameter hill
151
slope equation. The 50% and 80% inhibitory concentrations (IC50 and IC80) were reported as the
152
antibody concentrations required to inhibit infection by 50% and 80%, respectively.
153 154
Assessments of Solubility. To measure the turbidity of 10E8 variants at PBS, each variant in
155
IgG elution buffer (pH 2.8, Thermo Scientific, Rockford, IL) was subjected to buffer exchange
156
by either direct dilution with PBS or dialysis in PBS. For direct dilution method, we concentrated
157
antibodies in IgG elution buffer to 10 optical densities (OD) at 280 nm, diluted 20-fold with
7
Downloaded from http://jvi.asm.org/ on January 6, 2019 by guest
135
PBS, incubated overnight at room temperature, loaded 90 μl the diluent onto a 96-well
159
microplate (Corning, NY), and measured the absorbance at 350 nm using SPECTRA max PLUS
160
384 (Molecular Devices, Sunnyvale, CA). For dialysis method, 1 ml of 10E8 variants (1 OD at
161
280 nm, in elution buffer, pH 2.8, Thermo Scientific) was dialyzed against 1X PBS overnight at
162
20°C using a Slide-A-Lyzer dialysis cassette (10,000 MWCO, Thermo Scientific). The content
163
of dialysis bag was resuspended well by pipetting up and down, loaded 90 ul of the content onto
164
a 96-well microplate, and measured the absorbance at 350 nm. As a control, the elution buffer
165
was 20-fold diluted with PBS or dialyzed in PBS.
166 167
Kinetic Concentration. 3 ml of 10E8 variant (0.35 OD at 280 nm) in PBS were centrifuged at
168
4000g for 20 minutes using Amicon Ultra-4 Centrifugal Filter Units (30,000 NMWL, EMD
169
Millipore). Prior to concentration, the 10E8 variants were first passed through a 0.22 μm filter to
170
remove aggregates. The concentrated volume of each variant after centrifugation was measured
171
by weighing and its concentration was measured at 280 nm using NanoDrop (Thermo Scientific,
172
Rockford, IL).
173 174
Dynamic Light Scattering. Dynamic light scattering (DLS) measurements were performed at
175
25°C using DynaPro Plate Reader II (Wyatt Technology, Santa Barbara, CA). The samples were
176
dialyzed at 1X PBS, adjusted to 0.5 mg/ml, and filtered with 0.22 µm filters prior to analysis.
177
The data were analyzed using DYNAMICS version 7.1.7 software (Wyatt technology).
178 179
Assessment of Antibody Polyreactivity. Antibodies were assessed for autoreactivity on two
180
platforms: anti-nuclear antibodies by staining on HEp2 cells (ZEUS Scientific Cat. No: FA2400,
8
Downloaded from http://jvi.asm.org/ on January 6, 2019 by guest
158
ANA HEp2 Test System) and anti-cardiolipin ELISA (Inova Diagnostics Cat. No: 708625,
182
QUANTA LITE ACA IgG III) as per the manufacturer’s instructions. On HEp2 cells, antibodies
183
were tested at 50 and 25 µg/ml. Control antibodies VRC01-LS, 4E10 and VRC07-G54W were
184
included in each slide and assigned a score between 0 and 3+. Test antibodies that scored greater
185
than 1+ at 25 µg/ml were considered autoreactive. In the cardiolipin binding assay, mAbs that
186
scored greater than three times background at 33 µg/ml were considered autoreactive.
187 188
Assessment of Bioavailability in Mice. Balb/c mice were divided into three groups containing
189
three mice per group, and mice in each group were administered intraperitoneally with 100 µg of
190
10E8, 10E8v4 or 10E8v5 antibody. Blood was drawn from all animals at Days 1, 2, 4, 7 and 10
191
post antibody administration and serum was isolated and analyzed for levels of antibody in
192
individual mice. CoStar 96-Well EIA/RIA plates (Corning, NY) were coated with 100 ng per
193
well of goat anti-human IgG Fc-γ fragment (Jackson ImmunoResearch) overnight at 4°C. Plates
194
were washed three times with PBS + Tween and blocked with PBS containing 5% milk and
195
0.5% BSA for 2 hours at room temperature. Mouse serum from the treated animals, and purified
196
10E8, 10E8v4 or 10E8v5 mAb in PBS for the standard curves, were added to the wells in 1:2
197
serial dilutions in PBS containing 2% milk and 0.2% BSA and incubated for 2 hours. After
198
washing, peroxidase-conjugated goat anti-human IgG (Jackson ImmunoResearch) was incubated
199
for 1 hour at room temperature. Samples were detected by TMB Liquid Substrate System
200
(Sigma) and spectrophotometric readings were performed at 450 nm. All animals were bred and
201
maintained at the Comparative Bioscience Center of The Rockefeller University in accordance
202
with the regulations of its Institutional Animal Committee Care and Use Committee (IACUC).
203
All animal studies were conducted under protocols approved by this committee.
9
Downloaded from http://jvi.asm.org/ on January 6, 2019 by guest
181
204
Assessment of Bioavailability in Rhesus Macaques. Indian-origin Rhesus macaques were
206
administered low-endotoxin antibody preparations (