JVI Accepted Manuscript Posted Online 11 February 2015 J. Virol. doi:10.1128/JVI.00078-15 Copyright © 2015, American Society for Microbiology. All Rights Reserved.
1
An anti-H5N1 influenza virus FcDART antibody is an highly efficacious therapeutic and
2
prophylactic against H5N1 influenza virus infection.
3 4
Mark Zanin,a Zhen-Yong Keck,b G. Jonah Rainey,c+ Chia-Ying Kao Lam,c Adrianus C.
5
M. Boon,a* Adam Rubrum,a Daniel Darnell,a Sook-San Wong,a Yolanda Griffin,a Jinming
6
Xia,b Robert G. Webster,a Richard Webby,a# Syd Johnson,c Steven Foungb
7 8
Department of Infectious Diseases, St. Jude Children's Research Hospital, Memphis,
9
Tennessee, USAa; Department of Pathology, Stanford University School of Medicine,
10
Stanford, California, USAb; MacroGenics Inc., Rockville, Maryland, USAc.
11
12
Running head: FcDART is effective against H5N1 influenza viruses
13
#Address correspondence to Richard Webby, [email protected]
14
15
*Present address: Division of Biology and Biomedical Sciences, Washington University,
16
St Louis, Missouri, USA.
17
+
Present address: MedImmune, LLC. Gaithersburg, MD, USA
18
Abstract word count: 237
19
Importance word count: 128
20
Text word count: 4844
1
21
Abstract:
22
Highly pathogenic H5N1 avian influenza viruses are associated with severe disease in
23
humans and continue to be a pandemic threat. Whilst vaccines are available, other
24
approaches are required for patients that typically respond poorly to vaccination, such
25
as the elderly and the immunocompromised. To produce a therapeutic that is highly
26
efficacious at low doses and is broadly specific against antigenically-drifted H5N1
27
influenza viruses, we developed two neutralizing monoclonal antibodies and combined
28
them into a single bispecific-Fc fusion protein (FcDART). In mice, a single 2.5mg/kg
29
therapeutic or prophylactic dose of either monoclonal antibody provided 100%
30
protection against challenge with A/Vietnam/1203/04 (H5N1) or the antigenically-drifted
31
strain A/Whooper swan/Mongolia/244/05 (H5N1). In ferrets, a single 1mg/kg
32
prophylactic dose provided 100% protection against A/Vietnam/1203/04 challenge.
33
FcDART was also effective, as a single 2.5mg/kg therapeutic or prophylactic dose in
34
mice provided 100% protection against A/Vietnam/1203/04 challenge. Antibodies bound
35
to conformational epitopes in antigenic sites on the globular head of the hemagglutinin
36
protein, based on antibody escape mutations. Whilst it was possible to generate escape
37
mutants in vitro, they were neutralized in vivo by the antibodies, as mice infected with
38
escape mutants were 100% protected after only a single therapeutic dose of the
39
antibody used to generate the escape mutant in vitro. In summary, we have combined
40
the antigen specificities of two highly efficacious anti-H5N1 influenza virus antibodies
41
into a bispecific FcDART, which represents a strategy to produce broadly neutralizing
42
antibodies that are effective against antigenically diverse influenza viruses.
43
2
44
Importance:
45
Highly pathogenic H5N1 avian influenza viruses are associated with severe disease in
46
humans and are a pandemic threat. A vaccine is available but other approaches are
47
required for patients that typically respond poorly to vaccination, such as the elderly and
48
the immunocompromised. The variability of the virus means that such an approach
49
must be broad spectrum. To achieve this, we developed two antibodies that neutralize
50
H5N1 influenza viruses. These antibodies provided complete protection in mice against
51
a spectrum of H5N1 influenza viruses at a single low dose. We then combined the two
52
antibodies into a single molecule, FcDART, which combined the broad spectrum activity
53
and protective efficacy of both antibodies. This treatment strategy represents a novel
54
and effective therapeutic or prophylactic against highly pathogenic H5N1 avian
55
influenza viruses.
56 57
Introduction:
58
Since 2003, a series of events has occurred in Asia that many believe to be the
59
harbinger of an incipient influenza pandemic. The virus responsible for these events, an
60
H5N1 influenza virus, has infected and killed millions of chickens and ducks and
61
infected in excess of 650 people (www.who.int; reviewed in (1)). Fortunately this virus
62
has yet to gain the ability to transmit from human to human, however, mammalian
63
transmissibility can be conferred experimentally with relatively few mutations (2, 3). The
64
virulence of this virus (4, 5), its wide geographic distribution (6) and potential for human
65
aerosol transmission (2, 3) urge the development of targeted therapeutic and
66
prophylactic measures. Currently two H5N1 vaccines have been licensed, one of which
67
is available worldwide, and these offer the best means of mass protection (reviewed in 3
68
(7)). However, it is likely the first line of defense against an emerging pandemic will
69
require additional options, including antivirals. Currently, the two options for treating
70
influenza are the M2 ion channel blockers (e.g., amantadine) and the neuraminidase
71
inhibitors (e.g., oseltamivir). Unfortunately, many of the current H5N1 viruses are
72
resistant to amantadine (6) and a higher dose and longer treatment schedule of
73
oseltamivir is needed for protection in murine models than used previously (8), proof
74
that new therapeutic options are very much needed.
75 76
Monoclonal antibodies against the influenza virus hemagglutinin (HA) protein have
77
shown efficacy as therapeutics and prophylactics in murine challenge models against
78
H5N1 influenza virus infection (9-13) and other antibodies have also shown efficacy in
79
mice against other strains of influenza viruses, including H1N1, H2N2 and H3N2 (14-
80
19). Dosages at which these antibodies provided 100% protection against challenge in
81
mice ranged from 5 to 50mg/kg when delivered therapeutically and 1 to 30mg/kg when
82
delivered prophylactically. In ferrets, an anti-H5 antibody provided 100% protection
83
against challenge with an H5N1 influenza virus at 10mg/kg delivered prophylactically
84
and at 30mg/kg delivered therapeutically (20).
85 86
The variability of the HA protein, which contains the main antigenic sites of the virus,
87
necessitates that any broadly utilized antibody must be effective against antigenically
88
diverse influenza viruses. The most conserved regions of HA are the stalk and the
89
receptor-binding pocket, and antibodies directed at these sites are efficacious against
90
antigenically diverse strains of influenza virus (14, 21, 22). However, the most exposed
91
and antibody-accessible epitopes on HA are located on the globular head, although 4
92
these regions are highly variable. Due to the variability of HA, it is likely that more than
93
one neutralizing antibody will be required to provide broad protection against different
94
lineages of H5N1 and to protect against antibody escape mutants. However, this
95
presents certain problems for clinical testing and product development as cocktails of
96
antibodies are difficult and expensive to produce and test. Therefore, in this study we
97
have developed a bi-specific Fc fusion protein, Fc Dual-Affinity Re-Targeting (FcDART),
98
that combines the neutralizing capacity and specificities of two different neutralizing
99
monoclonal antibodies from human and murine sources that target the globular head of
100
HA. Individually, the two monoclonal antibodies provided 100% protection against
101
A/Vietnam/1203/04 (H5N1) (VN1203) challenge in mice and ferrets as therapeutics and
102
prophylactics at low doses (2.5mg/kg in mice and 1mg/kg in ferrets). They also provided
103
100% protection in mice against challenge with A/Whooper swan/Mongolia/244/05
104
(H5N1) (Mon244), which is an antigenically drifted strain. Further, antibody escape
105
mutants generated in vitro were not lethal in mice treated with a single therapeutic dose
106
of antibody. The FcDART molecule that combined the antigen specificities of these two
107
antibodies also provided 100% protection against challenge in mice as a therapeutic or
108
prophylactic and represents a strategy to produce antibody-based therapeutics that are
109
effective against antigenically diverse influenza viruses.
110 111
Materials and Methods:
112
Cells Lines and Culture Conditions. 293T, Chinese Hamster Ovary (CHO) and Madin
113
Darby canine kidney (MDCK) cell lines were obtained from ATCC. 293T cells were
114
grown in Opti-MEM (Invitrogen, CA) supplemented with 5% fetal calf serum (FCS,
115
Gemini Bioproducts Inc, CA), CHO cells were grown in F12K medium (Invitrogen, CA) 5
116
supplemented with 10% FCS and MDCK cells were grown in Dulbecco’s modified
117
minimal essential medium (DMEM, Invitrogen, CA) supplemented with 10% FCS and 2
118
mM glutamine. For MDCK cell infections, viruses were diluted in infection medium
119
(MEM supplemented with 5% (v/v) bovine serum albumin (BSA) and 2 mM glutamine
120
(Sigma, MO)).
121 122
Immunofluorescence Assay (IFA). MDCK cells were infected with VN1203 or Mon244
123
overnight. The cells were harvested by trypsinization and resuspended in PBS with 2%
124
FCS. Aliquots containing 3x104 cells were spotted onto HTC Super Cured 24-spot
125
slides (Erie Scientific Company, NH), dried, and fixed with 100% acetone for 10 min at
126
room temperature. Fixed cells were incubated with hybridoma supernatants for 30 min
127
at 37C and washed for 5 min with PBS. Slides were then incubated for 30 min at 37C
128
with 50 ng/mL propidium iodide and fluorescein isothiocyanate (FITC)-conjugated goat
129
anti-human IgG (Jackson ImmunoResearch Laboratories, West Grove, PA). Bound
130
antibody was revealed by fluorescence microscopy.
131 132
Hemagglutination inhibition (HI) assay. HI assays were conducted using standard
133
methodologies. In brief, 25 l of diluted antigen at four agglutination doses in PBS was
134
added to wells of 96 well plates containing a two-fold dilution series of the test antibody.
135
After 30 minutes incubation at room temperature, 50l of 0.5% (v/v) chicken or horse
136
red blood cells was added to each well and incubated at room temperature for another
137
30 min. Titers were recorded as the lowest dilution of antibody able to inhibit
138
hemagglutination.
139
6
140
Microneutralization (MN) assay. MN assays were conducted using MDCK cells
141
according to standard methodologies. In brief, a two-fold dilution series of each antibody
142
was incubated with virus at 100TCID50/50µL for one hour at 37oC. The antibody/virus
143
solutions were then added to MDCK cells for an additional hour at 37oC and were then
144
washed off and the cells incubated at 37oC for 72hrs with 200µL infection medium
145
containing 1µg/mL TPCK trypsin. Neutralizing titers were read by incubating 50µL of cell
146
culture medium with 0.5% (v/v) chicken or horse red blood cells followed by incubation
147
at room temperature for another 30 min and were expressed as the reciprocal of the
148
serum dilution that inhibited 50% of viral growth of 100TCID50 of virus.
149 150
Human Monoclonal Antibody Generation. Plasma samples were obtained from
151
individuals immunized with a recombinant, baculovirus-expressed HA protein from the
152
A/Hong Kong/156/97 (H5N1) virus (23) followed by the inactivated, subvirion vaccine
153
against VN1203 (24). Samples were screened by HI or by IFA using H5N1 (VN1203)
154
HA protein or H5N1-infected MDCK cells, respectively. In this study, a donor with serum
155
HI titer of 320 was used as the source of peripheral B-cells for human hybridoma
156
production as previously described (25). Three hybridomas, designated as BF1-1, BF1-
157
10 and BF1-19, were identified by HI assay secreting neutralizing H5N1 human
158
monoclonal antibodies (HMAbs). Against a panel of Southeast Asian H5N1 isolates,
159
these three donors showed reactivity to the antigenically distinct H5N1 virus HK156 but
160
not to the clade 2.3.4 virus A/Common magpie/Hong Kong/645/06. BF1-32 was
161
identified by IFA. Monoclonality was confirmed by sequencing IgG genes isolated from
162
10 individual cell clones derived from each hybridoma cell line. Cloning and analyzing
163
the VL and VH domains of these clones were performed as previously described (26). 7
164
HMAb production and purification were performed as described (25) and biotinylation of
165
the antibodies was carried out according to manufacturer’s instructions (Pierce
166
Biotechnology, Inc, Rockford, IL). The cell lines produced IgG1 antibody with λ light
167
chain for BF1-1 and κ light chains for BF1-10, BF1-19 and BF1-32 and secreted 10-40
168
µg human IgG per mL in spent cultured supernatant.
169 170
Murine Monoclonal Antibody Generation. Generation of murine monoclonal antibodies
171
(mMAbs) was achieved via a vaccination regimen consisting of an initial dose of 0.5
172
mouse lethal dose 50% (MLD50) of a reverse-genetics virus bearing the HA and
173
neuraminidase genes of VN1203 and the six internal genes of A/Puerto Rico/8/1934
174
(H1N1) (rgH5N1) delivered intranasally (IN) followed by two booster doses delivered
175
intraperitoneally (IP)
176
hemagglutination units (HAU) of Mon244 delivered IP followed by 1000 egg infectious
177
dose 50% (EID50) of rgH5N1 delivered intravenously (IV). Post vaccination, hybridomas
178
were produced by Rockland Immunochemicals and were screened by HI and IFA as
179
described above.
four
weeks apart. Booster
doses consisted
of
10,000
180 181
Humanization of Murine Monoclonal Antibodies. mMAbs were humanized by first
182
obtaining cDNA of the heavy (VH) and light (VL) variable segments from total RNA via
183
3’ RACE followed by cloning into the pCR2.1-topo vector (Life Technologies), which
184
was then transformed into E.coli, recovered and sequenced. The VH and VL segments
185
were also combined with human Cg1 and K constant region cDNA segments in the
186
mammalian expression vectors pEE13.4 (LC) and pEE6.4 (HC) (Lonza), then combined
187
into a single expression vector and the resulting chimeric HC and LC were then 8
188
transfected into CHO-S cells (Life Technologies). Conditioned media from transfected
189
cells was assayed by ELISA for human IgG production and anti-H5N1 activity. Human
190
germLine genes with the highest homology to the VH and VL segments were identified
191
using Ig BLAST and humanized VH and VL segments were then designed using CDR
192
grafting, VH1-18/VH1-69 was used for VH frameworks, and VK-O12 used for VL for
193
humanization of 10C3 and humanized VH and VL segments were then synthesized by
194
gene synthesis (GenScript). The resulting fragments were then cloned into the
195
mammalian expression vector pCIneo (Promega) with the appropriate C region and
196
clones were sequenced.
197 198
Generation of the FcDART. The antigen specificities of BF1-19 and h10C3 were
199
combined into a bispecific Fc fusion protein, termed FcDART. This was achieved by
200
joining the two antibody chains such that each chain of the FcDART comprised of the
201
VL from one antibody and the VH of the second antibody and vice-versa, with a short
202
linker (GGGSGGG) between the VL and VH. One of the two chains was then fused to a
203
human IgG1 hinge-Fc region (figure 1A). The two chains were transfected into CHO
204
cells and a stable pool of producing cells was selected. The material was purified over a
205
protein A column followed by size exclusion chromatography (figure 1B).
206 207
Affinity Measurement. The binding affinity of each monoclonal antibody (MAb) to H5N1
208
HA was characterized by surface plasmon resonance (SPR) analysis. The experiments
209
described below were carried out using the BIAcore 3000 system (Biacore). Affinity
210
purified anti-human IgG, Fcγ (Jackson Immuno Research) was amine coupled (10,000
211
RU) to all four flow cells of a CM5 sensor chip (Biacore). MAb samples at 0.5-1 µg/mL in 9
212
running buffer were injected into flow cells 2-4 while keeping flow cell 1 as a reference,
213
which led to roughly 150 RU of MAb being captured in each flow cell. Association
214
kinetics were observed by injecting purified H5N1 HA (Protein Sciences) in running
215
buffer through all four flow cells at a flow rate of 30 µl/min for a duration of 90 seconds.
216
Disassociation kinetics were then measured with running buffer flowing at a rate of 30
217
µl/min for a duration of 10 minutes. Flow cells were regenerated using a 15 µl injection
218
of 100mM H3PO4. Kinetics were tested at HA concentrations of 800, 400, 200, 100, 50,
219
25, 12.5, 6.25, 3.125, 1.56 nM. After each binding cycle, the same protocol was run with
220
running buffer substituted for the analyte to serve as a double reference. Data were
221
analyzed using BIAevaluation software version 4.1 (Biacore). The data were processed
222
by subtracting the double reference and trimming out MAb loading and H3PO4
223
regeneration information. Kinetic values were obtained by applying the 1:1 (Langmuir)
224
binding, simultaneous ka/kd, fit method.
225 226
Antibody Binding Profiles. Direct (antigen down) or capture (antibody down) ELISA
227
formats were used to measure antibody binding to HA from different clades of H5N1
228
influenza viruses. In antigen down format, HA protein was coated on an ELISA plate at
229
a concentration of 2µg/mL. HA protein was obtained from eEnzyme and Protein
230
Sciences and was produced in 293 cells or an insect cell line, respectively. HA was
231
uncleaved or cleaved at the furin cleavage site (furin). The indicated concentration of
232
antibody was then added to the plate and detection was performed using an
233
horseradish
234
chemiluminescent detection (SuperSignal, Pierce). For the MAb-Down assays, an
235
antibody specific for CH1 (4A11, Abcam) was used to coat an ELISA plate at 2 µg/mL.
peroxidase
(HRP)-conjugated
anti-human
antibody,
followed
by
10
236
The indicated concentration of antibody was then captured. His-tagged antigen
237
(100ng/mL) was then applied to the plate, followed by an HRP-conjugated anti-His
238
antibody, followed by chemiluminescent detection (SuperSignal, Pierce). To detect
239
binding of h10C3, this assay was performed in an antibody down format to maximize
240
sensitivity using antibody captured with polyclonal anti-human IgG and antigen binding
241
was detected with streptavidin-HRP and developed colorimetrically with 3,3’,5,5’-
242
Tetramethylbenzidine (TMB, Sigma).
243 244
Prevention and Therapeutic Treatment in animals. All animal studies were conducted
245
under applicable laws and guidelines of and after approval from the St. Jude Children’s
246
Research Hospital Animal Care and Use Committee. Female 6-8-week old BALB/c mice
247
were housed in cages with 5 mice per cage. Mice received the indicated amount of
248
antibody in a total volume of 300µL sterile PBS by intraperitoneal (IP) injection. At
249
various times pre and post antibody administration mice were lightly anesthetized with
250
isoflourane and inoculated intranasally with 100 MLD50 of VN1203 (1000 EID50) or 100
251
MLD50 (5x105 EID50) of Mon244 in 30µL of PBS.
252 253
Male 4 to 5-month-old ferrets were housed in cages with three animals per cage and
254
were administered 1mg/kg of indicated antibodies in 1mL PBS intramuscularly and on
255
the next day ferrets were lightly anaesthetized with isoflurane and administered 3×104
256
EID50 of VN1203 in 1mL of PBS IN. For all animals, weight change and survival were
257
monitored for 14 days following virus challenge.
258
11
259
Isolation and characterization of antibody escape mutants. Antibody escape mutants
260
were selected by inoculating 10-day-old specific pathogen free (SPF) embryonated
261
chicken eggs with rgH5N1 incubated with neutralizing antibody essentially as described
262
previously (27, 28). Eggs were screened for the presence of virus by hemagglutination
263
assay on allantoic fluid, which was harvested from infected eggs and viral RNA was
264
extracted and used to generate cDNA by RT-PCR. The HA gene was then amplified
265
and sequenced.
266 267
Data analysis and statistics. Data collected were inputted and graphed using Graphpad
268
Prism version 5.03. Statistical analysis was performed using the log-rank (Mantel-Cox)
269
test.
270 271
Results:
272
Characterization of H5N1 HMAbs. To determine the genetic diversity of the antibodies
273
identified, we sequenced the CDR 1, 2 and 3 regions of the antibody heavy chains.
274
Sequence analysis of the 11 HI-positive hybridomas showed that they clustered into two
275
different groups. Two HMAbs clustered into group one while the others clustered into
276
the second group. One antibody representing group one (BF1-10), and two representing
277
group two (BF1-1 and BF1-19) were chosen for further analysis. HI assays were
278
performed with spent supernatants from human hybridomas to assess the cross
279
reactivity of the generated antibodies against a panel of H5N1 viruses isolated in Asia.
280
All antibodies were reactive against VN1203, showing titers of 640 to 2560 (table 1).
281
BF1-19 also showed reactivity against the antigenically drifted strain Mon244 (HI=40),
282
whilst BF1-1 and BF1-10 did not (table 1). No antibodies showed reactivity against 12
283
A/Japanese white eye/Hong Kong/1038/06 (HK1038), which is antigenically drifted from
284
VN1203 to a greater degree than Mon244 (table 1). BF1-19 showed strong reactivity
285
against A/Hong Kong/156/97 (HK156) (HI=1280) whilst BF1-1 and BF1-10 showed only
286
weak reactivity (HI=10 and 20, respectively) (table 1). All antibodies also showed strong
287
neutralization activity in microneutralization assay against VN1203, with titers of 640 to
288
10240, however no neutralization activity was observed to Mon244 or HK1038 (table 2).
289
BF1-19 showed strong microneutralization activity against HK156 (titer>20,480), whilst
290
BF1-1 and BF1-10 showed only weak activity (titer=10 and 40, respectively) (table 2).
291 292
To determine the nature of the epitopes recognized by each HI-positive HMAb, Western
293
Blot analysis was performed using rgH5N1-infected MDCK cell lysate. All neutralizing
294
HMAbs showed no binding to denatured H5N1 antigens, suggesting that they recognize
295
conformational epitopes. The one HI-negative HMAb that was included in this study,
296
BF1-32, was Western blot-positive and was identified by IFA with VN1203-infected
297
MDCK cells. This antibody appears to target HA1, as the most prominent bands
298
corresponded to HA0 (82 Kda) and HA1 (58 Kda), with a less intense band at 78 Kda
299
(figure 2). To assess the affinities of each of the HMAbs for the H5 HA protein, surface
300
plasmon resonance studies were conducted using a BIAcore 3000 instrument. The KD
301
values for the binding of HMAbs BF1-1, BF1-10, and BF1-19 to the VN1203 HA were
302
20.3nM, 10.8nM, and 48.2nM, respectively, showing that each had similar affinities for
303
their antigens. Following these studies, we selected BF1-19 for subsequent
304
experiments.
305
13
306
We next assessed the breadth of reactivity of BF1-19 to a panel of HAs from diverse
307
H5N1 viruses via ELISA. We used antigens obtained from two commercial sources,
308
eEnzyme and Protein Sciences, to investigate whether the antigen production system
309
had an effect on the ability of the antibodies to bind, as the eEnzyme antigens were
310
produced in 293T cells whilst the Protein Sciences antigens were produced in insect
311
cells. Further, we used both wild type (wt) HA antigens and antigens where the furin
312
cleavage site were deleted (furin), to determine if cleavage had an impact on the
313
exposure of the epitope recognized by these antibodies. All eEnzyme antigens with the
314
exception of A/Indonesia/5/2005 (H5N1) were engineered to remove the furin cleavage
315
site, resulting in a more homogeneous product (figure 3). The Protein Sciences antigens
316
and the eEnzyme A/Indonesia/5/2005 (H5N1) antigen retained the furin cleavage site,
317
resulting in oligomers that consisted of both cleaved and uncleaved antigens. Antigens
318
obtained from Protein Sciences were tested in ‘antibody down’ format, where the
319
antibody was bound to the solid phase via a capture antibody and the antigen was the
320
soluble binding partner (figure 3). eEnzyme antigens contained a C-terminal HIS tag,
321
which allowed them to be used as a soluble binding partner or to be immobilized on the
322
solid-phase using a HIS binding partner, termed ‘antigen down’ (figure 3).
323 324
BF1-19 bound to Clade 1 antigens from both suppliers in an antigen down format (figure
325
3A) and also bound to Clades 2.2.1 and 2.3.4 antigens in antibody down format (figure
326
3B). Some binding to Clade 2.1.3.2 antigen was also detected at high concentrations
327
(figure 3B). Given the oligomeric structure of the antigens, the ability to detect binding in
328
antibody down format may be the result of more avid binding compared to the antigen
329
down format. BF1-19 also bound to Clade 1 antigens with or without the furin cleavage 14
330
site (figure 3A). Whilst these data are not able to definitively demonstrate a quantitative
331
difference between cleaved and uncleaved HA, they do clearly demonstrate that the
332
recognized epitope is available on oligomers derived from both wt and furin constructs.
333 334
Generation of mMAbs. We used a three dose vaccination regimen in mice, consisting of
335
VN1203 delivered IN then IP followed by Mon244 delivered IV, to generate antibodies
336
broadly specific to different clades of H5 influenza viruses. One of these antibodies,
337
10C3, bound to Clades 1 and 2.2.1 antigens from both suppliers in antigen down format
338
and, like BF1-19, bound to both wt and furin constructs (figure 3C). Similar results
339
were also observed in antibody down format (figure 3D).
340 341
10C3 was then humanized to reduce potential immunogenicity by grafting the
342
complementarity determining regions of the antibody onto the framework regions from
343
closely related human VH and VL gene sequences. Interestingly, there were indications
344
of h10C3 binding to Clade 2.1.3.2 and 2.3.4 antigens, which was not observed in 10C3
345
(figure 3E). Whilst this data is not conclusive, it is indicative that the humanization
346
process broadened the coverage of this antibody (figure 3D and E). By HI, h10C3 was
347
strongly reactive to VN1203 (HI=2560) and showed relatively weak reactivity against
348
Mon244 (HI=40) and no reactivity against HK1038. Unlike BF1-19, h10C3 did not show
349
reactivity against HK156 (table 1). In microneutralization assays, h10C3 only showed
350
neutralization activity against VN1203 (titer=2560), no neutralization activity was
351
observed against HK156, Mon244 or HK1038 (table 2).
352
15
353
Design and Development of the FcDART. The FcDART molecule consisted of two
354
polypeptide chains that heterodimerized to form a bi-specific molecule. Each chain
355
contained the VL of one specificity and the VH of the other, such that heterodimerization
356
of the two chains results in assembly of matching VL/VH domains. In order to favor
357
heterodimeric assembly, a strong heterodimerization domain has been appended to the
358
C-terminus of each chain. These domains consisted of coiled coil-forming sequences
359
where the hydrophobic core of the coiled coil interface was flanked by opposite charges
360
(E- on one coil and K+ on the other). These charges are believed to favor
361
heterodimerization by repelling homotypic association and preventing the initiation of
362
“zippering” of the hydrophobic core (figure 1A). These DART molecules were modified
363
further by appending an Fc domain to the C-terminus of the E-coil, causing two DART
364
units to dimerize and form a tetravalent bispecific structure (figure 1A). The addition of
365
the Fc confers long serum half-life on the molecules and restores the avidity associated
366
with the bivalent parental mAbs. Further, the presence of the Fc confers protein A
367
binding and makes these molecules amenable to platform purification processes
368
routinely used for mAbs in large scale cGMP production.
369 370
Following expression in CHO cells and purification, analytical size exclusion
371
chromatography showed that purified material consisted of predominantly one species
372
that eluted at a volume consistent with the predicted molecular mass of FcDART. Only a
373
small amount of higher order aggregate was present, as indicated by the slight left
374
shoulder on the peak. Increasing the concentration from 2 to 24.5mg/mL did not cause
375
a significant increase in the amount of aggregated material (figure 1B). These
376
characteristics indicate that the molecule was properly assembled, homogeneously 16
377
purified and not prone to aggregation. We then tested the FcDART against a panel of
378
antigens in antibody down format and found that, with the exception of Clade 2.1.3.2,
379
the FcDART was able to bind to the antigens in the panel and was generally consistent
380
with the binding profile of BF1-19 and h10C3, indicating that FcDART indeed combined
381
the binding profile of BF1-19 and h10C3 (figure 3F). By HI, FcDART showed reactivity
382
against VN1203 and HK156 (HI=640), relatively weak reactivity against Mon244 (HI=40)
383
and no reactivity against HK1038, which is again consistent with a binding profile that is
384
a combination of BF1-19 and h10C3 (table 1). The neutralization activity of FcDART
385
was also consistent with an antibody with the combined specificities of h10C3 and BF1-
386
19, as strong neutralization activity was observed against both VN1203 (titer=1280) and
387
HK156 (titer=10240) (table 2). Interestingly, FcDART also showed neutralization activity
388
against Mon244 (titer=80), which was not observed with any other antibody tested,
389
suggesting an increased breadth of specificity (table 2).
390 391
Escape mutant characterization. To further characterize the epitopes recognized by
392
these antibodies, we inoculated embryonated chicken eggs with rgH5N1 and antibody
393
and sequenced viruses present after incubation for 48hrs. Based on the mutations in
394
these viruses, MAbs primarily bound to antigenic sites on the globular head of HA (29).
395
L128S in antigenic site C and M265I in antigenic site A were present in escape mutants
396
to h10C3, K139E in antigenic site A was in escape mutants to 10C3, S140Y/F in
397
antigenic site A were in escape mutants to BF1-19, K188N/E in antigenic site B were in
398
escape mutants to BF1-19 and FcDART, T213A in antigenic site D was in escape
399
mutants to FcDART (figure 4). With the exception of h10C3, where L128S and M265I
400
were present in one mutant, each escape mutant did not harbor more that one of the 17
401
mutations indicated in figure 4. Further, escape mutants harboring the same mutation
402
were seen in all eggs (n=3) at the dilution of antibody that was permissive to the
403
generation of escape mutants.
404 405
We next performed HI assays to determine if binding of the antibodies to these escape
406
mutants was diminished or abrogated. BF1-19 showed little activity against the BF1-19
407
escape mutants K188N and S140F (HI titers
1
An anti-H5N1 influenza virus FcDART antibody is an highly efficacious therapeutic and
2
prophylactic against H5N1 influenza virus infection.
3 4
Mark Zanin,a Zhen-Yong Keck,b G. Jonah Rainey,c+ Chia-Ying Kao Lam,c Adrianus C.
5
M. Boon,a* Adam Rubrum,a Daniel Darnell,a Sook-San Wong,a Yolanda Griffin,a Jinming
6
Xia,b Robert G. Webster,a Richard Webby,a# Syd Johnson,c Steven Foungb
7 8
Department of Infectious Diseases, St. Jude Children's Research Hospital, Memphis,
9
Tennessee, USAa; Department of Pathology, Stanford University School of Medicine,
10
Stanford, California, USAb; MacroGenics Inc., Rockville, Maryland, USAc.
11
12
Running head: FcDART is effective against H5N1 influenza viruses
13
#Address correspondence to Richard Webby, [email protected]
14
15
*Present address: Division of Biology and Biomedical Sciences, Washington University,
16
St Louis, Missouri, USA.
17
+
Present address: MedImmune, LLC. Gaithersburg, MD, USA
18
Abstract word count: 237
19
Importance word count: 128
20
Text word count: 4844
1
21
Abstract:
22
Highly pathogenic H5N1 avian influenza viruses are associated with severe disease in
23
humans and continue to be a pandemic threat. Whilst vaccines are available, other
24
approaches are required for patients that typically respond poorly to vaccination, such
25
as the elderly and the immunocompromised. To produce a therapeutic that is highly
26
efficacious at low doses and is broadly specific against antigenically-drifted H5N1
27
influenza viruses, we developed two neutralizing monoclonal antibodies and combined
28
them into a single bispecific-Fc fusion protein (FcDART). In mice, a single 2.5mg/kg
29
therapeutic or prophylactic dose of either monoclonal antibody provided 100%
30
protection against challenge with A/Vietnam/1203/04 (H5N1) or the antigenically-drifted
31
strain A/Whooper swan/Mongolia/244/05 (H5N1). In ferrets, a single 1mg/kg
32
prophylactic dose provided 100% protection against A/Vietnam/1203/04 challenge.
33
FcDART was also effective, as a single 2.5mg/kg therapeutic or prophylactic dose in
34
mice provided 100% protection against A/Vietnam/1203/04 challenge. Antibodies bound
35
to conformational epitopes in antigenic sites on the globular head of the hemagglutinin
36
protein, based on antibody escape mutations. Whilst it was possible to generate escape
37
mutants in vitro, they were neutralized in vivo by the antibodies, as mice infected with
38
escape mutants were 100% protected after only a single therapeutic dose of the
39
antibody used to generate the escape mutant in vitro. In summary, we have combined
40
the antigen specificities of two highly efficacious anti-H5N1 influenza virus antibodies
41
into a bispecific FcDART, which represents a strategy to produce broadly neutralizing
42
antibodies that are effective against antigenically diverse influenza viruses.
43
2
44
Importance:
45
Highly pathogenic H5N1 avian influenza viruses are associated with severe disease in
46
humans and are a pandemic threat. A vaccine is available but other approaches are
47
required for patients that typically respond poorly to vaccination, such as the elderly and
48
the immunocompromised. The variability of the virus means that such an approach
49
must be broad spectrum. To achieve this, we developed two antibodies that neutralize
50
H5N1 influenza viruses. These antibodies provided complete protection in mice against
51
a spectrum of H5N1 influenza viruses at a single low dose. We then combined the two
52
antibodies into a single molecule, FcDART, which combined the broad spectrum activity
53
and protective efficacy of both antibodies. This treatment strategy represents a novel
54
and effective therapeutic or prophylactic against highly pathogenic H5N1 avian
55
influenza viruses.
56 57
Introduction:
58
Since 2003, a series of events has occurred in Asia that many believe to be the
59
harbinger of an incipient influenza pandemic. The virus responsible for these events, an
60
H5N1 influenza virus, has infected and killed millions of chickens and ducks and
61
infected in excess of 650 people (www.who.int; reviewed in (1)). Fortunately this virus
62
has yet to gain the ability to transmit from human to human, however, mammalian
63
transmissibility can be conferred experimentally with relatively few mutations (2, 3). The
64
virulence of this virus (4, 5), its wide geographic distribution (6) and potential for human
65
aerosol transmission (2, 3) urge the development of targeted therapeutic and
66
prophylactic measures. Currently two H5N1 vaccines have been licensed, one of which
67
is available worldwide, and these offer the best means of mass protection (reviewed in 3
68
(7)). However, it is likely the first line of defense against an emerging pandemic will
69
require additional options, including antivirals. Currently, the two options for treating
70
influenza are the M2 ion channel blockers (e.g., amantadine) and the neuraminidase
71
inhibitors (e.g., oseltamivir). Unfortunately, many of the current H5N1 viruses are
72
resistant to amantadine (6) and a higher dose and longer treatment schedule of
73
oseltamivir is needed for protection in murine models than used previously (8), proof
74
that new therapeutic options are very much needed.
75 76
Monoclonal antibodies against the influenza virus hemagglutinin (HA) protein have
77
shown efficacy as therapeutics and prophylactics in murine challenge models against
78
H5N1 influenza virus infection (9-13) and other antibodies have also shown efficacy in
79
mice against other strains of influenza viruses, including H1N1, H2N2 and H3N2 (14-
80
19). Dosages at which these antibodies provided 100% protection against challenge in
81
mice ranged from 5 to 50mg/kg when delivered therapeutically and 1 to 30mg/kg when
82
delivered prophylactically. In ferrets, an anti-H5 antibody provided 100% protection
83
against challenge with an H5N1 influenza virus at 10mg/kg delivered prophylactically
84
and at 30mg/kg delivered therapeutically (20).
85 86
The variability of the HA protein, which contains the main antigenic sites of the virus,
87
necessitates that any broadly utilized antibody must be effective against antigenically
88
diverse influenza viruses. The most conserved regions of HA are the stalk and the
89
receptor-binding pocket, and antibodies directed at these sites are efficacious against
90
antigenically diverse strains of influenza virus (14, 21, 22). However, the most exposed
91
and antibody-accessible epitopes on HA are located on the globular head, although 4
92
these regions are highly variable. Due to the variability of HA, it is likely that more than
93
one neutralizing antibody will be required to provide broad protection against different
94
lineages of H5N1 and to protect against antibody escape mutants. However, this
95
presents certain problems for clinical testing and product development as cocktails of
96
antibodies are difficult and expensive to produce and test. Therefore, in this study we
97
have developed a bi-specific Fc fusion protein, Fc Dual-Affinity Re-Targeting (FcDART),
98
that combines the neutralizing capacity and specificities of two different neutralizing
99
monoclonal antibodies from human and murine sources that target the globular head of
100
HA. Individually, the two monoclonal antibodies provided 100% protection against
101
A/Vietnam/1203/04 (H5N1) (VN1203) challenge in mice and ferrets as therapeutics and
102
prophylactics at low doses (2.5mg/kg in mice and 1mg/kg in ferrets). They also provided
103
100% protection in mice against challenge with A/Whooper swan/Mongolia/244/05
104
(H5N1) (Mon244), which is an antigenically drifted strain. Further, antibody escape
105
mutants generated in vitro were not lethal in mice treated with a single therapeutic dose
106
of antibody. The FcDART molecule that combined the antigen specificities of these two
107
antibodies also provided 100% protection against challenge in mice as a therapeutic or
108
prophylactic and represents a strategy to produce antibody-based therapeutics that are
109
effective against antigenically diverse influenza viruses.
110 111
Materials and Methods:
112
Cells Lines and Culture Conditions. 293T, Chinese Hamster Ovary (CHO) and Madin
113
Darby canine kidney (MDCK) cell lines were obtained from ATCC. 293T cells were
114
grown in Opti-MEM (Invitrogen, CA) supplemented with 5% fetal calf serum (FCS,
115
Gemini Bioproducts Inc, CA), CHO cells were grown in F12K medium (Invitrogen, CA) 5
116
supplemented with 10% FCS and MDCK cells were grown in Dulbecco’s modified
117
minimal essential medium (DMEM, Invitrogen, CA) supplemented with 10% FCS and 2
118
mM glutamine. For MDCK cell infections, viruses were diluted in infection medium
119
(MEM supplemented with 5% (v/v) bovine serum albumin (BSA) and 2 mM glutamine
120
(Sigma, MO)).
121 122
Immunofluorescence Assay (IFA). MDCK cells were infected with VN1203 or Mon244
123
overnight. The cells were harvested by trypsinization and resuspended in PBS with 2%
124
FCS. Aliquots containing 3x104 cells were spotted onto HTC Super Cured 24-spot
125
slides (Erie Scientific Company, NH), dried, and fixed with 100% acetone for 10 min at
126
room temperature. Fixed cells were incubated with hybridoma supernatants for 30 min
127
at 37C and washed for 5 min with PBS. Slides were then incubated for 30 min at 37C
128
with 50 ng/mL propidium iodide and fluorescein isothiocyanate (FITC)-conjugated goat
129
anti-human IgG (Jackson ImmunoResearch Laboratories, West Grove, PA). Bound
130
antibody was revealed by fluorescence microscopy.
131 132
Hemagglutination inhibition (HI) assay. HI assays were conducted using standard
133
methodologies. In brief, 25 l of diluted antigen at four agglutination doses in PBS was
134
added to wells of 96 well plates containing a two-fold dilution series of the test antibody.
135
After 30 minutes incubation at room temperature, 50l of 0.5% (v/v) chicken or horse
136
red blood cells was added to each well and incubated at room temperature for another
137
30 min. Titers were recorded as the lowest dilution of antibody able to inhibit
138
hemagglutination.
139
6
140
Microneutralization (MN) assay. MN assays were conducted using MDCK cells
141
according to standard methodologies. In brief, a two-fold dilution series of each antibody
142
was incubated with virus at 100TCID50/50µL for one hour at 37oC. The antibody/virus
143
solutions were then added to MDCK cells for an additional hour at 37oC and were then
144
washed off and the cells incubated at 37oC for 72hrs with 200µL infection medium
145
containing 1µg/mL TPCK trypsin. Neutralizing titers were read by incubating 50µL of cell
146
culture medium with 0.5% (v/v) chicken or horse red blood cells followed by incubation
147
at room temperature for another 30 min and were expressed as the reciprocal of the
148
serum dilution that inhibited 50% of viral growth of 100TCID50 of virus.
149 150
Human Monoclonal Antibody Generation. Plasma samples were obtained from
151
individuals immunized with a recombinant, baculovirus-expressed HA protein from the
152
A/Hong Kong/156/97 (H5N1) virus (23) followed by the inactivated, subvirion vaccine
153
against VN1203 (24). Samples were screened by HI or by IFA using H5N1 (VN1203)
154
HA protein or H5N1-infected MDCK cells, respectively. In this study, a donor with serum
155
HI titer of 320 was used as the source of peripheral B-cells for human hybridoma
156
production as previously described (25). Three hybridomas, designated as BF1-1, BF1-
157
10 and BF1-19, were identified by HI assay secreting neutralizing H5N1 human
158
monoclonal antibodies (HMAbs). Against a panel of Southeast Asian H5N1 isolates,
159
these three donors showed reactivity to the antigenically distinct H5N1 virus HK156 but
160
not to the clade 2.3.4 virus A/Common magpie/Hong Kong/645/06. BF1-32 was
161
identified by IFA. Monoclonality was confirmed by sequencing IgG genes isolated from
162
10 individual cell clones derived from each hybridoma cell line. Cloning and analyzing
163
the VL and VH domains of these clones were performed as previously described (26). 7
164
HMAb production and purification were performed as described (25) and biotinylation of
165
the antibodies was carried out according to manufacturer’s instructions (Pierce
166
Biotechnology, Inc, Rockford, IL). The cell lines produced IgG1 antibody with λ light
167
chain for BF1-1 and κ light chains for BF1-10, BF1-19 and BF1-32 and secreted 10-40
168
µg human IgG per mL in spent cultured supernatant.
169 170
Murine Monoclonal Antibody Generation. Generation of murine monoclonal antibodies
171
(mMAbs) was achieved via a vaccination regimen consisting of an initial dose of 0.5
172
mouse lethal dose 50% (MLD50) of a reverse-genetics virus bearing the HA and
173
neuraminidase genes of VN1203 and the six internal genes of A/Puerto Rico/8/1934
174
(H1N1) (rgH5N1) delivered intranasally (IN) followed by two booster doses delivered
175
intraperitoneally (IP)
176
hemagglutination units (HAU) of Mon244 delivered IP followed by 1000 egg infectious
177
dose 50% (EID50) of rgH5N1 delivered intravenously (IV). Post vaccination, hybridomas
178
were produced by Rockland Immunochemicals and were screened by HI and IFA as
179
described above.
four
weeks apart. Booster
doses consisted
of
10,000
180 181
Humanization of Murine Monoclonal Antibodies. mMAbs were humanized by first
182
obtaining cDNA of the heavy (VH) and light (VL) variable segments from total RNA via
183
3’ RACE followed by cloning into the pCR2.1-topo vector (Life Technologies), which
184
was then transformed into E.coli, recovered and sequenced. The VH and VL segments
185
were also combined with human Cg1 and K constant region cDNA segments in the
186
mammalian expression vectors pEE13.4 (LC) and pEE6.4 (HC) (Lonza), then combined
187
into a single expression vector and the resulting chimeric HC and LC were then 8
188
transfected into CHO-S cells (Life Technologies). Conditioned media from transfected
189
cells was assayed by ELISA for human IgG production and anti-H5N1 activity. Human
190
germLine genes with the highest homology to the VH and VL segments were identified
191
using Ig BLAST and humanized VH and VL segments were then designed using CDR
192
grafting, VH1-18/VH1-69 was used for VH frameworks, and VK-O12 used for VL for
193
humanization of 10C3 and humanized VH and VL segments were then synthesized by
194
gene synthesis (GenScript). The resulting fragments were then cloned into the
195
mammalian expression vector pCIneo (Promega) with the appropriate C region and
196
clones were sequenced.
197 198
Generation of the FcDART. The antigen specificities of BF1-19 and h10C3 were
199
combined into a bispecific Fc fusion protein, termed FcDART. This was achieved by
200
joining the two antibody chains such that each chain of the FcDART comprised of the
201
VL from one antibody and the VH of the second antibody and vice-versa, with a short
202
linker (GGGSGGG) between the VL and VH. One of the two chains was then fused to a
203
human IgG1 hinge-Fc region (figure 1A). The two chains were transfected into CHO
204
cells and a stable pool of producing cells was selected. The material was purified over a
205
protein A column followed by size exclusion chromatography (figure 1B).
206 207
Affinity Measurement. The binding affinity of each monoclonal antibody (MAb) to H5N1
208
HA was characterized by surface plasmon resonance (SPR) analysis. The experiments
209
described below were carried out using the BIAcore 3000 system (Biacore). Affinity
210
purified anti-human IgG, Fcγ (Jackson Immuno Research) was amine coupled (10,000
211
RU) to all four flow cells of a CM5 sensor chip (Biacore). MAb samples at 0.5-1 µg/mL in 9
212
running buffer were injected into flow cells 2-4 while keeping flow cell 1 as a reference,
213
which led to roughly 150 RU of MAb being captured in each flow cell. Association
214
kinetics were observed by injecting purified H5N1 HA (Protein Sciences) in running
215
buffer through all four flow cells at a flow rate of 30 µl/min for a duration of 90 seconds.
216
Disassociation kinetics were then measured with running buffer flowing at a rate of 30
217
µl/min for a duration of 10 minutes. Flow cells were regenerated using a 15 µl injection
218
of 100mM H3PO4. Kinetics were tested at HA concentrations of 800, 400, 200, 100, 50,
219
25, 12.5, 6.25, 3.125, 1.56 nM. After each binding cycle, the same protocol was run with
220
running buffer substituted for the analyte to serve as a double reference. Data were
221
analyzed using BIAevaluation software version 4.1 (Biacore). The data were processed
222
by subtracting the double reference and trimming out MAb loading and H3PO4
223
regeneration information. Kinetic values were obtained by applying the 1:1 (Langmuir)
224
binding, simultaneous ka/kd, fit method.
225 226
Antibody Binding Profiles. Direct (antigen down) or capture (antibody down) ELISA
227
formats were used to measure antibody binding to HA from different clades of H5N1
228
influenza viruses. In antigen down format, HA protein was coated on an ELISA plate at
229
a concentration of 2µg/mL. HA protein was obtained from eEnzyme and Protein
230
Sciences and was produced in 293 cells or an insect cell line, respectively. HA was
231
uncleaved or cleaved at the furin cleavage site (furin). The indicated concentration of
232
antibody was then added to the plate and detection was performed using an
233
horseradish
234
chemiluminescent detection (SuperSignal, Pierce). For the MAb-Down assays, an
235
antibody specific for CH1 (4A11, Abcam) was used to coat an ELISA plate at 2 µg/mL.
peroxidase
(HRP)-conjugated
anti-human
antibody,
followed
by
10
236
The indicated concentration of antibody was then captured. His-tagged antigen
237
(100ng/mL) was then applied to the plate, followed by an HRP-conjugated anti-His
238
antibody, followed by chemiluminescent detection (SuperSignal, Pierce). To detect
239
binding of h10C3, this assay was performed in an antibody down format to maximize
240
sensitivity using antibody captured with polyclonal anti-human IgG and antigen binding
241
was detected with streptavidin-HRP and developed colorimetrically with 3,3’,5,5’-
242
Tetramethylbenzidine (TMB, Sigma).
243 244
Prevention and Therapeutic Treatment in animals. All animal studies were conducted
245
under applicable laws and guidelines of and after approval from the St. Jude Children’s
246
Research Hospital Animal Care and Use Committee. Female 6-8-week old BALB/c mice
247
were housed in cages with 5 mice per cage. Mice received the indicated amount of
248
antibody in a total volume of 300µL sterile PBS by intraperitoneal (IP) injection. At
249
various times pre and post antibody administration mice were lightly anesthetized with
250
isoflourane and inoculated intranasally with 100 MLD50 of VN1203 (1000 EID50) or 100
251
MLD50 (5x105 EID50) of Mon244 in 30µL of PBS.
252 253
Male 4 to 5-month-old ferrets were housed in cages with three animals per cage and
254
were administered 1mg/kg of indicated antibodies in 1mL PBS intramuscularly and on
255
the next day ferrets were lightly anaesthetized with isoflurane and administered 3×104
256
EID50 of VN1203 in 1mL of PBS IN. For all animals, weight change and survival were
257
monitored for 14 days following virus challenge.
258
11
259
Isolation and characterization of antibody escape mutants. Antibody escape mutants
260
were selected by inoculating 10-day-old specific pathogen free (SPF) embryonated
261
chicken eggs with rgH5N1 incubated with neutralizing antibody essentially as described
262
previously (27, 28). Eggs were screened for the presence of virus by hemagglutination
263
assay on allantoic fluid, which was harvested from infected eggs and viral RNA was
264
extracted and used to generate cDNA by RT-PCR. The HA gene was then amplified
265
and sequenced.
266 267
Data analysis and statistics. Data collected were inputted and graphed using Graphpad
268
Prism version 5.03. Statistical analysis was performed using the log-rank (Mantel-Cox)
269
test.
270 271
Results:
272
Characterization of H5N1 HMAbs. To determine the genetic diversity of the antibodies
273
identified, we sequenced the CDR 1, 2 and 3 regions of the antibody heavy chains.
274
Sequence analysis of the 11 HI-positive hybridomas showed that they clustered into two
275
different groups. Two HMAbs clustered into group one while the others clustered into
276
the second group. One antibody representing group one (BF1-10), and two representing
277
group two (BF1-1 and BF1-19) were chosen for further analysis. HI assays were
278
performed with spent supernatants from human hybridomas to assess the cross
279
reactivity of the generated antibodies against a panel of H5N1 viruses isolated in Asia.
280
All antibodies were reactive against VN1203, showing titers of 640 to 2560 (table 1).
281
BF1-19 also showed reactivity against the antigenically drifted strain Mon244 (HI=40),
282
whilst BF1-1 and BF1-10 did not (table 1). No antibodies showed reactivity against 12
283
A/Japanese white eye/Hong Kong/1038/06 (HK1038), which is antigenically drifted from
284
VN1203 to a greater degree than Mon244 (table 1). BF1-19 showed strong reactivity
285
against A/Hong Kong/156/97 (HK156) (HI=1280) whilst BF1-1 and BF1-10 showed only
286
weak reactivity (HI=10 and 20, respectively) (table 1). All antibodies also showed strong
287
neutralization activity in microneutralization assay against VN1203, with titers of 640 to
288
10240, however no neutralization activity was observed to Mon244 or HK1038 (table 2).
289
BF1-19 showed strong microneutralization activity against HK156 (titer>20,480), whilst
290
BF1-1 and BF1-10 showed only weak activity (titer=10 and 40, respectively) (table 2).
291 292
To determine the nature of the epitopes recognized by each HI-positive HMAb, Western
293
Blot analysis was performed using rgH5N1-infected MDCK cell lysate. All neutralizing
294
HMAbs showed no binding to denatured H5N1 antigens, suggesting that they recognize
295
conformational epitopes. The one HI-negative HMAb that was included in this study,
296
BF1-32, was Western blot-positive and was identified by IFA with VN1203-infected
297
MDCK cells. This antibody appears to target HA1, as the most prominent bands
298
corresponded to HA0 (82 Kda) and HA1 (58 Kda), with a less intense band at 78 Kda
299
(figure 2). To assess the affinities of each of the HMAbs for the H5 HA protein, surface
300
plasmon resonance studies were conducted using a BIAcore 3000 instrument. The KD
301
values for the binding of HMAbs BF1-1, BF1-10, and BF1-19 to the VN1203 HA were
302
20.3nM, 10.8nM, and 48.2nM, respectively, showing that each had similar affinities for
303
their antigens. Following these studies, we selected BF1-19 for subsequent
304
experiments.
305
13
306
We next assessed the breadth of reactivity of BF1-19 to a panel of HAs from diverse
307
H5N1 viruses via ELISA. We used antigens obtained from two commercial sources,
308
eEnzyme and Protein Sciences, to investigate whether the antigen production system
309
had an effect on the ability of the antibodies to bind, as the eEnzyme antigens were
310
produced in 293T cells whilst the Protein Sciences antigens were produced in insect
311
cells. Further, we used both wild type (wt) HA antigens and antigens where the furin
312
cleavage site were deleted (furin), to determine if cleavage had an impact on the
313
exposure of the epitope recognized by these antibodies. All eEnzyme antigens with the
314
exception of A/Indonesia/5/2005 (H5N1) were engineered to remove the furin cleavage
315
site, resulting in a more homogeneous product (figure 3). The Protein Sciences antigens
316
and the eEnzyme A/Indonesia/5/2005 (H5N1) antigen retained the furin cleavage site,
317
resulting in oligomers that consisted of both cleaved and uncleaved antigens. Antigens
318
obtained from Protein Sciences were tested in ‘antibody down’ format, where the
319
antibody was bound to the solid phase via a capture antibody and the antigen was the
320
soluble binding partner (figure 3). eEnzyme antigens contained a C-terminal HIS tag,
321
which allowed them to be used as a soluble binding partner or to be immobilized on the
322
solid-phase using a HIS binding partner, termed ‘antigen down’ (figure 3).
323 324
BF1-19 bound to Clade 1 antigens from both suppliers in an antigen down format (figure
325
3A) and also bound to Clades 2.2.1 and 2.3.4 antigens in antibody down format (figure
326
3B). Some binding to Clade 2.1.3.2 antigen was also detected at high concentrations
327
(figure 3B). Given the oligomeric structure of the antigens, the ability to detect binding in
328
antibody down format may be the result of more avid binding compared to the antigen
329
down format. BF1-19 also bound to Clade 1 antigens with or without the furin cleavage 14
330
site (figure 3A). Whilst these data are not able to definitively demonstrate a quantitative
331
difference between cleaved and uncleaved HA, they do clearly demonstrate that the
332
recognized epitope is available on oligomers derived from both wt and furin constructs.
333 334
Generation of mMAbs. We used a three dose vaccination regimen in mice, consisting of
335
VN1203 delivered IN then IP followed by Mon244 delivered IV, to generate antibodies
336
broadly specific to different clades of H5 influenza viruses. One of these antibodies,
337
10C3, bound to Clades 1 and 2.2.1 antigens from both suppliers in antigen down format
338
and, like BF1-19, bound to both wt and furin constructs (figure 3C). Similar results
339
were also observed in antibody down format (figure 3D).
340 341
10C3 was then humanized to reduce potential immunogenicity by grafting the
342
complementarity determining regions of the antibody onto the framework regions from
343
closely related human VH and VL gene sequences. Interestingly, there were indications
344
of h10C3 binding to Clade 2.1.3.2 and 2.3.4 antigens, which was not observed in 10C3
345
(figure 3E). Whilst this data is not conclusive, it is indicative that the humanization
346
process broadened the coverage of this antibody (figure 3D and E). By HI, h10C3 was
347
strongly reactive to VN1203 (HI=2560) and showed relatively weak reactivity against
348
Mon244 (HI=40) and no reactivity against HK1038. Unlike BF1-19, h10C3 did not show
349
reactivity against HK156 (table 1). In microneutralization assays, h10C3 only showed
350
neutralization activity against VN1203 (titer=2560), no neutralization activity was
351
observed against HK156, Mon244 or HK1038 (table 2).
352
15
353
Design and Development of the FcDART. The FcDART molecule consisted of two
354
polypeptide chains that heterodimerized to form a bi-specific molecule. Each chain
355
contained the VL of one specificity and the VH of the other, such that heterodimerization
356
of the two chains results in assembly of matching VL/VH domains. In order to favor
357
heterodimeric assembly, a strong heterodimerization domain has been appended to the
358
C-terminus of each chain. These domains consisted of coiled coil-forming sequences
359
where the hydrophobic core of the coiled coil interface was flanked by opposite charges
360
(E- on one coil and K+ on the other). These charges are believed to favor
361
heterodimerization by repelling homotypic association and preventing the initiation of
362
“zippering” of the hydrophobic core (figure 1A). These DART molecules were modified
363
further by appending an Fc domain to the C-terminus of the E-coil, causing two DART
364
units to dimerize and form a tetravalent bispecific structure (figure 1A). The addition of
365
the Fc confers long serum half-life on the molecules and restores the avidity associated
366
with the bivalent parental mAbs. Further, the presence of the Fc confers protein A
367
binding and makes these molecules amenable to platform purification processes
368
routinely used for mAbs in large scale cGMP production.
369 370
Following expression in CHO cells and purification, analytical size exclusion
371
chromatography showed that purified material consisted of predominantly one species
372
that eluted at a volume consistent with the predicted molecular mass of FcDART. Only a
373
small amount of higher order aggregate was present, as indicated by the slight left
374
shoulder on the peak. Increasing the concentration from 2 to 24.5mg/mL did not cause
375
a significant increase in the amount of aggregated material (figure 1B). These
376
characteristics indicate that the molecule was properly assembled, homogeneously 16
377
purified and not prone to aggregation. We then tested the FcDART against a panel of
378
antigens in antibody down format and found that, with the exception of Clade 2.1.3.2,
379
the FcDART was able to bind to the antigens in the panel and was generally consistent
380
with the binding profile of BF1-19 and h10C3, indicating that FcDART indeed combined
381
the binding profile of BF1-19 and h10C3 (figure 3F). By HI, FcDART showed reactivity
382
against VN1203 and HK156 (HI=640), relatively weak reactivity against Mon244 (HI=40)
383
and no reactivity against HK1038, which is again consistent with a binding profile that is
384
a combination of BF1-19 and h10C3 (table 1). The neutralization activity of FcDART
385
was also consistent with an antibody with the combined specificities of h10C3 and BF1-
386
19, as strong neutralization activity was observed against both VN1203 (titer=1280) and
387
HK156 (titer=10240) (table 2). Interestingly, FcDART also showed neutralization activity
388
against Mon244 (titer=80), which was not observed with any other antibody tested,
389
suggesting an increased breadth of specificity (table 2).
390 391
Escape mutant characterization. To further characterize the epitopes recognized by
392
these antibodies, we inoculated embryonated chicken eggs with rgH5N1 and antibody
393
and sequenced viruses present after incubation for 48hrs. Based on the mutations in
394
these viruses, MAbs primarily bound to antigenic sites on the globular head of HA (29).
395
L128S in antigenic site C and M265I in antigenic site A were present in escape mutants
396
to h10C3, K139E in antigenic site A was in escape mutants to 10C3, S140Y/F in
397
antigenic site A were in escape mutants to BF1-19, K188N/E in antigenic site B were in
398
escape mutants to BF1-19 and FcDART, T213A in antigenic site D was in escape
399
mutants to FcDART (figure 4). With the exception of h10C3, where L128S and M265I
400
were present in one mutant, each escape mutant did not harbor more that one of the 17
401
mutations indicated in figure 4. Further, escape mutants harboring the same mutation
402
were seen in all eggs (n=3) at the dilution of antibody that was permissive to the
403
generation of escape mutants.
404 405
We next performed HI assays to determine if binding of the antibodies to these escape
406
mutants was diminished or abrogated. BF1-19 showed little activity against the BF1-19
407
escape mutants K188N and S140F (HI titers
