CVI Accepted Manuscript Posted Online 13 April 2016 Clin. Vaccine Immunol. doi:10.1128/CVI.00085-16 Copyright © 2016 McCraw et al. This is an open-access article distributed under the terms of the Creative Commons Attribution 4.0 International license.
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21
Characterization of influenza vaccine hemagglutinin complexes by cryo-electron microscopy and image analyses reveals structural polymorphisms Dustin M. McCraw1, John R. Gallagher1 and Audray K. Harris 1# 1Laboratory
#
of Infectious Diseases, National Institute of Allergy and Infectious Diseases, National Institutes of Health, Bethesda, MD 20892
To whom correspondence should be addressed: E-mail: A.K.H. (
[email protected])
22 23 24 25 26 27 28 29 30 31 32 33 34 35
1
36
ABSTRACT
38
threat of animal viruses crossing the species barrier to cause epidemics and pandemics resulting in
37
Influenza virus afflicts millions of people worldwide on an annual basis. There is the ever-present
39
great morbidity and mortality. Zoonosis outbreaks such as H7N9 underscore the need to better
40 41 42 43
44
understand the molecular organization of viral immunogens such as recombinant influenza
hemagglutinin (HA) proteins used in influenza subunit vaccines in order to optimize vaccine efficacy. Here we show using cryo-electron microscopy and image analysis that vaccine
recombinant H7 HA form macromolecular complexes consisting of a variable number of HA
subunits (6 to 8). In addition, HA-complexes were distributed across at least four distinct structural
45
classes (polymorphisms). 3D reconstruction and molecular modeling indicated that HA was in the
47
polymorphisms were due to hydrophobic interactions involving the transmembrane regions. These
49
used in subunit vaccines will lead to further understanding of the differences in vaccine efficacy and
46
48
50 51
52
53
pre-fusion state and suggested that the mechanism for oligomerization and observed structural
experiments suggest that characterization of the molecular structures of influenza HA complexes future optimization of subunit vaccines to prevent influenza infection. INTRODUCTION
The influenza virus is a negative-sense enveloped virus that infects millions of people
globally on an annual basis. Annual vaccination is required due to the constant change in the
54
antigenic composition of hemagglutinin (HA), which is the major viral antigenic constituent on the
56
Influenza viruses with subtypes H1, H2, and H3 have caused global epidemics and pandemics
55 57 58 59
60
viral membrane (1, 2). HAs are classified into eighteen antigenic subtypes (H1-H18) (3, 4).
within the human population (5). However, other subtypes can infect humans. Influenza has a
zoonotic nature, which allows animals viruses to infect humans to which the human population has no preexisting immunity. An example of this type of zoonosis was the influenza H7N9 virus that
infected humans in 2013 to which seasonal vaccines at that time could not provide protection (6-9).
2
61
However, later studies indicated that H7 subtype vaccines were efficacious against the H7N9
62
virus (10-14). These vaccines consisted of chemically inactivated and non-inactivated formulations
64
inactivated viruses (13) or purified influenza HA from recombinant protein expression systems (11,
63 65
66
of viruses and subunit vaccines. Subunit vaccines contain either viral glycoproteins isolated from
14). HA proteins isolated from influenza virus can form complexes that are referred to as rosettes, protein micelles, or glycoprotein subunits.(15, 16). Here we refer to recombinant HA proteins as
67
hemagglutinin complexes (14, 17, 18).
69
proteins. Three copies of the protomer form the trimeric HA molecule (19, 20). HA1 forms an apical
68
The protomer of trimeric HA is HA0 that is cleaved into disulfide linked HA1 and HA2
70
globular region while extended regions of HA1 and HA2 form the stem region. Also, HA2 contains a
72
HA mediates viral entry via a low-pH mediated conformational change that leads to membrane
71 73
74
transmembrane region located near the C-terminus which embeds HA in the viral membrane (21).
fusion via a conversion from a pre-fusion to a post-fusion state (20, 22-24).
Antibodies against hemagglutinin (HA) efficiently neutralize influenza virus. Thus, HA is the
75
principle antigen in influenza vaccines. Recombinant HA vaccines are advantageous because they
77
antigens (25) and recombinant proteins do not require the time needed to produce influenza
76 78 79 80 81
82
83 84
85
are not derived from chemical inactivation processes that can chemically modify viral surface
viruses used in the virus-based influenza vaccine formulations (26, 27). While the 3D molecular
organization and orientation of HA molecules on the surface of viruses have been studied (21, 28), the molecular structure and organization of recombinant HA complexes used within influenza vaccines has not been studied in great detail. Also, the conformation and mechanism of
oligomerization of HA molecules within recombinant HA complexes has not been addressed. In this study, we used cryo-electron microscopy and image analysis to identify and
characterize the polymorphisms of H7 HA complexes. HA was not cleaved into HA1 and HA2 but remained biochemically as HA0. HA-complexes distributed into four major classes with variable
3
86
numbers of HA molecules among the classes with a range of six to eight HA molecules per complex.
88
used to derive orientation information of constituent HA molecules within the complexes and to
87
89
90
The molecules did not present any rigid symmetry but starfish-like motifs. Averaging methods were obtain a 3D density map at ~25 Å resolution. Using the reconstructed density map and molecular
modeling with coordinates, we determined that HA was in a pre-fusion conformation. Models of
91
molecular orientations placed the transmembrane regions emanating from the center of the
93
polymorphisms and oligomerization of HA complexes. The ability to characterize the molecular
95
could be extended to other HA subtypes, and could be used to investigate the observed varying
92 94 96 97
98 99
complexes. This suggested that non-specific hydrophobic interactions were the basis for both
organizations of H7 HA molecules within recombinant HA complexes used in vaccine formulations stabilities and efficacies of subunit vaccines of differing compositions. Furthermore, structural
studies of subunit vaccines may describe correlates to immunogenicity for influenza HA and other viral glycoproteins used in subunit vaccine formulations. MATERIALS AND METHODS
100
SDS-PAGE and immunoblotting. Hemagglutinin proteins were from Protein Sciences Corporation
102
(A/Netherlands/219/2003 (H7N7) and (A/Anhui/1/1013) (H7N9). Proteins were in PBS and all
104
manufacture antigens for commercial influenza vaccines. SDS-PAGE analysis was carried out to
106
contained the reducing agent dithiothretol (DTT) at final concentration of 100 mM. Samples were
101 103 105 107
108
109 110
(Meriden, CT) and were recombinant full-length H7 proteins of influenza viruses:
produced from a proprietary baculovirus expression system (Protein Sciences Corporation) used to
assess the purity and HA0 cleavage states of hemagglutinin. For reducing conditions samples
heated at 95 degrees Celsius for 10 minutes for heating conditions and left at room temperature for
non-heating conditions of SDS-PAGE. Over-night coomassie blue staining was used for protein band
visualization of SDS-PAGE gels. For non-reducing and non-heated immunoblotting of antigens,
during SDS-PAGE no DTT or heating were used before transfer to nitrocellulose membranes. For
4
111
immunobloting, the primary antibody was anti-H7 mouse monoclonal antibody InA414 (anti H7N7)
113
alkaline-phosphatase conjugated secondary antibodies, and NBT/BCIP chromogenic substrate
112
from Novus Biologicals (Littleton, CO). Immunodetection and development was carried out using
114
(ThermoFisher Scientific, Waltham, MA).
116
applied to glow-discarded continuous carbon grids, and subsequently stained with 4% uranyl
115
117
118
Cryo-electron microscopy. For negative-staining screening of samples (4µl at 100µg/ml) were acetate before air-drying at room temperature and then imaged under cryo-conditions like vitrified
samples (see below). For cryo-electron microscopy samples (2µl at 500µg/ml in PBS) without
119
heavy stain added were applied to holey carbon films (Quantfoil, Großlöbichau, Germany) and
121
microscope (FEI Company, Hillsboro, OR) operated at liquid nitrogen temperatures and at 300 kV
120 122
plunge-frozen using a Vitrobot Mark IV (FEI Company, Hillsboro, OR). A Titan Krios electron
was used to digitally collect images via EPU on a 4,096 × 4,096 charge-coupled-device (CCD) camera
123
(Gatan Inc., Warrendale, PA) at a pixel size of 1.15 Å. The electron at doses ranged from ~10 to 20
125
astigmatism were excluded from further analysis. For further analyses, cryo-images (5299) were
124
126
127
128
129
130
131
e−/Å2 with defocus values ranging from ~2.5 to ~5.5 µm. Cryo-images indicating drift and visually screened for images that had the presence of particles.
Particle boxing and 2D-classification. Because automated methods of particle boxing (picking)
failed, as judged by high false positives, manual boxing was used to obtain well-separated particles
(i.e. non overlapping) by screening the cryo-images (5299) and manually boxing with a box size of
600 angstroms. Particles (N=3877) were centered aligned, and averaged. The 2D circular average
was then calculated. To measure diameter from the 1-D profile, the relative intensity along the x-
132
axis was plotted against the distance radius from the center via the Bsoft software package(29). 1D
134
manner. For 2D-classification complexes were classified as implemented in Relion (30).
133 135
profiles for the 2D class averages and the 3D rotational average profiles were calculated in a like
Convergence was considered reached with the assignment of 3875 of 3877 particle images (99.9%)
5
136
to ten 2D classes. The top four most populous 2D classes were subjected to 1D profile analysis to
137
estimate diameters (Bsoft).
139
protrusions (i.e. HA molecules) in the most populous 2D class averages, image libraries were
138
140 141 142 143 144 145 146 147
148
Computational estimation of HA numbers in complexes. In order to estimate the number of created of rotated images and then were correlated (Bsoft). The number of units per 360 degrees were varied to approximate 2, 3, 4, 5, 6, 7, 8, and 9 asymmetric units per 360 degrees resulting in rotated images by corresponding degree angles (0, 40, 45, 51.42, 60, 72, 80, 90, 102.84, 120, 135,
144, 154.6, 160, 180, 200, 205.68, 216, 225, 240, 250, 257.1, 270, 280, 288, 300,3 08.52, 315, 320,
360). Correlation maximums from plots were taken as the number of units (molecules) within the particular 2D class average of the complexes.
3D reconstruction and molecular modeling. Because automated methods of particle boxing (picking) failed, as judged by high false positives, to pick out the individual protrusions (HA
molecules) from the complexes due to overlap, manual boxing was used to obtain well-separated
149
HA protrusions to determine the 3D structural state of HA in the complexes. For screening
151
complexes. A total of 1327 individual protrusions (HA molecules) were manually selected and
150 152 153 154 155 156
complexes, a box size of 185.12 angstroms was used to select individual HA protrusions from the subjected to further analysis. Particles were first subjected to global shape analysis via calculation of a 3D average and subsequent profile analysis, and then 3D reconstruction with the molecular
docking of coordinates into the 3D map. 3D structural analyses were carried with Bsoft, EMAN2, and Relion and molecular docking with the Chimera software packages(29-32).
157
RESULTS
159
site of the hemagglutinin (HA) proteins, we used SDS-PAGE analysis. This allowed for a method to
158
Hemagglutinins as uncleaved HA0. To determine the cleavage status of the HA1-HA2 cleavage
160
define the relative purity and proportions of HA0, HA1, and HA2 proteins. For both HA proteins, H7
6
161
Anhui and H7 Netherlands, protein bands were observed at apparent molecular weights of about
163
would be expected to appear at about 50kDa and 20kDa, respectively. These bands were not
162 164 165 166 167
168
169
170
171
172
173
174
175
70kDa, which is the approximate molecular weight for HA0 (Figure 1A). Bands for HA1 and HA2
observed. Proteins were judged to be greater than 90% and 95% pure by densitometry. Assessment of protein purity and homogeneity was important for further analysis of HA by electron microscopy and by immunoblotting because the purity leveled ruled out both large amounts of contaminating proteins and heterogeneously cleaved HA proteins that would hamper further structural analysis and interpretation.
HA0 protomer detection and HA complexes. Hemagglutinins (HA) have conserved cysteine
residues (Figure S1) and their trimeric structures indicate inter and intra-disulfides within the HA1
and HA2 regions (33, 34). Thus, in order to address the affect of disulfide reduction on HA
conformation and oligomerization via immunoblotting, we screened a panel of H7 antibodies under
various reducing and non-reducing conditions and identified a monoclonal antibody (In414) with a
conformationally dependent epitope that required non-heating and non-reducing conditions
(Figures S2, S3). HA0 protomers were detected via immunoblotting using mAb414 (Figure 1B). Six
176
bands were visually apparent in the immunoblot (Figure 1B). Further analysis of protein banding
178
respectively (Figure S4). Based on molecular weight standards ranging from 20kDa to 220kDa the
180
respectively (Figure 1B). The top three bands of the six-band H7 Netherlands ladder were larger
177 179
patterns by densitometry (Figure S4) indicated six and four peaks for H7 Netherlands and H7 Anhui,
lower three bands of the HA had approximate molecular weights of 70kDa, 120kDa, and 220 kDa,
181
than the 220kDa marker and were judged to be higher molecular weight oligomers of HA0. (Figure
183
into constituent molecules during gel analysis to form a ladder (Figure 1B), electron microscopy
182 184
185
1B). Because six HA0 molecules (six protein bands) could represent two HA timers decomposed
was used to observe HA0 complexes. H7 Netherlands contained isolated molecular complexes as
observed by negative-staining electron microscopy (Figure 1C). The complexes were not compact
7
186
or spherical structures but appeared as starfish-like structures with protruding appendages from a
187
central point. Several complexes are denoted within circles (Figure 1C). Similar HA complexes were
189
analyses of the HA complexes.
188 190 191 192
193
seen for H7 Anhui by cryo-electron microscopy which were subjected to further 2D and 3D image Molecular Organization of HA-complexes by cryo-electron microscopy. While negative-
staining techniques in electron microscopy produce high contract images by way of heavy metal
counterstaining, the protein sample is subject to dehydration, and deposition of negative stain can introduce irregularities in individual particles that antagonize further image analysis. Thus, to
194
understand the molecular organization of HA-complexes we analyzed Anhui H7 HA-complexes by
196
structures with protruding appendages with starfish-like appearances as observed by cryo-electron
195
cryo-electron microscopy and image analysis. HA-complexes embedded in vitreous ice appeared as
197
microscopy (Figure 2A).
199
Observing the distribution of dot-shaped (HA head domain) and bar-shaped (HA stem domain)
198
200
Constituent appendages appeared as dark dots and bar-shaped densities (Figure 2A-2D).
densities among the HA complexes suggested the possibility of variable orientations of constituent
201
HA molecules. (Figure 2A-2D). In cryo-electron microscopy, sample is applied to a porous carbon
203
carbon film (Figure S5A), but some protein complexes adhere directly to the carbon film, where
202 204
205
film. Imaging conditions are best where protein is suspended in vitreous ice within holes of the
image contrast is reduced (Figure S5B). HA-complexes did distribute into vitreous ice, but had a
tendency to bind to the carbon film (Figure S5). For image analysis, we excluded complexes adhered
206
to the carbon film, and complexes that overlapped in the 2D projected image. Thus, over five
208
thousand individual manually boxed images of HA-complexes sufficient for analysis of size and
207 209 210
thousand 2D cryo-images had to be collected and manually screened in order to obtain about four subsequent molecular organization of HA complexes (methods). The average size of the HA-
complex was estimated by a circular average and derived 1D profile curve of the complexes (Figure
8
211
2E, 2F). The 1D profile curve had a smooth fall-off to a radius of about 180 angstroms (Figure 2F).
213
angstroms (36 nm) (Figure 2F).
212 214 215 216 217
218
219
Thus, the average population diameter (d = 2r) of the HA-complexes was estimated to be about 360 HA-complex classification and analysis. Molecular complexes often have rigid symmetry that can be observed by 2D image classification of macromolecules. To answer the questions of whether
there is rigid symmetry and discernable different subsets of structures we used computational 2D image classification and analyzed the subsequent classes for sizes and the number of constituent
HA molecules. HA-complexes (N=3875) distributed into ten 2D-classes (#1 thru #10) (Figure 3A). Each class had a unique appearance. Six of the classes (1,4,5,6,8,10) were poorly converged, and
220
contained less than 100 constituent particles (Figure 3B). The most populated classes were class
222
class averages with starfish-like structures with appendages protruding from the particles center,
221
#7 (n=1838), class #2 (n=1339), class #3 (n=522), and class # 9 (n=114). These classes displayed
223
and no apparent symmetry and indicated structural polymorphism.
225
calculated and used to produce 1D profile curves to measure their radii. 1D profiles curves
224
To assess class diameter, 2D circular averages of the four most populous classes were
226
appeared distinct in shapes. However, two general curve types of the 2D density profile curves
228
observed density peaks out from zero at approximately 55 and 35 angstroms, respectively, (Figure
230
(Figure 3D, 3E). Radii of the classes varied, and were estimated as ~ 41.1 nm, 38.6 nm, 39.7 nm and
232
antibody reactivity, and size, the constituent protrusions/appendages within the complexes were
227 229 231 233
234 235
calculated from circular averages were observed (Figure 3C, 3D, 3E, 3F). Classes #2 and #9 had
3D, 3F) while classes #3 and # 7 had density peak maximums near the center set to zero distance
35.1 nm, for classes 2, 3,7, and 9, respectively (Figure 3C-3F). Based upon purity of the sample, assigned to HA molecules.
Estimation of the number of constituent HA molecules per class. Classes displayed differences
in the number of HA molecules emanating from the complex (Figure 3A). Because 2D image
9
236
classification produced classes with discernable HA density protrusions even in the absence of rigid
238
densities within these classes (Figure 3A). We developed a strategy by using image rotation and
237 239
symmetry, further computational analyzes were done to estimate the number of protruding HA correlation to estimate the number of HA protrusions (assumed as slightly shifted asymmetric
240
units) per 360 degrees in 2D classes 2, 3, 7 and 9. For each class an image library was created by
242
Image rotation resulted in protrusion rotation about the center. As expected 0 and 360 degree
241 243
244
rotating each image at angular increments within the rotation range of 0 to 360 degrees (methods). images were identical as indicated with some examples of rotated images for class # 2 (Figure 4A). Classes had different estimated numbers of HA protrusions based on the highest correlation peak
245
from the correlations plots (Figure 4B, 4C, 4D, 4E). The numbers of unit protrusions ranged from 6
247
#7 has 7 units. Class #3 had the highest estimation with 8 units (Figure 4B-4E).
246 248
to 8 units per 360-degree rotations. Classes #2 and #9 were estimated to have 6 units, while class Molecular conformation of HA molecules within HA complexes. In order to address the
249
questions of HA orientation and 3D molecular conformation of the HA molecules within HA-
251
refined 3D reconstruction map. Computational filtering of images to improve visual contrast of HA
250 252 253 254
255
256
257
258 259
260
complexes, we used 3D reconstruction techniques to calculate a circular 3D average and then a
complexes was used to help observe the relative orientations of HA molecules within the complexes (Figure 5A-5D). Complexes had constituent HA molecules that appeared to be in different apical, lateral, and tilted arrangements with respect to the image plane. In some instances lateral (side-
views) of the HA molecules within the complexes resolved to be peanut-shaped densities (Figure
5A-5C, black arrows). Apical (top-views) of the HA molecules within the HA complexes appeared as dark dots or triangular-like views (Figure 5A-5C, white arrows).
In cryo-electron microscopy samples are quickly frozen and are suspended in a 3D layer of
vitreous ice (Figure 2A). Cryo-EM images are 2D projection images resulting from the electron
beam transmitting through the 3D structure of the sample in a frozen hydrated-state (Figure S6).
10
261
Schematics of apical and lateral projection images from HA molecules are illustrated in
263
the impression of planar HA complexes, notable details indicated that they were 3-dimensional in
262 264
265
supplemental Figure S6. Although the raw images (Figure 2A) and 2D classes (Figure 3A) might give nature. Views of the HA molecules in apparent apical (Figure 5, white arrows) and lateral (black
arrows) orientations were observed, which was consistent with a non-planar configuration of the
266
HA-complexes (Figure 5A-5D). The radial density observed in the 2D class averages also indicated a
268
center of the complex, consistent with a projection image of a globular structure (Figure 3C-F). Thus
267
269 270 271
272
3-dimensional nature to the complexes. The highest density in 2D class averages was at or near the the HA-complexes are likely to be irregularly distributed about the complex center in all 3 dimensions.
Individual HA appendages from the HA complexes were boxed and a 3D circular average was
calculated. The 3D average had a bi-lobed molecular shape in the body of the structure that was on
273
top of a base (Figure 5E). The 3D average consisted of three molecular regions as detected by the
275
average had a similar pattern as the profile curve derived from a viral HA map that had been solved
274 276 277
278 279
280
281 282
283
284 285
curve shape of its density profile (Figure 5F, blue curve). The derived profile curve from the 3D
previously by cryo-electron microscopy (28)(Figure 5F, grey curve). Peaks 1, 2, and 3 of the HA
complex data were approximately registered with peaks for HA1, HA2, and transmembrane regions of viral HA, respectively (Figure 5F). Ectodomain and transmembrane lengths of the HA molecular average were measured from the width of peaks 1, 2, and 3. The ectodomain region assigned to
peaks 1 and 2 spanned about 158 angstroms (15.8 nm) and the transmembrane region assigned to peak 3 had a width of about 40 angstroms (4 nm) (Figure 5F, blue curve).
To understand the molecular disposition of HA1 and HA2 regions in HA complexes, a 3D
reconstruction was determined from computationally boxed appendages of the HA complexes. The
refined 3D reconstruction was bilobed in shape (Figure 6A). The lobe at the top was larger than the
lower lobe, which gave a peanut-shape appearance to the 3D map (Figure 6A). Docking of HA H7
11
286
coordinates into the 3D map matched the top larger lobe to the HA1 head (red) region and the
288
the HA2 C-terminus of the ectodomain at the bottom of the 3D map (Figure 6B, 6C). This HA
290
molecules with proximal transmembrane regions centrally located within the HA complexes
292
DISCUSSION
287 289 291
293
294
lower lobe to the HA2 stem (blue) region of the coordinates (Figure 6B). The docking positioned
orientation was integrated into schematic models of HA-complexes invoking bi-lobed shaped
(Figure 7).
In order to answer questions about the molecular organization and conformation of
hemagglutinin (HA)-complexes used in influenza recombinant subunit vaccines, we studied HA H7
295
complexes biochemically and by cryo-electron microcopy coupled to image analyses and 3D
297
arrangements of constituent HA molecules. The 3D map from cryo-electron microscopy displayed a
296 298 299
300
reconstruction. Our results suggested HA-complexes have heterogeneous numbers and
structure with the HA in a pre-fusion state with the HA1 globular head oriented at the top of the complexes and the HA2 stem regions in a pre-fusion state suggesting that the orientation of the transmembrane regions are approximately towards the center of the HA-complexes.
301
Hierarchy of molecular interactions within HA complexes.
303
complexes may reduce vaccine potency (17, 35), we analyzed banding patterns of the H7 HA
305
strength of interactions between HA proteins. Single major bands at ~70kDa suggested HAs were
302 304
Because partial cleavage of HA0 during purification and storage of recombinant HA
proteins by SDS-PAGE and immunobloting to determine both the HA cleavage status and relative
306
homogenous proteins and in the uncleaved HA0 state (Figure 1A). This is important because
308
protein sequences, H7 Anhui and H7 Netherlands have predicted molecular weights of 62.1kDa and
310
glycosylation of the HAs. HA molecules formed oligomers as indicated by a ladder of bands in
307 309
proteolytic degradation can decrease antigen stability in HA-complex based vaccines (36). Based on
62.5 kDa, respectively, and the higher apparent molecular weights of HA are likely explained by
12
311
immunoblots. HA-complexes appeared as “starfish-like” clusters of HA molecules in cryo-electron
313
cryo-electron microscopy image processing (Figure 4C). This would represent 24 (8x3) protomers
312 314
microscopic images (Figures 2A, 3A). We estimated up to eight HA molecules in a HA complex by of HA. However, in SDS-PAGE/immunoblots only six bands were observed and these can be
315
interpreted as representing one, two, three, four, five, and six protomers of HA0 (Figure 1B). This
317
complex were abrogated to a single HA monomer state as judged by the appearance of bands
316 318 319 320 321
322
implies that the SDS conditions affected the complex, but that not all interactions within the representing integral weights of HA (Figure 1B). Detergent is used and removed during the
purification of HA0 from the membranes of recombinant systems. This suggests a model for a hierarchy of interactions in which HA-complexes may be formed by the association of one HA trimer to another as detergent is removed during the protein purification process.
Redox conditions could also affect the hierarchy of these interactions, because reduction of
323
disulfide bonds in the presence or absence of heat denaturation conditions both abrogated the
325
cysteines forming intra and inter-disulfides bonds of HA not only in order for the 3D structure of HA
324
326
327
observation of HA oligomers (Figure S2, S3). This implies required integrity of the conserved
to be maintained (37) but also to enforce other interactions within the HA-complex. Indeed, the
formation of non-native disulfides can reduce recombinant HA vaccine potency (35) and certain
328
chemical additives can lessen non-native disulfides (38). The correct disposition of disulfides bonds
330
residues are important for assembly and epitope display (39). Our results with mAb414 indicate
329
has be studied in other systems such as hepatitis B surface antigen particles, where cysteine
331
that antibodies with conformational, disulfide-dependent epitopes can be used to assess proper
333
Arrangements and numbers of constituent HA molecules within complexes.
335
particles, such as virus-like particles of human papillomavirus (HPV) (40, 41) and surface antigen
332
334
folding and oligomerization hierarchy of HA molecules within HA-complexes.
Some viral vaccines are composed of recombinant viral antigens that are arranged as rigid
13
336
337
particles (sAg) of hepatitis B virus (42-44) and can have hundreds of copies of the viral antigen.
However, the rigidity, flexibility, and HA numbers within HA-complexes used in influenza vaccines
338
had not been analyzed in detail (17). From 2D classification we would expect to detect rigid
340
processing as shown for symmetrical particles such as HPV and designed influenza nanoparticles
339 341 342
343 344 345
346 347 348
349
350
symmetrical arrangements of molecules because such arrangements can be detected by image (41, 45). Our results using cryo-electron microscopy and 2D classification indicated non-rigid structures with no apparent consistent symmetry for a population of several thousand HA-
complexes (Figure 3A, 3B). However, HA-complexes did classify into different classes and curve analyses revealed differences in relative diameters of HA-complex classes (Figure 3). The
differences between the shapes and peak maximums of class profile curves suggest that HA-
complexes can exist in at least to major states: open and close states. In open states the constituent HA molecules of the complexes are more spread out radially from the center like the arms of an open starfish. In closed states there is more side-by-side or collimated arrangement of HA
molecules like the arms of a closed starfish. Further evidence for this is seen in images of individual complexes where apparent apical (top-views) and lateral (side-views) of HA molecules emanate
351
from a common center (Figure 2A-2D, Figure 5A-5D). For example, particles with more top-views
353
represent an opened state of the HA complex (Figure 2D). A possible third mixed state with HA-
355
2C). Although appearing non-rigid with asymmetric arrangements of constituent HA molecules, HA
352 354 356 357
358
359
of HA could represent a closed state (Figure 2B) while particles with more lateral-views of HA could
complexes with both apical and lateral views of HA molecules could represent a mixed state (Figure complexes did display HA molecules as protrusions emanating from a common center to give rise to starfish-like arrangements of HA molecules within 2D classes. However, visually counting the
number of HA molecules in individual HA complex images (Figure 2A-2D) and 2D images classes
(Figure 3A) would be subjective.
14
360
Thus, we devised a more objective computational strategy to estimate the number of HA
361
molecules per complex within the 2D classes by correlating rotated images and plotting the
363
number of epitopes in the HA-complex. This is important because multivalency of viral antigens is
362
364
365
correlations (Figure 4). Also, the number of HA molecules informs on the multivalency via the
thought to improve immunogenicity (40, 45). The number of estimated HA molecules ranged from 6 to 8 molecules for HA complexes (Figure 4B-4E). Thus, trimeric HA within HA-complexes could
366
have a predicted maximum of 24 (3x8) epitopes per complex. This number of epitopes is much less
368
this predicted maximum of 24 epitopes is much less than that for the virus-like particle of HPV
370
formulated into efficacious influenza vaccines (17, 18, 46). Moreover, designed octahedral
372
models (45, 47). These octahedral nanoparticles contain 24 copies of the HA polypeptide like the
367 369 371
373
than influenza virus particles, which have hundreds of HA molecules and epitopes (21, 28). Also,
vaccine which has 360 copies of the major capsid protein L1 (40). However, HA-complexes can be
nanoparticles that display epitopes for hemagglutinin can elicit protective antibodies in animal maximum 24 epitope copies for the HA-complex with 8 trimeric HA molecules. Taken together
374
these results suggest that the display of hemagglutinin (HA) epitopes can occur on particles that
376
influenza virus surface and HA designed nanoparticles were shown to be in the pre-fusion state by
375
377 378
379
380
differ in both numbers and arrangements of constituent HA molecules. However, HA on the
cryo-electron microscopy (28, 45) and in this study we determined that HA within HA complexes to be in the pre-fusion state, as well.
Hemagglutinin (HA) conformation in a pre-fusion state. Because influenza hemagglutinin (HA) can exist in pre-fusion and post-fusion states (20, 22,
381
28) the conformational state of HA within HA-complexes was determined by image analysis and 3D
383
observed in soluble HA ectodomains without transmembrane regions and HA with transmembrane
382 384
reconstruction coupled with molecular modeling. Although structures of pre-fusion HA have been
regions on viral surfaces (20, 22, 28), the 3D conformation of HA within recombinant HA-complexes
15
385
used as vaccine immunogens had not been addressed in great detail. Previous studies have used
387
structural flattening and deformations can occur. Also, images of only a small number of particles
386 388
389 390
391
392
heavy metal stains and dehydrated samples fixed to a carbon support and under these conditions (
