Rational Design of A Triple-Type Human Papillomavirus Vaccine by Compromising Viral-Type Speci Ficity

ARTICLE
DOI: 10.1038/s41467-018-07199-6 OPEN
Rational design of a triple-type human

papillomavirus vaccine by compromising
viral-type specificity
Zhihai Li1, Shuo Song1, Maozhou He2, Daning Wang2, Jingjie Shi1, Xinlin Liu1, Yunbing Li1, Xin Chi2,
Shuangping Wei1, Yurou Yang1, Zhiping Wang1, Jinjin Li2, Huilian Qian2, Hai Yu1, Qingbing Zheng1,
Xiaodong Yan2,3, Qinjian Zhao2, Jun Zhang2, Ying Gu1,2, Shaowei Li 1,2 & Ningshao Xia 1,2
1234567890():,;
Sequence variability in surface-antigenic sites of pathogenic proteins is an important obstacle

in vaccine development. Over 200 distinct genomic sequences have been identified for
human papillomavirus (HPV), of which more than 18 are associated with cervical cancer.
Here, based on the high structural similarity of L1 surface loops within a group of phylo-
genetically close HPV types, we design a triple-type chimera of HPV33/58/52 using loop
swapping. The chimeric VLPs elicit neutralization titers comparable with a mix of the three
wild-type VLPs both in mice and non-human primates. This engineered region of the chimeric
protein recapitulates the conformational contours of the antigenic surfaces of the parental-
type proteins, offering a basis for this high immunity. Our stratagem is equally successful in
developing other triplet-type chimeras (HPV16/35/31, HPV56/66/53, HPV39/68/70,
HPV18/45/59), paving the way for the development of an improved HPV prophylactic
vaccine against all carcinogenic HPV strains. This technique may also be extrapolated to
other microbes.
1 State Key Laboratory of Molecular Vaccinology and Molecular Diagnostics, School of Life Sciences, Xiamen University, Xiamen, China 361102. 2 National
Institute of Diagnostics and Vaccine Development in Infectious Disease, School of Public Health, Xiamen University, Xiamen, China 361102. 3 Department of
Chemistry and Biochemistry and Division of Biological Sciences, University of California-San Diego, San Diego, CA 92093-0378, USA. These authors
contributed equally: Zhihai Li, Shuo Song, Maozhou He. Correspondence and requests for materials should be addressed to Y.G. (email: [email protected])
or to S.L. (email: [email protected]) or to N.X. (email: [email protected])
NATURE COMMUNICATIONS | (2018)9:5360 | DOI: 10.1038/s41467-018-07199-6 | www.nature.com/naturecommunications 1

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-07199-6
V
accines are highly efficient weapons against infectious Previous studies on HPV structures have shown that the HPV
disease. However, multiple antigenic types or subtypes L1 monomer forms a canonical, eight-stranded β-barrel (BIDG-
derived from the evolution of pathogenic microbes CHEF), and both biochemical22–27 and structural analyses28–30
through sequence variation presents a serious obstacle in vaccine indicate that HPV type-specificity is largely dependent on five
development. One way to tackle this variation is to include more highly variable surface loops—BC, DE, EF, FG and HI—on the
antigenic variants into a single vaccine, as exemplified with the HPV capsid31,32. It stands to reason that a chimeric recombinant
Streptococcus pneumoniae vaccine1 and Human papillomavirus L1 capsid comprising the antigenic determinants of different
(HPV) prophylactic vaccine2. Yet, because pathogens, such as the HPV types may present an alternative strategy for the develop-
influenza viruses and human immunodeficiency virus (HIV), ment of a cross-type or wide-spectrum HPV vaccine. Previous
have very high levels of antigenic variability, this approach is reports have observed that the recognition region (i.e., homo-
fraught with difficulties, as an increase in type coverage will logous sequences on the surface loops) of type-specific neu-
dramatically enhance protein amount and adjuvant level per dose, tralizing antibodies of one HPV type (type A) could be
as well as increase the manufacturing complexity and associated transferred to create a chimeric VLP of another type
production costs. Studies that focus on designing immunogens (type B)23,33–35. However, previous studies show that such a
capable of inducing a broader protection against multiple sub- chimeric VLP of type B only generated low levels of neutralizing
types or variants require technical methods, such as computa- antibodies that could cross-react with type A22,34, and because of
tionally optimized broadly reactive antigen (COBRA)3, which this poor response, there was little further interest in developing
uses the consensus sequence from multiple variants to increase chimeric L1 VLPs based on such epitope grafting.
the immunogenicity of the conserved epitopes that are shared In this study, we sought to develop a highly efficacious, cross-
between subtypes and targeted by broadly neutralizing antibodies type vaccine candidate against the infection of two or more HPV
among subtypes4–6. As yet, however, few studies have been suc- types using an epitope grafting approach. We initially found that
cessful, and there is thus an urgent need to identify or design the L1 proteins of various HPV types shared an overall conserved
antigens that can elicit antibodies with high and broad anti-virus structure in their core regions and even in their basic surface loop
potency. configurations, with phenotype variations between different HPV
Oncogenic HPV infection is associated with several malig- types able to be induced by minor structural movements of the
nancies, including cervical and anogenital cancer7. To date, more surface loops. Our design is based on a rational understanding of
than 200 distinct HPV genotypes have been identified, of which at the high structural conservation of L1 surface loops within phy-
least 18 belong to the “high-risk” group and are chiefly respon- logenetically close HPV types despite sequence variation; and we
sible for the development of cancer8–10. HPVs are non-enveloped, used this knowledge to select among hundreds of HPV variants to
double-stranded DNA viruses comprising multiple copies of the engineer chimeric particles. Here we characterized a chimeric
major (L1) and minor (L2) capsid proteins. The native T = 7 HPV cross-type vaccine candidate with comparable efficacy to
HPV virion can be mimicked by an empty icosahedral shell, that of mixed wild-type (WT) VLPs. Our study confirms that the
called a virus-like particle (VLP) which consists of 72 L1-only strict type specificity of HPV can be undermined by the intro-
pentamers11,12. High-level neutralizing antibodies elicited by L1 duction of a few residue substitutions between genetically close
VLPs can block HPV infection; however, protection against HPV HPV types, introducing an appealing strategy for the design and
infection is type-restricted, and there tends to be little cross- development of wider-spectrum vaccines against HPV as well as
reactivity among HPV types13–15. Pre-clinical and clinical data other pathogenic microbes.
show that vaccinated patients exhibit low titers of neutralizing
antibodies against genetically related, non-vaccine HPV types in
highly reactive immune sera, and these antibodies may drop Results
below protective levels sooner than type-specific ones16–18. For Genetically close HPV types share high structural similarity.
these reasons, the current market-available prophylactic HPV The type-specific neutralization epitopes of HPV, which are
vaccines: Cervarix19, Gardasil20 and Gardasil 9 are multivalent located on the surface of the HPV capsid, are mostly mediated by
vaccines of the various HPV L1 proteins. For example, Gardasil 9 the pentamers36. To ascertain whether structural divergence drives
provides type-specific protection against infection from nine HPV type specificity, we scrutinized the crystal structures of L1
HPV types (HPV6, -11, 16, -18, -31, -33, -45, -52 and -58), the pentamers of genetically close or distant HPV types (Fig. 1a, b).
last seven of which are responsible for about 90% of cervical First, L1 proteins of HPV33 and HPV52 were cloned and
cancers. expressed in E.coli, purified, digested by trypsin, and crystalized.
The evolutionary history of HPV indicates a slow rate of HPV The crystal structures of HPV33 and HPV52 pentamers were
mutation21; however, it is possible that genetic mutation and determined to resolutions of 2.9Å and 2.75Å, respectively (Table 1,
recombination between different viruses may result in vaccine Supplementary Fig. 1a), with refinement to Rwork/Rfree values of
evasion; for example, a recombination event between HPV16 and 21.0/24.0% and 18.7/23.5%, respectively. Both structures show a
HPV31 may result in a virus carrying HPV16 oncogenes but hollow lumen in the central axis, and two neighboring monomers
coding for HPV31 structural proteins21. Indeed, previous work intertwined by surface loops (Fig. 1b, Supplementary Fig. 1a).
shows that substitution of just a few residues on the FG loops of Each L1 monomer folds into a canonical, eight-stranded β-barrel
L1 proteins between HPV16 and HPV31 can create a new ser- (BIDG-CHEF), with the strands joined via flexible surface loops
otype, which would avoid the immunity conferred by both (BC, DE, EF, FG, HI; Supplementary Fig. 1b, c). The same sec-
types22. Furthermore, it remains unclear whether widespread ondary structural elemental profile is found among various
immunization with vaccines like Gardasil 9 would lead to an L1 structures (HPV11/16/18/35/58/59)12,31.
increase in infection rates from the other carcinogenic HPV types Next, we sought to determine the extent of the differences in
that are responsible for the remaining 10% of cervical cancer. the tertiary structures of L1 proteins among different HPV types;
Ideally, an even broader vaccine that could protect against these these are well defined in the electron density map, even for the
other types would help to avoid such concerns. However, the high variable loops located on the outer surface (Supplementary Fig. 2).
dosage already required for Gardasil 9 (270 μg protein in a single It should be noted that the surface loops in the non-
dose) presents a significant challenge for the development of such crystallographic symmetry (NCS)-related monomers within the
a broad-acting vaccine through standard manufacturing practices. asymmetric unit of HPV33, HPV52, and HPV58 might differ
2 NATURE COMMUNICATIONS | (2018)9:5360 | DOI: 10.1038/s41467-018-07199-6 | www.nature.com/naturecommunications

NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-07199-6 ARTICLE
a b 147 175
290 316
Group 5 Group 4
DE
FG
HPV39
HPV68
HPV26 HPV69
374 384
HPV70 HPV51
HPV59
HI
HPV45 HPV53
HPV18 HPV56
90°
HPV66
Group 6
Group 3
195 213
76 93
HPV6
HPV58
HPV11
HPV33
Group 7 BC EF
HPV52 HPV16
Group 1
HPV31 HPV35
0.1
Group 2
Fig. 1 Phylogenetic classification among the major capsid proteins (L1) of multiple types, and a structural comparison of HPV L1s. a Phylogenetic tree of
HPV L1 proteins based on the amino acid sequences of 20 HPV types. The bar at the bottom provides a scale for the change in evolutionary lineages over
time. Twenty HPV types were divided into seven groups based on evolutionary distance of their L1 proteins: Group 1 (blue): HPV33, -52 and -58; Group 2
(cyan): HPV16, -31 and -35; Group 3 (purple): HPV53, -56 and -66; Group 4 (yellow): HPV26, -51 and -69; Group 5 (orange): HPV39, -68 and -70; Group 6
(pink): HPV18, -45 and -59; and Group 7 (green): HPV6 and -11. b Structural comparison of loop structure among HPV types of HPV33, -58 (PDB code:
5Y9E), -52, -18 (PDB code: 2R5I), and -59 (PDB code: 5J6R). The loops are colored by type: HPV33 (light blue), HPV58 (deep blue), HPV52 (marine),
HPV18 (light pink) and HPV59 (red). The loop sequence alignments among types are shown above the corresponding loop structure, with residue
differences compared with HPV58 colored red
Table 1 Data collection and refinement statistics
HPV33 HPV52 H58-33BC H58-33BC-52HI

Data collection
Space group P21 C2 P21 P1
Cell dimensions
a, b, c (Å) 98.8, 171.9, 145.7 306.8, 105.1, 196.9 153.7, 105.8, 154.7 136.5, 209.8, 212.6
α, β, γ (°) 90.0, 97.0, 90.0 90.0, 125.8, 90.0 90.0, 99.5, 90.0 60.5, 85.1, 90.1
Resolution (Å) 50.0–2.9 (2.97–2.92)a,b 50.0–2.7 (2.80–2.75) 50.0–2.5 (2.54–2.50) 50.0–3.5 (3.56–3.50)
Rsym (%) 9.5 (75.1) 19.9 (123.8) 13.4 (83.3) 17.1 (64.5)
I / σI 14.2 (2.1) 12.1 (1.8) 11.1 (1.8) 4.8 (1.3)
Completeness (%) 99.8 (99.9) 99.9 (100.0) 99.8 (99.7) 97.0 (87.2)
Redundancy 3.7 (3.7) 7.6 (7.6) 3.7 (3.7) 1.9 (1.8)
Refinement
Resolution (Å) 48.2–2.9 42.7–2.8 40.8–2.5 40.4–3.5
No. reflections 105,166 131,736 169,776 248,099
Rwork / Rfree 21.0/24.0 18.7/23.5 18.2/21.6 31.5/34.0
No. atoms 33,033 33,256 34,661 133,168
Protein 33,033 33,256 33,241 133,168
Water — — 1420 —
B-factors 76.1 53.0 40.7 105.5
Protein 76.1 53.0 40.9 105.5
Water — — 35.0 —
R.m.s. deviations
Bond lengths (Å) 0.003 0.007 0.004 0.003
Bond angles (°) 0.65 0.93 0.67 0.62
aOne crystal was used for each structure.
bValues in parentheses are for highest-resolution shell

slightly in conformation due to their intrinsic flexibilities HPV, which even extends across different groups, despite minor
(Supplementary Fig. 3). First, we divided 20 of the major HPV variations in the outer antigenic surface loops.
types into seven groups based on evolutionary distance of L1
proteins: 18 of these are responsible for more than 99% of cervical
cancer, with HPV6 and 11, in particular, accountable for about Design and expression of chimeric HPV VLPs. Viral epitopes
90% of genital warts (Fig. 1a, Supplementary Table 1). Then, five that induce neutralizing antibodies against HPV seem to cover
L1 pentamer structures representing two groups (group 1: more than one surface loop on a single L1 protein25,28,29,37,38; this
HPV58, -33 and -52; group 6: HPV18 and -59) were aligned to may explain why high variability in the loop region leads to type
compare the structural differences among their five surface loops. specificity among different HPV genotypes. Therefore, our ana-
We show that the overall loop structures are conserved for all five lysis on the structure of HPV33/58/52/18/59 types demonstrated
surface loops among various types of L1 structures. Nevertheless, that the L1 surface loops showed high structural similarities
the variable sequence identity of individual genotypes induces within a phylogenetic group, which remarkably extended across
small loop movements between types, as shown in the super- different groups, with only minor variation in the antigenic loops.
imposed structures. We also observe that the surface loops of This type of conserved structure despite genotypic evolution
types within group 1 or group 6 have quite similar main-chain illustrates the variability that can be induced while conserving
traces, with minor differences noted among the loops between the function, and then provides preliminary structural information to
two groups. The extent of loop movement between L1s are explain the cross-protection achieved by HPV vaccines against
correlated with aa sequence diversity in terms of residue type and infection with a limited number of phylogenetically related HPV
sequence length (Fig. 1b). Furthermore, pairwise structural types39,40. We thus propose that highly potent cross-type pro-
comparisons were carried out between NCS copies of L1 tection against HPV infection could be achieved by creating a
monomers from five HPV types and the mean root mean chimeric HPV VLP that contains antigenic determinants from
squared deviations (RMSDs) were generated for both the core genetically close HPV types in same phylogenetic group rather
regions and the surface loops of L1 for the various genotypes than evolutionarily distant HPV types across groups. Given the
(Supplementary Table 2). In the comprehensive structural extent of the structural diversity among the five surface loops, it is
comparisons, both the core region structures and the surface assumed that comprehensive loop swapping trials could be
variable regions—constituted by five surface loops among the implemented to screen for an appropriate chimera that may
various types of L1 structures—shared conserved structural confer the expected cross-type antigenicity and immunogenicity.
configurations, with RMSD values of Cα ranging from 0.22 to To test our hypothesis, we genetically swapped the loop
0.59Å and from 0.53 to 2.06Å, respectively (Supplementary sequences on five surface regions (BC/DE/EF/FG/HI) between the
Table 2). The minor loop movements, particularly between types two most phylogenetically related HPV types—HPV33 and
from group 1 and group 6 (Fig. 1b), slightly increased the RMSD HPV58—and generated 10 chimeric L1 constructs: H33-58BC,
values with respect to that of the relatively rigid core region, H33-58DE, H33-58EF, H33-58FG, H33-58HI, H58-33BC, H58-
which was believed to be associated with the various type-specific 33DE, H58-33EF, H58-33FG and H58-33HI (Fig. 2a and
phenotype variations underlying virus phylogenetic evolution. Supplementary Fig. 4). These mutants were expressed in E.coli
Collectively, the structural information suggests high conserva- (Supplementary Fig. 5a, b), and self-assembly of the purified
tion of the overall structures of L1 within a phylogenetic group of mutated L1 proteins was confirmed by transmission electron
a BC DE EF FG HI
b HPV33 specific mAbs HPV58 specific mAbs
9A 7
C 2
A1 9
1D 8
5G 1
C 3
D
A
5A
0B
11 2
A3
A4
B1
C 4
2C
9
1
14
16
4
5
3
3
2
4
F
7C
5H
4E
7A
4A
A1
3E
11
13
10
1 121 150 263 290 499
>10k
HPV33
50 67 169 187 348 358
H33-58BC
100~1k 1k~5k 5k~10k
H33-58DE
H33-58EF
H33-58FG
H33-58HI
149 178 289 316

1 524
<100
HPV58
76 93 197 216 374 384
H58-33BC
EC50(ng)
H58-33DE
H58-33EF
H58-33FG
H58-33HI
Fig. 2 Molecular design and antigenic characterization of loop-swapped HPV33/58 chimeric VLPs. a Schematic representation of the wild types (WTs) and
the chimeric HPV33/58 L1 proteins. The location of surface loops for each WT or mutated L1 protein are labeled by amino acid position. b Heatmap
representations of the EC50 values of chimeric HPV33/58 VLPs based on ELISA assays against a type-specific mAb panel of HPV33 and HPV58 VLPs. The
key indicates the heatmap gradient. A detailed characterization of each mAb is shown in Supplementary Table 3

microscopy (Supplementary Fig. 5c). Particle size and homo- HPV58 VLPs against both HPV33 and HPV58 (Fig. 3a).
geneity of the chimeric HPV VLPs, as determined by high- Therefore, these two mutants were subjected to further study,
performance size-exclusion chromatography (HPSEC) and ana- testing immunogenicity assays in mice. We found that immuni-
lytical ultra-centrifugation (AUC), were similar to those measures zation with higher concentrations per dose led to higher titers:
observed for WT L1 VLPs of their corresponding backbone type animals receiving 0.1 μg per dose of chimeric VLP H33-58HI and
(Supplementary Fig. 5c). Using differential scanning calorimetry H58-33BC were capable of inducing anti-HPV58 and anti-
(DSC), we found that the transition corresponding to the melting HPV33 neutralization titers of up to 49,500 and 29,000,
temperature (Tm) in the DSC trace occurred at 77.28 °C for WT respectively (Fig. 3b); 1 μg per dose led to 1.5-fold
HPV33 VLPs and 68.61 °C for WT HPV58 VLPs, whereas the (50,000–92,000) and 3.3-fold (64,000–126,000) higher neutraliza-
Tm values for the chimeric HPV33 VLPs (H33-58BC, H33-58DE, tion titers, respectively, than with 0.1 μg per dose; and 10 μg per
H33-58EF, H33-58FG and H33-58HI) and chimeric HPV58 dose led to 2.0-fold (102,000–192,000) and 1.5-fold
VLPs (H58-33BC, H58-33DE, H58-33EF, H58-33FG and H58- (100,000–195,000) higher neutralization titers, respectively, than
33HI) ranged from 61 to 74 °C and 62 to 67 °C, respectively. with 0.1 μg per dose. These results were reflected by the ED50
These results indicated no obvious effects on the structural values of the chimeric VLPs; the ED50 values define the minimum
stabilities of the L1 VLPs following the amino acids substitutions effective dose that provides seroconversion in 50% of the animals
on the capsid surface regions (Supplementary Fig. 5). in a given group42. As shown in Table 2 and Supplementary
Table 4, the ED50 values of H33-58HI were as low as 0.053 μg for
HPV33 and 0.049 μg for HPV58, whereas, for H58-33BC, the
Antigenicity and immunogenicity of chimeric HPV33/58 values were 0.043 μg and 0.019 μg, respectively. This indicates
VLPs. To investigate the antigenicity of the chimeric HPV33/58 that VLPs of both H33-58HI and H58-33BC could provide cross
VLPs, we employed a panel of neutralizing antibodies against protection against HPV33 and HPV58 at levels comparable with
HPV33 and HPV58 (Supplementary Table 3); this allowed us to that of WT HPV58 and HPV33, respectively (Table 2).
evaluate the binding ability of the VLP mutants to those type- Furthermore, we questioned whether the cross protection
specific antibodies. The median effective concentrations (EC50) of afforded by these two chimeras was achieved by multiple
the mAb panel to various chimeric HPV33/58 VLPs was deter- simultaneous mutations. We generated eight single mutation
mined by ELISA assay (Fig. 2b). We observed the following three chimeras of HPV33 VLPs (mutated to the HPV58 HI loop: H33-
main findings: (1) The binding of HPV33 (or HPV58) mAbs was S350K, H33-D351E, H33-S352G, H33-E357D) and HPV58 VLPs
affected, to varying degrees, by loop substitutions at one of (mutated to the HPV33 BC loop: H58-S80N, H58-N82T, H58-
the five surface loops (Fig. 2b), suggesting that particular loop(s) N84A, H58-V87L). The neutralization titers of the anti-sera of
are responsible for mAb recognition. Notably, five of the nine these chimeras were measured and we found that most of the
HPV33-specific mAbs (C2C3, C14A2, C16D7, 9A3 and 7A3) single-point mutants (except for H58-V87L) could elicit hetero-
failed to bind to the BC loop-swapped mutant of HPV33 (H33- typic neutralization antibodies (Supplementary Fig. 6) to some
58BC), suggesting that the BC loop of HPV33 may involve a extent; this indicated that every residue associated with type-
major neutralization epitope. (2) Enhanced binding activity of restricted neutralization contributes to the cross-neutralization
multiple non-backbone-type mAbs to chimeric VLPs was found when substituted to the other type. Intriguingly, the H33-E357D
following the substitution of several amino acids on different chimera was able to induce a cross-type neutralization titer as
surface loops of HPV58 into HPV33 (and vice versa). Specially, high as that of H33-58HI (Supplementary Fig. 6), possibly
eight HPV58-specific mAbs showed binding to the DE loop- demonstrating that residue D377 of HPV58 L1 takes a critical
swapped mutant of HPV33 (H33-58DE; Fig. 2b and Supple- role during the immune recognition process of HPV58.
mentary Table. 3). (3) Binding of several mAbs to WT VLPs To check if cross-protection could be achieved by introducing
could be disrupted by single loop substitutions with the hetero- mutations from more distantly related HPV types, we generated
logous type; i.e., replacement of one of the four loops (BC, DE, EF two chimeric VLPs—H33-59HI and H58-59BC—using the
and FG) in WT HPV33 VLP abrogated the binding of the specific distant type HPV59, which belongs to another group in the
mAb 9A3, suggesting that these four loops are involved in mAb phylogenetic tree (Fig. 1a; Supplementary Fig. 7a, b). To evaluate
9A3 recognition of HPV33 VLP. their potential cross-type immunogenicity, mice were immunized
To assess the immunogenicity of the cross-type chimeric HPV33/ three times with a higher dose (100 μg) of proteins absorbed with
58 VLPs, BALB/c mice were inoculated three times with chimeric a stronger Freund’s adjuvant (Supplementary Fig. 7c). In contrast
VLPs formulated with aluminum adjuvant. The neutralization titers to the high neutralization titers against heterologous types elicited
of the sera were measured with a pseudo-virion–based cell by H33-58HI and H58-33BC, we found no increase in the anti-
neutralization assay. We noticed that the WT HPV33 and HPV59 antibody response in the antisera of either H33-59HI or
HPV58 VLPs can elicit moderate reciprocal cross-type neutraliza- H58-59BC (Supplementary Fig. 7c). Thus, cross-protective
tion titers ranging from 101.2 to 102.6, which is consistent with immunities was not be elicited by proteins engineered using
previous results41. Meanwhile, all of the chimeric HPV33/58 VLPs genetically distant types.
showed improved heterotypic antibody responses, and this occurred
without changing the elicitation of neutralization antibodies against
their own backbone types (Fig. 3a). Notably, the anti-HPV58 Immunogenicity of triple-type HPV33/58/52 chimeras. To
antibody titers of H33-58DE and H33-58HI were 25-fold (titers explore the feasibility of including more genetically close geno-
ranging from 10,000 to 102,400; P < 0.01) and 80-fold (102,400 to types into a single, chimeric L1 VLP type, HPV52—which is
212,400; P < 0.001) higher than that found in mice immunized with closest to both HPV33 and HPV58 (Fig. 1a)—was engineered
WT HPV33 VLPs (2.5 to 102,400), respectively (Fig. 3a), whereas into the lead HPV33/58 chimeric VLPs, H33-58HI and H58-
H58-33BC, H58-33DE, and H58-33HI could induce, respectively, 33BC, by individually swapping each of the remaining four loops
325-fold (24,000 to 204,800; P < 0.001), 94-fold (2,000 to 51,200; with those from HPV52. This resulted in eight HPV33/58/52
P < 0.01), and 52-fold (1,700 to 33,000; P < 0.01) higher titers chimeras: H33-58HI-52BC, H33-58HI-52DE, H33-58HI-52EF,
against PsV33 than that elicited by WT HPV58 VLPs (Fig. 3a). H33-58HI-52FG, H58-33BC-52DE, H58-33BC-52EF, H58-33BC-
H33-58HI and H58-33BC mutants showed comparable cross- 52FG and H58-33BC-52HI (Fig. 4 and Supplementary Fig. 8a–c).
protection potency as that conferred by bivalent HPV33 and The VLPs from these chimeras were then used to immunize mice

a 5 μg per mouse
**
7 *** **
Neturalization titer (log 10)

** ***
6
5
4
3
2
1
0
33
58
8
8H
3H
&5
8E
3E
8D
3D
8B
3B
8F
3F
PV
PV
-5
-3
33
-5
-3
-5
-5
-5
-3
-3
-3
H
33
58
33
58
33
33
33
58
58
58
PV
H
H
H
H
H
H
PsV33 PsV58
b 0.1 μg per mouse 1 μg per mouse 10 μg per mouse
7 7 7
*** **** **** **** *** ***

6 6 6
5 5 5
4 4 4
3 3 3
2 2 2
1 1 1
0 0 0
33
58
33
58
33
58
8H
8H
8H
3B
3B
3B
PV
PV
PV
PV
PV
PV
-5
-5
-5
-3
-3
-3
H
33
33
33
H
58
58
58
H
H
H
H
Fig. 3 Immunogenicity of loop-swapped HPV33/58 chimeric VLPs. a Neutralization titers of immune sera from BALB/c mice (n = 5) collected four weeks
after the third immunization (immunization at weeks 0, 2 and 4 with 5 μg per dose of aluminum adjuvant-containing chimeric or WT VLPs). A group of
mice immunized with mixed WT HPV33 and HPV58 VLPs was labeled. (*) denotes significant differences as compared with the neutralization titer induced
by the corresponding backbone type. b Groups of mice (n = 6) were immunized with 0.1, 1 or 10 μg per dose of the selected chimeric or WT HPV33/58
VLPs at weeks 0, 2 and 4. Immune sera were collected at week 8 and analyzed by neutralization assays. All the data were analyzed by one-way analysis of
variance (ANOVA), **P < 0.01; ***P < 0.001; ****P < 0.0001. The error bars are standard deviation and symbols represent individual mice. The dotted line
indicates the limit of detection for the assay. Non-neutralizing sera were assigned a value of one-half of the limit of detection for visualization
to assess their cross-type immunogenicity. Of the eight triple-type

chimeric VLPs, H33-58HI-52DE and H33-58HI-52FG showed
Table 2 Half-effective dosage (ED50) of double-type and the highest anti-HPV52 antibody titers (4700–2200 and
triple-type chimeras in mice 1300–47900, respectively; P < 0.001) as compared with the H33-
58HI control. H58-33BC-52DE and H58-33BC-52HI also
Antigen HPV VLP ED50 (μg) showed high titers (2,280–23,440 and 2,180–87,430, respectively;
category P < 0.001) as compared with the H58-33BC control (Fig. 4b). No
anti- anti- anti- anti-HPV52 neutralizing antibodies were found for any of the
HPV33 HPV58 HPV52 controls (WT HPV33, WT HPV58, or HPV33/58 chimeras).
Wild-type HPV33 0.066a > 0.3 — Intriguingly, two of the triple chimeras—H33-58HI-52FG and
control H58-33BC-52FG—showed lower anti-HPV58 neutralizing titers
HPV58 > 0.3 0.006 — than those of their backbone VLPs (30-fold [1,700–10,640] and
Double-type H33-58HI 0.053 0.049 —
330-fold [2.5–5,400] for H33-58HI and H58-33BC, respectively);
chimera
H58-33BC 0.043 0.019 —
this suggested that the FG loop (in HPV58 in particular) was part
Wild-type HPV33 0.013b >0.3 >0.3 of the immunodominant region associated with the neutralizing
control antibody production. Anti-HPV33 neutralization titers for all
HPV58 > 0.3 0.021 >0.3 triple chimeras remained comparable with that of their parental
HPV52 >0.3 >0.3 0.046 double chimeras.
H33&58&52 0.009 0.009 0.017 Next, we further measured the anti-HPV52 immunogenicity of
Triple-type H33-58HI-52DE 0.083 0.069 0.557 four selected candidates—H33-58HI-52DE, H33-58HI-52FG,
chimera H58-33BC-52DE and H58-33BC-52HI—using three different
H33-58HI-52FG 0.046 0.243 0.087 immunization doses (0.1, 1 and 10 μg), as before. All four
H58-33BC-52DE 0.009 0.014 0.341
mutants showed dose-dependent anti-HPV52 neutralization
H58-33BC-52HI 0.065 0.058 0.041
antibody responses (Fig. 5a). However, H33-58HI-52FG, which
aRefer to the detailed ED50 calculation in Supplementary Table 4 showed moderate antibody titers (2,280–23,440) in the cross-type
bRefer to the detailed ED50 calculation in Supplementary Table 5
immunogenicity assays, failed to show high potency against

a BC DE EF FG HI b
PsV33 PsV58 PsV52
1 147 176 292 321 529
H52 VLP
76 95 195 213 379 389
H33 VLP
H58 VLP
H33-58HI
****
H33-58HI-52BC
H33-58HI-52DE
***
H33-58HI-52EF
H33-58HI-52FG
H58-33BC
****
H58-33BC-52DE
***
H58-33BC-52EF
H58-33BC-52FG
H58-33BC-52HI
H33&58&52
0 1 2 3 4 5 6
Neutralization titer (log10)
Fig. 4 Immunogenicity of loop-swapped HPV33/58/52 chimeric VLPs. a Schematic representation of the chimeric HPV33/52/58 L1 proteins based on
chimeric L1 proteins H33-58HI and H58-33BC. A group of mice were immunized with mixed WT HPV33, -52 and -58 VLPs. b Neutralization titers of the
immune sera from BALB/c mice (n = 4) collected 4 weeks after the third immunization (immunization at weeks 0, 2 and 4 with 5 μg per dose of aluminum
adjuvant-containing tripe-type chimeric or WT VLPs). All the data were analyzed by one-way ANOVA, **P < 0.01; ***P < 0.001; ****P < 0.0001. The error
bars are standard deviation and symbols represent individual mice. The dotted line indicates the limit of detection for the assay. Non-neutralizing sera were
assigned a value of one-half of the limit of detection for visualization
HPV52 infection in groups receiving doses as low as 0.1 μg or 1 Immunogenicity and immune durability in non-human pri-
μg (Fig. 5a, left and middle). Also, the ED50 values for these four mates. We next immunized cynomolgus macaques (Macaca
chimeras were measured to test their immunogenicity against Fascicularis) with the lead HPV33/58/52 vaccine candidate to
infection of all three HPV types. Overall, only H58-33BC-52HI evaluate vaccine-induced immunity. Monkeys (n = 4 per group)
conferred high immune efficacy against infections from all three were administered with 5 μg of the H58-33BC-52HI chimeric
genotypes (HPV33 [ED50, 0.065 μg], HPV52 [0.041 μg] and VLPs formulated with aluminum adjuvant at week 0 and 8.
HPV58 [0.058 μg]), reaching comparable levels to those provided Neutralization titers were measured from antisera drawn at week
by a mix of the WT VLPs of HPV33, -52, and -58 (0.009 μg, 10. Immunization with WT HPV VLPs resulted in minor cross-
0.017 μg, and 0.009 μg, respectively; Table 2 and Supplementary neutralization of HPV33 and HPV58 in monkeys (titers ranging
Table 5). from 50 to 500); however, no cross-reactive antibodies to HPV52
were observed with HPV33 or HPV58 WT VLPs, and similarly,
Immunogenicity of other chimeric VLP combinations. To test no cross-reactive antibodies to HPV33 or HPV58 were detectable
whether this method could be applied to produce other triple after HPV52 VLP immunization (Fig. 6a). Nevertheless, the
HPV types, we attempted to graft immunodominant epitopes into cross-reactive antibody titers for HPV33 were significantly (26-
a single L1 molecule within different groups. Four chimeras fold) higher with the chimeric VLPs (960 to 6,290, P < 0.001;
(H35-16FG-31BC, H66-56HI-53FG, H39-68FG-70HI and H45- Fig. 4a, left) than with the HPV58 WT VLPs; reciprocally, the
18BC-59HI) were generated for Groups 2, -3, -5 and -6 (Fig. 1a) anti-HPV58 antibody titer (6,060 to 40,960) of H58-33BC-52HI
based on the protocol summarized for the generation of HPV58- was also 54-fold higher than that of HPV33 WT VLPs (P < 0.001;
33BC-52HI, and were tested in mice (5 μg per dose). All four Fig. 6a, right). The chimeric VLPs could also elicit a mean neu-
chimeric VLPs showed high cross-protective efficacy for each tralization titer of 5,400 (1,780 to 8,460) against HPV52 (Fig. 6a,
group (Fig. 5b and Supplementary Fig. 9), with at least 150-fold middle). Thus, engineered loop swapping can produce high
increased heterotypic neutralization titer achieved by these lead immune potency and provide cross-type protection of phylo-
candidates as compared with the backbone type, without sig- genetically related HPV types.
nificantly affecting the original immunization response (Supple- We then tested the kinetics of the vaccine-induced neutralizing
mentary Table 6). Thus, we believe that engineering in the surface antibodies in immunized monkeys. Two weeks after the first
loops between closely related HPV types can be used to optimally vaccination, the anti-HPV33 neutralizing titer reached ~200, with
modulate the immunodominance among different types. the titers then slightly decreasing until the second immunization

a PsV33 PsV58 PsV52
0.1 μg per mouse 1 μg per mouse 10 μg per mouse
**** ****
* ** ****
7 * 7 **** 7 ****
**** **** ****
6 6 6
5 5 5
4 4 4
3 3 3
2 2 2
1 1 1
0 0 0
8H DE
3B 2FG
3B DE
I
33
52
&5 8
58
8H DE
3B 2FG
3B DE
I
33
52
&5 8
58
8H DE
3B 2FG
3B DE
I
33
52
&5 8
58
2H
2H
2H
5
5
PV
PV
PV
2&
PV
PV
2&
PV
PV
PV
2&
2
2
-5
-5
-5
PV HP
I -5
I -5
58 C-5
I -5
I -5
58 C-5
I -5
I -5
58 C-5
C
H
8H
8H
8H
33
33
33
-5
-5
-3
-5
-5
-3
-5
-5
-3
PV
PV
-3
-3
-3
33
33
33
33
33
33
58
58
58
H
H
H
H
H
H
b Group 2 Group 3 Group 5 Group 6
PsV16 PsV35 PsV31 PsV56 PsV66 PsV53 PsV39 PsV68 PsV70 PsV18 PsV45 PsV59
7 7 7 7 ****
**** *** ***
6 **** 6 *** 6 *** 6 **

5 5 5 5
4 4 4 4
3 3 3 3
2 2 2 2
1 1 1 1
0 0 0 0
C
16
31
35
35
56
66
53
53
39
68
70
70
18
45
59
59
0H
9H
1B
3F
PV
PV
PV
1&
PV
PV
PV
6&
PV
PV
PV
8&
PV
PV
PV
5&
-7
-5
-3
I-5
&3
&6
&6
&4
H
H
G
6H
8B
8F
16
56
39
18
6F
-5
-6
-1
PV
PV
PV
PV
-1
66
39
45
35
H
H
H
H
Fig. 5 Immunogenicity of loop-swapped HPV33/58/52 chimeric VLPs at three dosages and other triple-type chimeric VLPs. a Groups of mice (n = 5) were
immunized with 0.1, 1 or 10 μg per dose of selected chimeric or WT VLPs at weeks 0, 2 and 4. Immune sera were collected at week 8 and analyzed by
neutralization assays. b Neutralization titers of immune sera (at week 8) in BALB/c mice (n = 5) immunized with 5 μg per dose of chimeric and WT L1
proteins of Groups 2, -3, -5, and -6 (also see Fig. 1a) in aluminum adjuvant at weeks 0, 2 and 4. All the data were analyzed by one-way ANOVA, **P < 0.01;
***P < 0.001; ****P < 0.0001. The error bars are standard deviation and symbols represent individual mice. The dotted line indicates the limit of detection
for the assay. Non-neutralizing sera were assigned a value of one-half of the limit of detection for visualization
at week 8 (Fig. 6b, left). Titer levels peaked at four weeks after the antibody production can be enhanced in pre-existing immunity
second vaccination (week 12), slightly decreasing over time but and immunization with these chimeric VLPs to provide long-
remaining high even after four and a half months (~2,200 at week lasting (~4.5 months) protection, at levels comparable with that
26); albeit, this titer was ~2-fold lower than that of mixed WT achieved by mixing three WT HPV VLPs at three times the
VLPs of HPV33/52/58 (Fig. 6b, left). Similar kinetic curves were dosage (versus one dose with the chimeric VLPs; Fig. 6b).
observed for anti-HPV52 (Fig. 6b, middle) and anti-HPV58
neutralizing titers (Fig. 6b, right), with the overall titer against
HPV52 and HPV58 reaching 1,100 and 22,000 at week 26, Structures of chimeric pentamers and VLPs-Fab complexes.
respectively (Fig. 6b). Neutralizing antibody titers in monkeys We solved the crystal structures of pentamers of the lead double
immunized with WT HPV33 or WT HPV58 VLPs showed and triple chimeric candidates H58-33BC and H58-33BC-52HI to
increased reciprocal neutralization after the first and second resolutions of 2.5Å and 3.5Å, respectively (Fig. 7b and Supple-
vaccinations but these also reduced over time, with low cross- mentary Fig. 11a; Table 1). The two pentamer structures of the
protective antibody titers of ~25 (anti-HPV58) for HPV33 and chimeric mutants were superimposed against wild-type structures
~33 (anti-HPV33) for HPV58 by week 26 (Fig. 6b, left and right). of HPV58, 33 and 52 to identify marked structural differences
In monkeys receiving HPV52 VLPs, no cross-reactive antibodies (Fig. 7b-d). Consistent with the high structural similarities within
were found for either HPV33 or HPV58 at week 26 (Fig. 6b, left this phylogenetic group, as indicated in Fig. 1, the type-specific aa
and right), as anticipated. Overall, the data confirm the high replacement in both the BC loop and HI loop still did not sub-
immune potency and cross-type protection afforded by the stantially change the main-chain conformation of these two loops
HPV33/52/58 chimeric VLPs designed through the exchange of in the chimeras (Fig 7c, d top left). Further pairwise comparisons
several amino acids between phylogenetically close related types of the surface contours of wild types and mutants illustrated that
(Fig. 7a and Supplementary Fig. 10), and also indicate that the most distinct bumpy protrusions between parental and

a PsV33 PsV52 PsV58

*** *** ****
**** **** ***
6 6 6
5 5 5
4 4 4
3 3 3
2 2 2
1 1 1
0 0 0
33
52
58
58
33
52
58
58
33
52
58
58
I
I
2H
2H
2H
PV
PV
PV
2&
PV
PV
PV
2&
PV
PV
PV
2&
-5
-5
-5
&5
&5
&5
H
H
C
C
3B
3B
3B
33
33
33
-3
-3
-3
58
58
58
H
H
b H58-33BC-52HI 33&52&58 HPV33 HPV52 HPV58
PsV33 PsV52 PsV58

6 6 6
5 5 5
4 4 4
3 3 3
2 2 2
1 1 1
0 0 0
0 4 8 12 16 20 24 0 4 8 12 16 20 24 0 4 8 12 16 20 24
PsV33 Weeks Weeks
Fig. 6 Immunogenicity of triple-type chimeric HPV33/52/58 VLPs in Cynomolgus Macaques. a Neutralization titers against PsV33, PsV52 and PsV58,
respectively, from the immune sera in monkeys (n = 4) collected two weeks after the second immunization (immunization at weeks 0 and 8 with 5 μg per
dose of aluminum adjuvant-containing chimeric H58-33BC-52HI or WT VLPs). Each symbol represents one monkey. Statistical significance was assessed
by one-way ANOVA, **P < 0.01; ***P < 0.001; ****P < 0.0001. b Kinetics of serum neutralization titers in monkeys after immunization with the chimeric or
WT VLPs. Immune sera were collected weekly after immunization, and serum neutralization titers were determined and plotted. The dotted line indicates
the limit of detection for the assay. Non-neutralizing sera were assigned a value of half of the limit of detection for visualization. All the neutralizing titers
were calculated to express the mean value with SD
targeting types were located on the side-chains of K53H33, capsid; although, much lower and discontinuous densities for the
N57H33, K59H33 and K60H33 for HPV58 vs. HPV33 (Fig. 7c, 4E5 Fabs bound to 6-coordinated pentamer were also observed
bottom left), and K380H52, S383H52 and E388H52 for HPV58 vs. (Fig. 8d, left), which could be mainly caused by steric hindrance
HPV52 (Fig. 7d, bottom left). By loop swapping with a few aa among bound Fabs30. Analysis of the extracted cryoEM map of
mutations on HPV58 L1—i.e., S80N, N82T, N84A and V87L for one 5-coordinated HPV33 pentamer and one 4E5 Fab revealed
BC loop, T375K, G378S and D383E for HI loop (Fig. 7a)—the that the 4E5 Fab is bound to the outer rim of the HPV33
resurfaced structural contours of the BC and HI loops of the capsomer at an angle of ~58° to the pentamer surface
chimera nearly recapitulate those of the HPV33 BC loop (Fig. 7c, (Supplementary Fig. 13c). Similar binding occupancies and
middle and right) and HPV52 HI loop (Fig. 7d, right), respec- binding angles (60° for H58-33BC:4E5 and 60° for H58-33BC-
tively. These subtle structural resurfacing changes in the engi- 52HI:4E5) were observed for 4E5 Fabs bound to L1 monomers of
neered regions might provide structural evidence for the cross- both H58-33BC:4E5 and H58-33BC-52HI:4E5 complexes (Fig. 8d,
protective immunity of the chimeric proteins. middle and right, and Supplementary Fig. 13c, middle and right).
To verify that the residues exchanged on the surface loops of This suggests that the same recognition mode was adopted by
HPV58 introduce a similar functional surface area as that of the mAb 4E5 to the VLPs of HPV33 and the chimeric HPV58/33 and
corresponding heterologous type, we recruited mAb 4E5—a HPV58/33/52. Further analysis of the mapped epitopes suggests
conformational and type-specific neutralizing antibody of that the 4E5 Fab binding orientation and footprints covering both
HPV33, the binding of which could be redirected to H58-33BC the BC loop and the EF loop are similar for all three VLPs-Fab
and H52-33BC-52HI (Figs 2b, 8a, c, d, Supplementary Fig. 12 and complexes (Fig. 8e and Supplementary Fig. 13d). These data
Supplementary Table 7)—and determined cryoEM structures of provide strong evidence that loop swapping of the BC loop from
the 4E5 Fab in complex with HPV33 WT VLPs (HPV33:4E5) and HPV33 onto the HPV58 basal framework can mimic the
both chimeric VLPs (H58-33BC:4E5 and H58-33BC-52HI:4E5). molecular surface of HPV33 and give rise to heterotypic
The cryo-EM structures of HPV33:4E5, H58-33BC:4E5 and H58- neutralizing antibodies in vivo.
33BC-52HI:4E5 were reconstructed to 12.3Å, 12.0Å and 10.0Å,
respectively (Fig. 8d, Supplementary Fig. 13; Supplementary
Table 8). From the cryoEM micrographs and 2D class averages, Discussion
we can discern obvious protrusions, which we consider as the Genetic mutation and recombination are routinely used by
bound 4E5 Fabs, for both chimeric VLPs:4E5 complexes, similar pathogens to avoid host immune system recognition, particularly
to that observed for the HPV33:4E5 complex (Supplementary in surface-exposed regions which are targeted by the immune
Figs. 13a, 14). The HPV33:4E5 complex map shows strong machinery; this also happens to be where most of the neutralizing
density corresponding to five bound 4E5 Fabs around the epitopes are located. The immunodominant epitopes of
periphery of the upper rim of the 5-coordinated pentamer of the HPV pathogens usually involve multiple surface regions that

a b
BC loop HI loop
HI
Conservation
76
. 93
. 374
. 384
.
HPV58 FSIKSPNNNKKVLVPKVS VTKEGTYKNDN
50
. 67
.
HPV33 ----N-T-A--L------
BC
379
. 389
.
HPV52 -K--S----E-
76
. 93
. 374
. 384
.
H58-33BC ----N-T-A--L------ -----------
76
. 93
. 374
. 384
.
H58-33BC-52HI ----N-T-A--L------ -K--S----E-
c BC loop d HI loop
HPV58 vs. H58-33BC HPV58 vs. H58-33BC-52HI HPV58 vs. H58-33BC-52HI
T56H33 H33 N83 G378H58
N54 N83 S383H52
K79
N82H58 S80H58 S378
T375H58
A58H33
N84H58 K85 K380H52

K85 E388H52 K375
V87H58 K86 D383H58
L61H33 K86 E383
HPV58 vs. HPV33 HPV33 vs. H58-33BC HPV33 vs. H58-33BC-52HI HPV58 vs. HPV52 HPV52 vs. H58-33BC-52HI
N57
K53
K79
S383
K79
K59
K85
K85
K380
K60 K86 E383
K86 E388
Fig. 7 Crystal structures of the chimeric pentamers. a The modified L1 sequence on the BC and HI loop(s) of the lead candidates of the double- and triple-
type chimeric VLPs: H58-33BC and H58-33BC-52HI. The consensus is determined based on an alignment of L1 sequences from 20 different HPV
serotypes (also see Supplementary Fig. 4). b A structural comparison of pentamers among HPV58 (gray), HPV33 (light blue), HPV52 (pale cyan), and the
chimeric proteins H58-33BC (pale green) and H58-33BC-52HI (light pink). Loops BC and HI are boxed in red and black, respectively. c Structural
comparisons of the BC loops of WT (HPV58 [gray], HPV33 [light blue]) and chimeric L1 proteins (H58-33BC [pale green], H58-33BC-52HI [pale pink]).
Upper left: Ribbon display of superimposed main-chain carbon atoms. Side-chains of the different amino acids between HPV58 and HPV33 are labeled and
shown as sticks. Upper middle, right and lower: Pairwise structural comparisons with surface display among HPV58, HPV33, H58-33BC and H58-33BC-
52HI. d Structural comparisons of HI loops of WT (HPV58 [gray], HPV52 [pale cyan]) and the chimeric L1 protein (H58-33BC-52HI [light pink]). Upper
left: Ribbon display of superimposed main-chain carbon atoms. Side-chains of the different amino acids between HPV58 and HPV52 are labeled and shown
as sticks. Upper right and lower: Pairwise structural comparisons with surface display among HPV58, HPV52 and H58-33BC-52HI. For presentation clarity,
the surface of the chimera in the pairwise surface comparisons in c, d is shown with 50% transparency and embracing the model rendered in stick mode;
the surface of the wild-type model is shown with full opaqueness
are discontinuous in the primary sequence, so that type VLPs (Fig. 3a). One possible explanation is that the EF loop is
specificity persists even following one or a few amino acids not a major component in determining HPV type-specificity.
substitutions22–27,34. Based on our immunogenicity assays in Alternatively, it is possible that the higher flexibility displayed on
mice and non-human privates, we have successfully generated the EF loop than the other loops makes it more difficult to
triple-type chimeric HPV proteins that can simultaneously pre- provide a similar heterotypic functional area through EF loop
sent three independent immunodominant epitopes, using seven exchange.
or fewer amino acid substitutions. We believe that our success is Although the chimeric HPV VLPs had similar physical prop-
borne from an understanding of the structural and ancestral erties to that of WT VLPs (Supplementary Figs. 8, 9), the chimeric
relationships among multiple HPV types, and, thereafter, the VLPs provide a major increment in HPV protection by inducing
selection of closely related types for epitope grafting. efficient neutralizing antibodies against a broad spectrum of HPV,
Our data show that different loops have variable effects on with five chimeric VLPs able to induce immunity against a total of
transferring cross-protection between different groups (Figs 4, 5); 15 HPV types in mice presented in our study. Even though evo-
this could be primarily caused by their inconsistent antigenic lutionary changes tend to be slow in papillomavirus43,44, recom-
determinants for type specificity. Still, additional structural bination events between different types of viruses may lead to
information of the chimeric VLPs needs to be considered to vaccine evasion; for example, as mentioned earlier, recombination
rationalize how to select particular loops for chimeras with high events between HPV16 and HPV31 can result in a virus carrying
cross-protective efficacy. We also noticed that, among the five HPV16 oncogenes but coding for HPV31 structural proteins or a
surface loops, there is no EF loop-exchange constructs for any of completely new serotype21. Therefore, a cross-type HPV vaccine
the lead triple-type candidates among the five groups (Group-1, with a much broader spectrum against similar types could become
-2, 3, -5 and -6) (Fig. 5), even though an increased heterotypic an alternative means to provide potential protection against the
neutralization could be observed in EF loop-swapped constructs; effects of mutation events and/or infection of parental types not
i.e., the anti-HPV58 antibody titer of H33-58EF was ~2- to 3-log included in current vaccines. We propose that, with necessary
higher than that found following immunization with WT HPV33 modifications, an improved HPV vaccine could be engineered to

a HPV33 HPV58 HPV52 b c 7.6

HPV33 L1 kDa
5 H58-33BC H58-33BC-52HI
55 7.4
4 SDS
PAGE
7.2
–log (KD)
3 40
OD450
7.0
2
55
WB 6.8
1 2
40 1
0 mAb 4E5 4B3 0
–1 0 1 2 3 4
33
58
52
I
2H
3B
PV
PV
PV
mAb 4E5 concentration (ng ml–1, log10)
-5
-3
H
C
58
3B
H
-3
58
H
d HPV33:4E5 H58-33BC:4E5 H58-33BC-52HI:4E5
230 280 320
e 68.0
φ
68.0
φ
68.0
φ
46.0 46.0 46.0
86.0 86.0 86.0
46.0 46.0 46.0
θ θ θ
θ θ θ
58.0 58.0 58.0

68.0 58.0 68.0 58.0 68.0 58.0
φ 86.0 φ 86.0 φ 86.0
BC DE EF FG HI
Fig. 8 Structural characterization of the chimeric pentamers and VLPs-Fab immune complexes. a Binding profiles of mAb 4E5 to chimeric mutant and WT
VLPs of HPV33, HPV52 and HPV58. Data indicates mAb 4E5 can bind to WT VLPs of HPV33 and chimeric VLPs of H58-33BC and H58-33BC-52HI but
fails to bind to WT VLPs of HPV52 and HPV58. The error bars are standard deviation of experiments performed in duplicate. b HPV33 L1 protein was
subjected to reducing SDS-PAGE and western blot (WB) analysis with mAb 4E5. mAb 4B3, a linear antibody capable of recognizing HPV L1 protein, was
employed as a positive control for the WB experiment. c Affinity constants of the chimeric mutants (H58-33BC and H58-33BC-52HI) and WT VLPs
(HPV33, HPV52 and HPV58) against mAb 4E5 were measured by surface plasma resonance (SPR) in a Biacore 3000. The error bars are standard
deviation of experiments performed in triplicate. d CryoEM structures of the VLPs-Fab complexes of HPV33:4E5 (left), H58-33BC:4E5 (middle) and H58-
33BC-52HI:4E5 (right). Density maps are colored by radius from cyan (230 Å), to yellow (280 Å), to pink (320 Å). Icosahedral 2-, 3- and 5-fold axes are
indicated by black symbols. For the purpose of comparability, the cryoEM maps at higher resolution were low-pass filtered to 12.3 Å. e Close-up view of the
roadmaps of HPV33:4E5 (left), HPV58-33BC:4E5 (middle) and HPV58-33BC-52HI (right) surfaces, projected by an area surrounding a 5-fold vertex
defined with a polar angle θ range of 46° to 58° and a polar angle ϕ range of 68°–86°. Color scheme for surface residues are according to surface loops
(BC, amber; DE, orange; EF, yellow; FG, green and HI, Cyan). Blue contour lines were drawn according to the projection of the Fab-4E5 entity with a density
greater than 1 and within a radial section of the VLP-4E5 complex ranging from 310 Å to 360 Å. The icosahedral symmetry axes are labeled. The roadmaps
were generated by the software RIVEM68
elicit protective immunity against 20 types of HPV through seven hypothetically cover more than 99% of HPV infection that cause
chimeric VLPs, along with carcinogenic HPV types from group 4, cervical cancers45.
and other low-risk types (e.g., HPV6 and HPV11, which are not Increasing the immunity of conserved neutralizing epitopes
associated with cervical cancer, but are linked with precancerous or offers a logical strategy for the design of broadly protective
dysplastic lesions and genital warts) (Fig. 1a). This would immunogens; this would be based on a molecular understanding

of the interactions between isolated broadly neutralizing anti- solutions supplemented with 30% glycerol, and flash-cooled at 100 K.Diffraction
bodies and protective antigens4. However, no immunogen to date data from the HPV33, HPV52, HPV58-33BC, HPV58-33BC-52HI L1 pentamer
crystals were collected at the Shanghai Synchrotron Radiation Facility beamline
has successfully elicited antibodies with high potent cross-type BL17U using a Quantum-315r CCD Area Detector. All datasets were indexed,
neutralizing ability, probably due to the limited information integrated, scaled and merged using the HKL-2000 program package (http://www.
available about the conserved regions of natural antigens or the hkl-xray.com). The structures were solved by molecular replacement with PHA-
poor immunogenicity of the conserved areas, since such regions SER47. The crystal structure of HPV58 pentamer (PDB 5Y9E [http://dx.doi.org/
appear to be generally inaccessible in the pathogen’s native state. 10.2210/pdb5Y9E/pdb]) served as the search model. The interest models were built
manually in Coot, refined by PHENIX48 and analyzed by MolProbity49 (96.3%
Here, we provide a rational approach for structure-based cross- favored, 0.3% outliers for HPV33; 96.0% favored, 0.2% outliers for HPV52; 97.0%
type vaccine design: (1) Structure similarity (where the structure favored, 0.2% outliers for H58-33BC; 95.4% favored, 0.2% outliers for H58-33BC-
is available) and evolutionary distance analyses were performed to 52HI). Briefly, one round of rigid-body refinement was performed after molecular
define type groups that could be incorporated into one chimera; replacement. Afterwards, the resulting models were manually corrected and refined
by COOT50. Coordinates and individual B factors of HPV52 and HPV58-33BC
(2) Chimeric loop-substituted mutants created between two most pentamers were refined in reciprocal space without non-crystallographic symmetry
closely related types were initially generated to determine can- (NCS) restraints, whereas the coordinates and group B factors of HPV33 and
didate(s) that would elicit an ideal double-type immunization HPV58-33BC-52HI pentamers were refined in reciprocal space with NCS restraints
response; constructs where swapping occurred between more and secondary structure restraints to avoid overfitting. Data collection and struc-
ture refinement statistics are given in Table 1.
structurally similar loops were prioritized, as only limited con-
structs could be generated and assayed for immunogenicity; (3)
The remaining four loops of the double-type candidate(s) were Sequence alignment and phylogenic tree construction. Multiple sequence
alignment of the L1 proteins from 20 different HPV types was conducted using
then independently exchanged to the corresponding loop of a Clustal Omega51 and the phylogenic tree was constructed by the neighbor-joining
third genetically close type, and immunogenicity was evaluated to method using software MEGA 552 (http://www.megasoftware.net/). The tree was
achieve the optimal triple-type chimeric VLPs; again, constructs drawn to scale, with branch lengths measured in the number of substitutions per
where swapping had occurred on more structurally similar loops site. The five additional types of HPV L1 protein sequences downloaded from
GenBank (accession numbers: HPV6, AAC80442 [https://www.ncbi.nlm.nih.gov/
were prioritized. Although HPV sequence diversity is not as high protein/AAC80442]; HPV11, P04012 [https://www.ncbi.nlm.nih.gov/protein/
as that of RNA viruses, such as the human immunodeficiency and P04012]; HPV26, NP_041787 [https://www.ncbi.nlm.nih.gov/protein/NP_041787];
influenza viruses, or single-stranded DNA viruses, with extensive HPV51, AJS10540 [https://www.ncbi.nlm.nih.gov/protein/AJS10540]; HPV69,
structural information, the method provided here still could ALJ32828 [https://www.ncbi.nlm.nih.gov/protein/ALJ32828]) along with 15 types
of sequences as mentioned above were used for phylogenetic analysis.
markedly benefit the development of a new generation of vaccines
for a wide variety of pathogens.
SDS-PAGE and western blotting. SDS-PAGE was performed using the Laemmli
method with minor modifications53. Briefly, samples were diluted with Laemmli
Methods sample buffer (0.0625 M Tris-HCl, pH 6.8, 2% (w/v) SDS, 10% (w/v) glycerol,
Protein cloning, expression and purification. N-terminally truncated HPV16 100 mM dithiothreitol, and 0.001% (w/v) bromophenol blue) to a final protein
(GenBank: ANA05496 [https://www.ncbi.nlm.nih.gov/protein/ANA05496]), concentration of 1 mg mL−1 or 0.2 mg ml−1. Samples were heated to 80 °C for 10
HPV18 (AAQ92369 [https://www.ncbi.nlm.nih.gov/protein/AAQ92369]), HPV31 min, loaded into the wells of 10% acrylamide gels, electrophoresed, and stained
(P17388, [https://www.ncbi.nlm.nih.gov/protein/P17388]), HPV33 (AMY16565, with Coomassie Brilliant Blue.
[https://www.ncbi.nlm.nih.gov/protein/AMY16565]), HPV35 (P27232, [https:// For western blotting, resolved proteins were electrically transferred to
www.ncbi.nlm.nih.gov/protein/P27232]), HPV39 (P24838, [https://www.ncbi.nlm. nitrocellulose membranes for 55 min at 35 mA. Membranes were blocked with 5%
nih.gov/protein/P24838]), HPV45 (P36741, [https://www.ncbi.nlm.nih.gov/ skim milk, incubated with anti-HPV L1 linear mAb 4B3 (recognizing a broad-
protein/P36741]), HPV52 (AML80965, [https://www.ncbi.nlm.nih.gov/protein/ spectrum linear epitope, 1:1000 dilution), washed, and then incubated with goat
AML80965]), HPV53 (NP_041848, [https://www.ncbi.nlm.nih.gov/protein/ anti-mouse alkaline phosphatase-conjugated antibody (Dako, Denmark). Color
NP_041848]), HPV56 (P36743, [https://www.ncbi.nlm.nih.gov/protein/P36743]), was developed over 5 min using NBT/BCIP reagent (Pierce Biotechnology;
HPV58 (AFS33402, [https://www.ncbi.nlm.nih.gov/protein/AFS33402]), HPV59 Rockford, IL).
(CAA54856, [https://www.ncbi.nlm.nih.gov/protein/CAA54856]), HPV66
(ABO76893, [https://www.ncbi.nlm.nih.gov/protein/ABO76893]), HPV68
Indirect ELISA. The wells of 96-well microplates were coated with wild type (WT)
(AGU90787, [https://www.ncbi.nlm.nih.gov/protein/AGU90787]) and HPV70
or chimeric VLPs (300 ng per well) and then blocked overnight at 4 °C. The wells
(P50793, [https://www.ncbi.nlm.nih.gov/protein/P50793]) L1 genes were cloned
were then incubated with two-fold serial dilutions of the antibody for 45 min at
into pTO-T7 vector. Mutations were constructed using Gibson assembly46, a
37 °C. Antibody titers were detected using a horseradish peroxidase (HRP)-con-
potent reconstitution polymerase chain reaction (PCR) strategy. Briefly, two gene
jugated goat anti-mouse IgG antibody (diluted 1:5000 in HS-PBS, Abcam; Cam-
fragments from their templates were cloned by specific primers (Refer to the pri-
bridge, UK), followed by 50 µL of 3, 3′, 5, 5′-Tetramethylbenzidine Liquid Substrate
mer list in Supplementary Data 1). Gibson assembly was then used to construct the
(Sigma-Aldrich, St Louis, MO) per well for 15 min at 37 °C. The reaction was
targeted clone with these two fragments as templates. All recombinant HPV L1
stopped by the addition of 50 μl of 2 M H2SO4, and the absorbance read at 450 nm
proteins were produced in ER2566 E. coli strain as described previously12. Briefly,
(reference, 620 nm) using an automated ELISA reader (TECAN, Männedorf,
the E. coli strain containing L1 expression vectors was induced with 0.4 mM iso-
Switzerland). The median effective concentration (EC50, ng mL−1) is defined as the
propyl-β-D-thiogalactopyranoside for 10 h at 25 °C. The harvested cells were re-
antibody concentration for achieving 50% binding with the antigen. Data analysis
suspended in buffer (50 mM Tris-HCl, pH 7.2, 10 mM EDTA, 300 mM NaCl) and
was performed using GraphPad Prism (GraphPad Software, San Diego, CA).
disrupted by sonication. The supernatant was purified by two chromatography
procedures using SP sepharose (GE Healthcare, America) and CHT II resin (Bio-
Rad, America). For VLP preparation, the purified L1 proteins were dialyzed with Analytical ultracentrifugation (AUC). The homogeneity and molecular mass of
the above-mentioned buffer without reductant to allow VLP self-assembly; To HPV L1 proteins were determined using sedimentation velocity (SV) experiments.
prepare pentamers for crystallization, the Cys175 (as numbered according to the SV was performed at 20 °C on a Beckman XL-A analytical ultracentrifuge,
sequence of HPV16 L1) in L1 proteins was replaced with Ser and the mutated equipped with absorbance optics and an An60-Ti rotor (Beckman Coulter; Full-
proteins were then subjected to limited trypsin digestion (mass ratio of trypsin vs. erton, CA). All the samples were diluted to an OD280nm of 1 in a 1.2-cm light path.
L1 protein was approximately 1:1000) at room temperature overnight and purified A total of 150 scans for samples were recorded and the rotor speed was set to 7,000
by Superdex 200. rpm for VLPs, and 30,000 r.p.m. for pentamers. The sedimentation coefficient was
obtained using the c(s) method with the Sedfit software54, kindly provided by Dr.
P. Schuck at the National Institutes of Health (Bethesda, MD).
Crystallization and structural determinations. Purified HPV L1 pentamers of
concentration at ~5 mg mL−1 were performed for crystallization trial in sitting-
drop vapor diffusion method in 96-well plates at 20 °C. Best crystals were grown by High-performance size-exclusion chromatography (HPSEC). Purified WT or
mixing 1 µL sample with 1 µL reservoir solution using the hanging-drop vapor- chimeric HPV L1 VLPs were chromatographed using a 1120 Compact LC HPLC
diffusion method in 24-well plates (Reservoir solutions: for HPV33 pentamer: 0.2 system (Agilent Technologies; Santa Clara, CA) separately, equipped with a TSK
M magnesium formate, 13.5% (w/v) PEG 3350; for HPV52 pentamer: 8% tasimate, Gel PW5000xl 7.8 × 300 mm column (TOSOH, Japan); columns were pre-
17.5% (w/v) PEG 3350; for HPV58-33BC pentamer: 0.2 M potassium thiocyanate, equilibrated in phosphate buffer, pH 6.5 with 500 mM NaCl. The column flow rate
23% (w/v) PEG 3350; for HPV58-33BC-52HI pentamer: 0.2 M calcium acetate, and protein signal detection for the SEC analysis were 0.5 ml/min and 280 nm,
18% (w/v) PEG 3350). Crystals were cryo-protected in their respective reservoir respectively.

Differential scanning calorimetry (DSC). DSC was performed on HPV L1 VLPs To verify cross protection of H33-59HI and H58-59BC chimeric VLPs, WT
using a MicroCal VP-DSC instrument (GE Healthcare, MicroCal Products Group; VLPs and chimeric H33-58HI and H58-33BC VLPs (100 μg per dose), absorbed
Northampton, MA). Samples (0.5 mg mL−1) were measured at a heating rate of 1.5 with Freund’s adjuvant, were used to immunize SPF BALB/c mice (n = 3)
°C min−1 using a scanning temperature that ranged from 10 °C to 90 °C. Software subcutaneously three times at an interval of 2 weeks (weeks 0, 2 and 4). Serum
MicroCal Origin 7.0 (Origin-Lab Corp., Northampton, MA) was used to calculate samples were taken at week 8 after the first immunization to determine the
the melting temperatures (Tm) based on the melting curves. neutralizing titers by PBNA.
For dose-dependent response immunogenicity evaluations, SPF BALB/c mice
were immunized intraperitoneally three times with three different doses (0.1, 1 and
Monoclonal antibodies (mAbs). BALB/c mice were immunized subcutaneously 10 μg, n = 5 or 6 per dose) at an interval of 2 weeks (weeks 0, 2 and 4) with
three times at 2-week intervals with HPV33 VLPs or HPV58 VLPs (100 μg per chimeric or WT VLPs diluted in aluminum adjuvant. Serum samples were
mice) absorbed with aluminum adjuvant for prime immunization. Mice also collected at week 8 for PBNA.
received two additional boost inoculations. Fusion was performed 2 weeks after the For ED50 calculations, 40 SPF BALB/c mice (n = 8 per group) were vaccinated
final immunization. The resulting hybridomas were screened using indirect HPV33 by intraperitoneal injection with a single dose (0.3, 0.1, 0.033, 0.011, or 0.004 μg) of
or HPV58 VLP-based ELISA and pseudovirus-based neutralization assays (PBNA, each chimeric or WT VLP (serially diluted in aluminum adjuvant). The dose of
see below). Positive cells were cloned by limiting dilution at least three times until a each VLP in the mixed WT VLP group was equal to that of the same VLP in the
single cell clone was attained. MAbs were prepared by injecting hybridoma cells single-type groups. Serum samples were taken 4 weeks after immunization, and the
into the peritoneal cavities of pristane-primed BALB/c mice; ascitic fluid was seroconversion at each dose was calculated. The ED50 was reported in units of
collected after 9–12 days, and the mAbs were purified by Protein A affinity micrograms of antigen per mouse, resulting in 50% seroconversion. The ED50
chromatography. The purified anti-HPV33 mAbs (4E5, 7C12, 11A3, 11F4, C2C3, results were analyzed based on the dose-response curve using a Reed–Muench
C14A2, C16D7, 9A3 and 7A3) and anti-HPV58 mAbs (4A2, 13A4, 5H2, 10B11, model60.
5G9, A15A9, A10B8, 1D4 and 3E4) were diluted to 1.0 mg ml−1 in PBS and stored To assess the immunogenicity of H58-33BC-52HI VLPs in non-human
at −20 °C. mAb concentration was determined with a BCA method. primates, 20 cynomolgus macaques (Macaca Fascicularis)—with no detectable
neutralizing antibodies against infection of HPV33, -52 or -58—were divided into
BIAcore biosensor analysis. CM-5 sensor chips were amine-coupled to a goat five groups (n = 4). The first group was inoculated with H58-33BC-52HI VLP (5 μg
anti-mouse antibody Fc fragment (GAM-Fc) (BIAcore 3000, GE). One flow cell of per dose); the other four groups were inoculated with HPV33 (5 μg per dose),
a chip was coated with 15,000 RU (resonance units) of the GAM-Fc, whereas the HPV52 (5 μg per dose), HPV58 (5 μg per dose) VLPs, and the mixed VLPs of
other flow cell was left uncoated and blocked as a control. The affinity measure- HPV33/52/58 (5 μg of HPV33, -52 and -58 per dose), respectively. All five groups
ments of mAb 4E5 binding with HPV33, HPV58, HPV52, H58-33BC and H58- were immunized twice (months 0 and 2) by Intramuscular injection. Pre-immune
33BC-52HI pentamer, respectively, were initiated by passing HBS (10 mM HEPES, serum samples of each monkey were collected individually, and serum samples
pH 7.4 and 150 mM NaCl) over the sensor surface for 100 s at 30 μl min−1, fol- were collected at 2-week intervals after immunization. The titers of HPV
lowed by injection of 1 μg ml−1 of mAb 4E5 at 30 μl min−1 for 3.3 min, and then neutralizing antibodies in these sera were detected by PBNA. All data for the
injections of serially diluted antigens at 30 μL min−1 for 3.3 min. Every measure- neutralization titers of per group analyzed by PBNA are plotted as the mean
ment on the BIAcore 3000 biosensor was performed three times and the individual with SD.
values were used to produce the mean affinity constant and standard deviation. The sample size of animal studies were estimated according to our previous
studies61. All the mice used in this study were female and about 4–6 weeks old and
the age of mice were much the same in the same set experiments. The male and
Preparation of HPV pseudoviruses. The L1/L2 expression vector and pN31- female cynomolgus macaques were about 3 years old. All the animals were divided
EGFP used in the experiment were kindly provided by Dr. J. T. Schiller55. Briefly, randomly into a certain number groups. No blinding was done.
plasmids carrying codon-optimized HPV L1 genes were individually co-transfected
with an L2 expression plasmid and the marker plasmid into 293FT cells, which
were purchased from Thermo Fisher (Catalogue number: R70007) and had been
authenticated by Short Tandem Repeat (STR) assay and single nucleotide poly- Cryo-EM and three-dimensional (3D) reconstruction. Purified mAb 4E5 Fab
morphism (SNP) genotyping, and tested for the absence of mycoplasma con- fragments were incubated with HPV58, HPV58-33BC, or HPV58-33BC-52HI
tamination. The cells were harvested 72 h after transfection, lysed with cell lysis VLPs at 37 °C for 2 h (molar ratios of 1:1.2 for all immune complexes). Small
buffer containing 0.5% Brij58 (Sigma-Aldrich), 0.2% benzonase (Merck Millipore; aliquots (3 μL) of mixtures were deposited onto glow discharged holey carbon
Darmstadt, Germany), 0.2% PlasmidSafe ATP-Dependent DNase (Epicenter Bio- Quantifoil Cu grids (R2/2, 200 mesh, Quantifoil Micro Tools), blotted with filter
technologies, Madison, WI) and DPBS-Mg solution, and incubated at 37 °C for 24 paper for 6 s, and then vitrified by plunging into liquid ethane inside an FEI Mark
h. Afterward, 5 M NaCl solution was added to the samples to extract the cell IV Vitrobot at 100% humidity. CryoEM micrographs were collected at 300 kV with
lysates. The TCID50 (tissue culture infective dose) of the supernatant was then a FEI Tecnai F30 transmission electron microscope, equipped with Falcon II direct
measured to determine the titers of the PsVs, calculated according to the classical electron detector at a defocus level from 1.5 to 3.5 µm evaluated using Gctf62. The
Reed–Muench method56. Maturation and purification of these samples followed nominal magnification was set to 93,000 (pixel size = 1.128Å) with an electron
the procedures described before57–59. dose of 25 e−/Å 2 at 1 s. Particles were manually picked using the e2boxer.py
program in EMAN2.163. After several rounds of reference-free 2D classifications
using Relion2.064, good particles were sorted for further 3D reconstruction and
Pseudovirion-based neutralization assay (PBNA). 293FT cells were incubated at refinement with AUTO3DEM65. Final map resolutions were determined by the
37 °C in the wells of a 96-well plate at a density of 1.5 × 104 cells per well for 6 h. gold standard FSC curve between two-half maps with a cutoff of 0.366. HPV58
Sera were 2-fold diluted and PsVs were diluted to 2 × 105 TCID50/µL. Equal pentamer crystal structure were manually fitted using the program Chimera, and
volumes (60 µl) of the PsV diluent and the serially diluted sera were mixed and further optimized into the corresponding maps using the “fit in map” command in
incubated at 4 °C for 1 h. The negative control was prepared by mixing 60 µL of the Chimera67.
PsV diluent with an equal volume of culture medium. Then, 100 µL of each
mixture was added to designated wells and incubated at 37 °C for 72 h. The neu-
tralization titers were calculated as the log10 of the highest sera dilution with a
percentile of infection inhibition higher than 50%. Every sample was repeated at Stereographic roadmap projections of HPV VLP-Fab complexes. The crystal
least three times, and the mean values and SD are reported. structures of HPV33, HPV58-33BC, HPV58-33BC-52HI were fitted into the cor-
responding cryoEM density map using the tool “fit in map” in Chimera, and were
Ethics statement. Animal experiments in mice were approved by Xiamen Uni- then plotted onto stereographic spheres using RIVEM68 (Radial Interpretation of
versity Laboratory Animal Center (XMULAC). The immunization and serum Viral Electron Density Maps), where the polar angles θ and ϕ represent latitude
harvest of cynomolgus macaques were conducted at WuXi AppTec Co., Ltd. and longitude, respectively. The Fab density was projected as contour lines onto the
(Suzhou, P. R. China) (Contract no. XMU-20160525A). The manipulation protocol sphere, which is presented as a stereographic diagram.
was approved by WuXi Institutional Animal Care and Use Committee (IACUC).
All procedures were conducted in accordance with animal ethics guidelines and
approved protocols to minimize suffering during vaccination, blood collection, and
euthanasia.
Data availability:
The data supporting the findings of this study are available from the corresponding
authors upon reasonable request. Atomic coordinates and structure factors for the
Animals, immunizations and serological analysis. To initially assess the HPV33, HPV52, H58-33BC and H58-33BC-52HI L1 pentamers have been
immunogenicity of all chimeric VLPs, special pathogen-free (SPF) BALB/c mice deposited in the Protein Data Bank (accession code 6IGE, 6IGF, 6IGD, and 6IGC).
(n = 4 or 5) were immunized intraperitoneally three times at an interval of 2 weeks The cryoEM density maps for HPV33:4E5, H58-33BC:4E5 and H58-33BC-
(weeks 0, 2 and 4) with chimeric or WT VLPs diluted in aluminum adjuvant (5 μg 52HI:4E5 have been deposited in the Electron Microscopy Data Bank (accession
per dose). Serum samples were collected at week 8, and the neutralizing titers were code EMD-9665, EMD-9666 and EMD-9667). A reporting summary for this
analyzed by PBNA, as described above. Article is available as a Supplementary Information file.

Received: 22 December 2017 Accepted: 18 October 2018 27. Pastrana, D. V. et al. Reactivity of human sera in a sensitive, high-throughput
pseudovirus-based papillomavirus neutralization assay for HPV16 and
HPV18. Virology 321, 205–216 (2004).
28. Guan, J. et al. Structural comparison of four different antibodies interacting
with human papillomavirus 16 and mechanisms of neutralization. Virology
483, 253–263 (2015).
References 29. Lee, H. et al. A CryoEM study identifies the complete H16.V5 epitope and
1. World Health Organization. Pneumococcal vaccines WHO position paper-- reveals global conformational changes initiated by binding of the neutralizing
2012. Releve Epidemiol. Hebdomadaire 87, 129–144 (2012). antibody fragment. J. Virol. 89, 1428–1438 (2014).
2. Schiller, J. T. & Muller, M. Next generation prophylactic human 30. Li, Z. et al. Crystal structures of two immune complexes identify determinants
papillomavirus vaccines. Lancet Oncol. 16, e217–e225 (2015). for viral infectivity and type-specific neutralization of human papillomavirus.
3. Kirchenbaum, G. A. & Ross, T. M. Eliciting broadly protective antibody mBio 8, e00787 (2017).
responses against influenza. Curr. Opin. Immunol. 28, 71–76 (2014). 31. Bishop, B. et al. Crystal structures of four types of human papillomavirus L1
4. Burton, D. R., Poignard, P., Stanfield, R. L. & Wilson, I. A. Broadly capsid proteins: understanding the specificity of neutralizing monoclonal
neutralizing antibodies present new prospects to counter highly antigenically antibodies. J. Biol. Chem. 282, 31803–31811 (2007).
diverse viruses. Science 337, 183–186 (2012). 32. Chen, X. S., Garcea, R. L., Goldberg, I., Casini, G. & Harrison, S. C. Structure
5. Bommakanti, G. et al. Design of an HA2-based Escherichia coli expressed of small virus-like particles assembled from the L1 protein of human
influenza immunogen that protects mice from pathogenic challenge. Proc. papillomavirus 16. Mol. Cell 5, 557–567 (2000).
Natl Acad. Sci. USA 107, 13701–13706 (2010). 33. Carter, J. J. et al. Identification of human papillomavirus type 16 L1 surface
6. Steel, J. et al. Influenza virus vaccine based on the conserved hemagglutinin loops required for neutralization by human sera. J. Virol. 80, 4664–4672
stalk domain. mBio 1, e0018 (2010). (2006).
7. zur Hausen, H. Papillomaviruses in the causation of human cancers - a brief 34. Christensen, N. D. et al. Hybrid papillomavirus L1 molecules assemble
historical account. Virology 384, 260–265 (2009). into virus-like particles that reconstitute conformational epitopes and
8. Ho, G. Y., Bierman, R., Beardsley, L., Chang, C. J. & Burk, R. D. Natural induce neutralizing antibodies to distinct HPV types. Virology 291, 324–334
history of cervicovaginal papillomavirus infection in young women. N. Engl. J. (2001).
Med. 338, 423–428 (1998). 35. Wang, X., Wang, Z., Christensen, N. D. & Dillner, J. Mapping of human
9. Munoz, N. et al. Epidemiologic classification of human papillomavirus types serum-reactive epitopes in virus-like particles of human papillomavirus types
associated with cervical cancer. N. Engl. J. Med. 348, 518–527 (2003). 16 and 11. Virology 311, 213–221 (2003).
10. de Villiers, E. M. Cross-roads in the classification of papillomaviruses. 36. Jagu, S., Kwak, K., Garcea, R. L. & Roden, R. B. Vaccination with multimeric
Virology 445, 2–10 (2013). L2 fusion protein and L1 VLP or capsomeres to broaden protection against
11. Baker, T. S. et al. Structures of bovine and human papillomaviruses. Analysis HPV infection. Vaccine 28, 4478–4486 (2010).
by cryoelectron microscopy three-dimensional Image reconstruction.Biophys. 37. Christensen, N. D., Kreider, J. W., Cladel, N. M., Patrick, S. D. & Welsh, P. A.
J. 60, 1445–1456 (1991). Monoclonal antibody-mediated neutralization of infectious human
12. Li, Z. et al. The C-Terminal arm of the human papillomavirus major capsid papillomavirus type 11. J. Virol. 64, 5678–5681 (1990).
protein is immunogenic and involved in virus-host interaction. Structure 24, 38. Guan, J. et al. The U4 antibody epitope on human papillomavirus 16 identified
874–885 (2016). by cryo-EM. J. Virol. 89, 12108–12117 (2015).
13. Breitburd, F. et al. Immunization with viruslike particles from cottontail rabbit 39. Brown, D. R. et al. The impact of quadrivalent human papillomavirus (HPV;
papillomavirus (CRPV) can protect against experimental CRPV infection. J. types 6, 11, 16, and 18) L1 virus-like particle vaccine on infection and disease
Virol. 69, 3959–3963 (1995). due to oncogenic nonvaccine HPV types in generally HPV-naive women aged
14. Roden, R. B. et al. In vitro generation and type-specific neutralization of a 16-26 years. J. Infect. Dis. 199, 926–935 (2009).
human papillomavirus type 16 virion pseudotype. J. Virol. 70, 5875–5883 40. Paavonen, J. et al. Efficacy of human papillomavirus (HPV)-16/18 AS04-
(1996). adjuvanted vaccine against cervical infection and precancer caused by
15. Roden, R. B. et al. Assessment of the serological relatedness of genital human oncogenic HPV types (PATRICIA): final analysis of a double-blind,
papillomaviruses by hemagglutination inhibition. J. Virol. 70, 3298–3301 randomised study in young women. Lancet 374, 301–314 (2009).
(1996). 41. Bissett, S. L. et al. Pre-clinical immunogenicity of human papillomavirus
16. Kemp, T. J. et al. HPV16/18 L1 VLP vaccine induces cross-neutralizing alpha-7 and alpha-9 major capsid proteins. Vaccine 32, 6548–6555 (2014).
antibodies that may mediate cross-protection. Vaccine 29, 2011–2014 (2011). 42. Giersing, B. K. et al. Potency assay design for adjuvanted recombinant
17. Ochi, H. et al. Neutralizing antibodies against human papillomavirus types 16, proteins as malaria vaccines. Vaccine 24, 4264–4270 (2006).
18, 31, 52, and 58 in serum samples from women in Japan with low-grade 43. Ong, C. K. et al. Evolution of human papillomavirus type 18: an ancient
cervical intraepithelial neoplasia. Clin. Vaccin. Immunol.: CVI 15, 1536–1540 phylogenetic root in Africa and intratype diversity reflect coevolution with
(2008). human ethnic groups. J. Virol. 67, 6424–6431 (1993).
18. Smith, J. F. et al. Antibodies from women immunized with Gardasil cross- 44. Shah, S. D., Doorbar, J. & Goldstein, R. A. Analysis of host-parasite
neutralize HPV 45 pseudovirions. Hum. Vaccin. 3, 109–115 (2007). incongruence in papillomavirus evolution using importance sampling. Mol.
19. Deschuyteneer, M. et al. Molecular and structural characterization of the L1 Biol. Evol. 27, 1301–1314 (2010).
virus-like particles that are used as vaccine antigens in Cervarix, the AS04- 45. Schiffman, M. et al. Human papillomavirus testing in the prevention of
adjuvanted HPV-16 and -18 cervical cancer vaccine. Hum. Vaccin. 6, 407–419 cervical cancer. J. Natl Cancer Inst. 103, 368–383 (2011).
(2010). 46. Gibson, D. G., Smith, H. O., Hutchison, C. A. 3rd, Venter, J. C. & Merryman,
20. McCormack, P. L. Quadrivalent human papillomavirus (types 6, 11, 16, 18) C. Chemical synthesis of the mouse mitochondrial genome. Nat. Methods 7,
recombinant vaccine (gardasil((R))): a review of its use in the prevention of 901–903 (2010).
premalignant anogenital lesions, cervical and anal cancers, and genital warts. 47. McCoy, A. J., Grosse-Kunstleve, R. W., Storoni, L. C. & Read, R. J. Likelihood-
Drugs 74, 1253–1283 (2010). enhanced fast translation functions. Acta Crystallogr. D. Biol. Crystallogr. 61,
21. Van Doorslaer, K. Evolution of the papillomaviridae. Virology 445, 11–20 458–464 (2005).
(2013). 48. Adams, P. D. et al. PHENIX: a comprehensive Python-based system for
22. Fleury, M. J., Touze, A. & Coursaget, P. Human papillomavirus type 16 macromolecular structure solution. Acta Crystallogr. D. Biol. Crystallogr. 66,
pseudovirions with few point mutations in L1 major capsid protein FG loop 213–221 (2010).
could escape actual or future vaccination for potential use in gene therapy. 49. Chen, V. B. et al. MolProbity: all-atom structure validation for
Mol. Biotechnol. 56, 479–486 (2014). macromolecular crystallography. Acta Crystallogr. D. Biol. Crystallogr. 66,
23. Carter, J. J., Wipf, G. C., Benki, S. F., Christensen, N. D. & Galloway, D. A. 12–21 (2010).
Identification of a human papillomavirus type 16-specific epitope on the C- 50. Emsley, P. & Cowtan, K. Coot: model-building tools for molecular graphics.
terminal arm of the major capsid protein L1. J. Virol. 77, 11625–11632 (2003). Acta Crystallogr. D. Biol. Crystallogr. 60, 2126–2132 (2004).
24. Christensen, N. D. et al. v5-Surface conformational and linear epitopes on 51. Sievers, F. et al. Fast, scalable generation of high-quality protein multiple
HPV-16 and HPV-18 L1 virus-like particles as defined by monoclonal sequence alignments using Clustal Omega. Mol. Syst. Biol. 7, 539 (2011).
antibodies. Virology 223, 174–184 (1996). 52. Tamura, K. et al. MEGA5: molecular evolutionary genetics analysis using
25. Ludmerer, S. W., Benincasa, D. & Mark, G. E. 3rd Two amino acid residues maximum likelihood, evolutionary distance, and maximum parsimony
confer type specificity to a neutralizing, conformationally dependent epitope methods. Mol. Biol. Evol. 28, 2731–2739 (2011).
on human papillomavirus type 11. J. Virol. 70, 4791–4794 (1996). 53. Laemmli, U. K., Beguin, F. & Gujer-Kellenberger, G. A factor preventing the
26. Roden, R. B. et al. Characterization of a human papillomavirus type 16 major head protein of bacteriophage T4 from random aggregation. J. Mol.
variant-dependent neutralizing epitope. J. Virol. 71, 6247–6252 (1997). Biol. 47, 69–85 (1970).

54. Schuck, P. Size-distribution analysis of macromolecules by sedimentation 81701637), the National Science and Technology Major Projects for Major New Drugs
velocity ultracentrifugation and lamm equation modeling. Biophys. J. 78, Innovation and Development (No. 2018ZX09738008). The funders had no role in the
1606–1619 (2000). study design, data collection and analysis, decision to publish, or preparation of the
55. Buck, C. B., Pastrana, D. V., Lowy, D. R. & Schiller, J. T. Generation of HPV manuscript.
pseudovirions using transfection and their use in neutralization assays.
Methods Mol. Med. 119, 445–462 (2005). Author contributions
56. Chen, Y. et al. Antigenic analysis of divergent genotypes human Enterovirus S.L., Y.G. and N.X. designed the study. Z.L., S.S., M.H., D.W., J.S., X.L., Y.L., X.C., S.W.,
71 viruses by a panel of neutralizing monoclonal antibodies: current Y.Y., Z.W., J.L., H.Q. and Q.Z. performed the experiments. Z.L., S.S., M.H., D.W., H.Y.,
genotyping of EV71 does not reflect their antigenicity. Vaccine 31, 425–430 X.Y., Y.G., S.L. and N.X. analyzed the data. Z.L. and S.L. wrote the manuscript. Z.L., S.S.,
(2013). M.H., D.W., H.Y., X.Y., Q.Z., J.Z.,Y.G., S.L. and N.X. participated in discussion and
57. Buck, C. B., Pastrana, D. V., Lowy, D. R. & Schiller, J. T. Efficient intracellular interpretation of the results. All authors contributed to experimental design.
assembly of papillomaviral vectors. J. Virol. 78, 751–757 (2004).
58. Buck, C. B. & Thompson, C. D. Production of papillomavirus-based gene
transfer vectors. Curr. Protoc. Cell Biol. Chapter 26, Unit 26 (2007). 21. Additional information
59. Cardone, G. et al. Maturation of the human papillomavirus 16 capsid. mBio 5, Supplementary Information accompanies this paper at https://doi.org/10.1038/s41467-
e01104–e01114 (2014). 018-07199-6.
60. Reed, L. J. & Muench, H. A simple method of estimating fifty per cent
endpoints 1 2. Am. J. Epidemiol. 27, 493–497 (1938). Competing interests: The authors declare no competing interests.
61. Gu, Y. et al. Characterization of an Escherichia coli-derived human
papillomavirus type 16 and 18 bivalent vaccine. Vaccine 35, 4637–4645 Reprints and permission information is available online at http://npg.nature.com/
(2017). reprintsandpermissions/
62. Zhang, K. Gctf: Real-time CTF determination and correction. J. Struct. Biol.
193, 1–12 (2016).
63. Ludtke, S. J. Single-particle refinement and variability analysis in EMAN2.1.
Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in
Methods Enzymol. 579, 159–189 (2016).
published maps and institutional affiliations.
64. Scheres, S. H. RELION: implementation of a Bayesian approach to cryo-EM
structure determination. J. Struct. Biol. 180, 519–530 (2012).
65. Yan, X., Sinkovits, R. S. & Baker, T. S. AUTO3DEM--an automated and high
throughput program for image reconstruction of icosahedral particles. J.
Open Access This article is licensed under a Creative Commons
Struct. Biol. 157, 73–82 (2007).
Attribution 4.0 International License, which permits use, sharing,
66. Rosenthal, P. B. & Henderson, R. Optimal determination of particle
adaptation, distribution and reproduction in any medium or format, as long as you give
orientation, absolute hand, and contrast loss in single-particle electron
appropriate credit to the original author(s) and the source, provide a link to the Creative
cryomicroscopy. J. Mol. Biol. 333, 721–745 (2003).
67. Pettersen, E. F. et al. UCSF Chimera--a visualization system for exploratory Commons license, and indicate if changes were made. The images or other third party
research and analysis. J. Comput. Chem. 25, 1605–1612 (2004). material in this article are included in the article’s Creative Commons license, unless
68. Xiao, C. & Rossmann, M. G. Interpretation of electron density with indicated otherwise in a credit line to the material. If material is not included in the
stereographic roadmap projections. J. Struct. Biol. 158, 182–187 (2007). article’s Creative Commons license and your intended use is not permitted by statutory
regulation or exceeds the permitted use, you will need to obtain permission directly from
the copyright holder. To view a copy of this license, visit http://creativecommons.org/
Acknowledgements licenses/by/4.0/.
We acknowledge use of beamline 17U of the Shanghai Synchrotron Radiation Facility
(SRFF) for X-ray diffraction data collection. This work was supported by grants from
the National natural Science Foundation of China (No. U1705283, 31670935 and © The Author(s) 2018

Rational Design of A Triple-Type Human Papillomavirus Vaccine by Compromising Viral-Type Speci Ficity

Uploaded by

Copyright:

Available Formats

Rational Design of A Triple-Type Human Papillomavirus Vaccine by Compromising Viral-Type Speci Ficity

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Rational Design of A Triple-Type Human Papillomavirus Vaccine by Compromising Viral-Type Speci Ficity

Uploaded by

Copyright:

Available Formats

ARTICLE

DOI: 10.1038/s41467-018-07199-6 OPEN

Rational design of a triple-type human

Sequence variability in surface-antigenic sites of pathogenic proteins is an important obstacle

NATURE COMMUNICATIONS | (2018)9:5360 | DOI: 10.1038/s41467-018-07199-6 | www.nature.com/naturecommunications 1

2 NATURE COMMUNICATIONS | (2018)9:5360 | DOI: 10.1038/s41467-018-07199-6 | www.nature.com/naturecommunications

Table 1 Data collection and reﬁnement statistics

HPV33 HPV52 H58-33BC H58-33BC-52HI

NATURE COMMUNICATIONS | (2018)9:5360 | DOI: 10.1038/s41467-018-07199-6 | www.nature.com/naturecommunications 3

1 121 150 263 290 499

149 178 289 316

4 NATURE COMMUNICATIONS | (2018)9:5360 | DOI: 10.1038/s41467-018-07199-6 | www.nature.com/naturecommunications

NATURE COMMUNICATIONS | (2018)9:5360 | DOI: 10.1038/s41467-018-07199-6 | www.nature.com/naturecommunications 5

Neturalization titer (log 10)

*** **** **** **** *** ***

to assess their cross-type immunogenicity. Of the eight triple-type

6 NATURE COMMUNICATIONS | (2018)9:5360 | DOI: 10.1038/s41467-018-07199-6 | www.nature.com/naturecommunications

NATURE COMMUNICATIONS | (2018)9:5360 | DOI: 10.1038/s41467-018-07199-6 | www.nature.com/naturecommunications 7

a PsV33 PsV58 PsV52

0.1 μg per mouse 1 μg per mouse 10 μg per mouse

6 **** 6 *** 6 *** 6 **

8 NATURE COMMUNICATIONS | (2018)9:5360 | DOI: 10.1038/s41467-018-07199-6 | www.nature.com/naturecommunications

a PsV33 PsV52 PsV58

PsV33 PsV52 PsV58

NATURE COMMUNICATIONS | (2018)9:5360 | DOI: 10.1038/s41467-018-07199-6 | www.nature.com/naturecommunications 9

N84H58 K85 K380H52

10 NATURE COMMUNICATIONS | (2018)9:5360 | DOI: 10.1038/s41467-018-07199-6 | www.nature.com/naturecommunications

a HPV33 HPV58 HPV52 b c 7.6

230 280 320

58.0 58.0 58.0

NATURE COMMUNICATIONS | (2018)9:5360 | DOI: 10.1038/s41467-018-07199-6 | www.nature.com/naturecommunications 11

12 NATURE COMMUNICATIONS | (2018)9:5360 | DOI: 10.1038/s41467-018-07199-6 | www.nature.com/naturecommunications

NATURE COMMUNICATIONS | (2018)9:5360 | DOI: 10.1038/s41467-018-07199-6 | www.nature.com/naturecommunications 13

14 NATURE COMMUNICATIONS | (2018)9:5360 | DOI: 10.1038/s41467-018-07199-6 | www.nature.com/naturecommunications

NATURE COMMUNICATIONS | (2018)9:5360 | DOI: 10.1038/s41467-018-07199-6 | www.nature.com/naturecommunications 15

You might also like

* * ***

6 ** 6 * 6 * 6