You are on page 1of 6

Production and stabilization of the trimeric influenza

hemagglutinin stem domain for potentially broadly


protective influenza vaccines
Yuan Lua, John P. Welsha, and James R. Swartza,b,1
Departments of aChemical Engineering and bBioengineering, Stanford University, Stanford, CA 94305

Edited by Peter Palese, Icahn School of Medicine at Mount Sinai, New York, NY, and approved November 15, 2013 (received for review May 7, 2013)

The rapid dissemination of the 2009 pandemic H1N1 influenza Because the HA head domain evolves with a high mutation rate,
virus emphasizes the need for universal influenza vaccines that the influenza vaccine has to be regularly updated. The stem
would broadly protect against multiple mutated strains. Recent domain, on the other hand, is much more conserved (SI Ap-
efforts have focused on the highly conserved hemagglutinin (HA) pendix, Fig. S1) (6, 7). At low pH (5∼6), as the influenza virus is
stem domain, which must undergo a significant conformational trafficking in endosomal vesicles, the trimeric HA stem domain
change for effective viral infection. Although the production of undergoes a conformational change to trigger fusion to the
isolated domains of multimeric ectodomain proteins has proven endosomal membrane, thereby enabling RNA release and suc-
difficult, we report a method to rapidly produce the properly cessful infection (8). Several research groups have devised con-
folded HA stem domain protein from influenza virus A/California/ structs consisting of portions of this stem domain in hopes of
05/2009 (H1N1) by using Escherichia coli-based cell-free protein producing a broadly protective influenza vaccine (1, 9, 10).
synthesis and a simple refolding protocol. The T4 bacteriophage However, none have produced properly folded stem trimer.
fibritin foldon placed at the C terminus of the HA stem domain Recent work suggests that antibodies generated against this stem
induces trimer formation. Placing emphasis on newly exposed pro- domain can cross-react between different virus subtypes, lending
tein surfaces, several hydrophobic residues were mutated, two hope toward using this domain as an antigen for broadly pro-
polypeptide segments were deleted, and the number of disulfide tective influenza vaccines (10–13). A vaccine, multivalent against
bonds in each monomer was reduced from four to two. High pH several stem subtypes, would not need annual updates and could
and Brij 35 detergent emerged as the most beneficial factors for dramatically reduce the risk of influenza pandemics.
improving the refolding yield. To stabilize the trimer of the HA The natural production mechanisms suggest that HA stem
stem-foldon fusion, new intermolecular disulfide bonds were fi- production would be technically challenging. The first challenge
nally introduced between foldon monomers and between stem is that the HA stem domain has not evolved to fold and trimerize
domain monomers. The correct immunogenic conformation of as an independent unit. Further, its contemporaneous folding
the stabilized HA stem domain trimer was confirmed by using along with that of the head domain most likely occurs cotrans-
antibodies CR6261, C179, and FI6 that block influenza infection lationally as first part of the stem, then the head, and finally the
by binding to the HA stem domain trimer. These results suggest rest of the stem domain are extruded from the ER membrane
great promise for a broadly protective vaccine and also demon- and are orientated by this association. During this process,
strate a unique approach for producing individual domains of com- disulfide bonds must form within each monomer while avoiding
plex multimeric proteins. intermonomer linkage. Finally, the absence of the head domain
exposes internal HA polypeptides that may now be disordered,

V accination is the most effective way to protect against in-


fluenza virus infection, with most of the neutralizing anti-
bodies recognizing the hemagglutinin (HA) protein on the
hydrophobic, or otherwise inappropriate as surface epitopes for

Significance
surface of the virus. As of 2007, all commercial influenza vac-
cines were produced in embryonated chicken eggs, but the The discovery of neutralizing antibodies that block influenza

APPLIED BIOLOGICAL
generation of a new vaccine takes six to eight months, making it infection by binding to the hemagglutinin (HA) stem domain
ineffective for combating potential pandemics (1). In 2007, to raised the hope for broadly protective vaccines. These could

SCIENCES
partially address this concern, the European Union approved avoid the need for annual vaccinations and reduce pandemic
Optaflu, a vaccine produced by Novartis using a mammalian cell threats, and the stem subdomain of the trimeric HA ectodo-
line (2). In 2013, the recombinant HA vaccine (Flublok) manu- main would be an ideal antigen. However, its production has
factured in insect cells by Protein Sciences was also licensed in proven extremely difficult. Here, we describe a simple pro-
the United States (3). However, the new processes would still cedure resulting in high yields of HA stem trimer recognized
likely only improve production times by several weeks (4). identically by a panel of neutralizing antibodies as compared
Hence, rapid production (for example in 1–2 mo) of large with recognition of the full-length HA ectodomain. Cell-free
amounts of vaccine to stop an epidemic/pandemic remains an protein synthesis is followed by a simple refolding procedure
important objective. This challenge can be addressed in two to produce a rationally mutated stem in which newly exposed
ways, either by producing a more broadly protective vaccine to protein surfaces are modified and trimerization is induced and
lower the probability of a pandemic occurrance or by developing covalently stabilized.
new, rapid, and scalable technologies for vaccine production.
This work seeks to address both. Author contributions: Y.L., J.P.W., and J.R.S. designed research; Y.L. and J.P.W. performed
The trimeric HA ectodomains consist of a head domain and research; Y.L. contributed new reagents/analytic tools; Y.L., J.P.W., and J.R.S. analyzed
a stem domain (Fig. 1A). The assembled HA ectodomain also data; and Y.L., J.P.W., and J.R.S. wrote the paper.
consists of HA1 and HA2 domains, which are defined by a spe- The authors declare no conflict of interest.
cific proteolytic cleavage (5). The majority of protective anti- This article is a PNAS Direct Submission.
bodies bind to the HA globular head domain. Consequently, new 1
To whom correspondence should be addressed. E-mail: jswartz@stanford.edu.
viral strains arise with the most common mutations occurring in This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.
the head domain, thereby avoiding initial antibody suppression. 1073/pnas.1308701110/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1308701110 PNAS | January 7, 2014 | vol. 111 | no. 1 | 125–130


Fig. 1. (A) Structural features of the influenza virus HA trimer and its head and stem domains. HA1 regions are shown in blue, and HA2 regions are in cyan.
The figure was derived from the X-ray structure of HA from the (H1N1) isolate A/California/05/2009 (H1N1) (PDB ID code 3LZG) and was drawn by using the
program PyMOL. Protein yields (B) and autoradiograms (C) of nonreducing SDS/PAGE gels of CFPS products. The arrow indicates the monomer. Stem: no
trimerization domain, 32 kDa; Stem-foldon: with the foldon trimerization domain, 36 kDa; Stem-CAT: with the CAT trimerization domain, 58 kDa; S, soluble;
T, total protein. Autoradiogram from nonreducing SDS/PAGE gels of refolded proteins (D) and pI analysis of HA protein fragments (E). Polypeptides were
refolded with different pH environments (pH 6.0, 8.0, or 10.5) and with (+) or without (−) Tween 20 (0.05%). The theoretical pI values were calculated by
using the program ProtParam. All of the residue numbers were in H3 numbering.

stable and soluble protein. Because of these complications, in the head domain (1, 9). HA is classified into 17 subtypes (H1–
vivo Escherichia coli expression produced low yields of only 2 mg/L, H17) based on antigenicity and sequence diversity (17), and, in
and the HA stem polypeptides accumulated as inclusion bod- particular, the stem domains in the same subtype are highly
ies (1). Although some refolding methods have been attempted conserved (SI Appendix, Fig. S1). Two subtypes of influenza A,
(14–16), the recovery yields of soluble products have been H1N1 and H3N2, most commonly infect humans. Therefore, for
low, and properly folded stable trimeric assembly has not been example, a vaccine presenting these two stem domains could
confirmed. serve as a broadly protective vaccine.
In this study, we demonstrated the production of HA stem In historical records of previous influenza pandemics, H1N1
domain from the 2009 pandemic H1N1 strain by using E. coli- influenzas occurred most often and were severe, so the HA stem
based cell-free protein synthesis (CFPS) production methods domain from the influenza virus A/California/05/2009 (H1N1)
followed by protein refolding. For inducing the trimerization of was chosen as the first target. It is highly likely that the HA stem
the HA stem domain, either a CAT or foldon domain was fused domains from other subtypes could also be produced by using
to the C terminus. To reduce the chance of the formation of the same protocol described in this work. Because HA is active
intermolecular disulfide bonds, several HA stem domain muta- as a trimer on the viral surface, two different sequences were
tions were introduced, including removal of hydrophobic and fused to the C terminus of the stem domain to induce trimeri-
positively charged residues and decreasing the number of disul- zation. The first was a 29-aa “foldon” sequence that forms a β-
fide bonds. Protein refolding methods were also optimized. The propeller structure comprising the C terminus of the fibritin
trimer was further stabilized by introducing disulfide bonds be- domain of the T4 bacteriophage (18). This domain has been used
tween foldon domains and in the distal region of the stem trimer. as a trimerization domain for HA (12, 19). The second candidate
The correct immunogenic conformation of this stabilized HA trimerization domain was chloramphenicol acetyl transferase
stem trimer was confirmed by using antibodies CR6261, C179, (CAT), which expresses and folds well in typical CFPS reactions
and FI6 that block influenza infection by binding to the HA stem and also naturally forms a trimer (20). Both domains were linked
trimer at a site that inhibits the structural changes required for to the HA sequence by a cleavable tobacco etch virus (TEV)
membrane fusion. Production consistency and high recovery protease site and also contained a His6 tag at the C terminus.
yields were also demonstrated. The E. coli-based CFPS system was used for this study because
it has many advantages for rapidly screening protein constructs
Results and Discussion compared with traditional E. coli expression systems (21). First,
Production and Refolding of HA Stem Domain Trimer. The stem direct use of plasmids or PCR templates as expression tem-
domain of the HA protein is considerably more conserved than plates avoids time-consuming molecular cloning steps. Second,

126 | www.pnas.org/cgi/doi/10.1073/pnas.1308701110 Lu et al.


milligrams of protein can be produced in a few hours in multiwell Mutations of Hydrophobic Regions. Based on these considerations,
plates and then directly purified without a cell-lysis step. Third, five sets of mutations were designed (Fig. 2A) to either mitigate
the lack of a cell wall barrier allows for easy manipulation of newly exposed hydrophobicity (SI Appendix, Fig. S2), reduce the
reaction conditions, including the incorporation of isotopically potential for intermolecular ion pairing, or both. Five different
labeled amino acids. The E. coli-based CFPS system is linearly groups of mutations were evaluated, M1 (I69T + I72E + I74T +
scalable from micro-scale (20 μL) to large-scale reactions (100 L) C77T), M2 (I69T + I72E + I74T + C77T + F164D), M3 (I69T +
(22, 23) with nearly identical process performance. The system I72E + I74T + C77T + F164D + L174D), M4 (F164D), and M5
has also been shown to produce active pharmaceutical proteins (F164D + L174D). The mutations decreased the pIs of two
(22–24). newly exposed stem domain fragments (Stem-HA1-Fragment1
Initial microscale (20 μL) CFPS reactions tested expression of and Stem-HA2-Fragment1) (SI Appendix, Table S1). It should be
the HA stem construct either alone or with the trimerization noted that the targeted mutations are distant from the surface
domains. These reactions were conducted at 30 °C for 6 h and recognized by the neutralizing antibody (12). However, the CFPS
produced high total yields but much lower soluble yields (Fig. soluble yields of these mutants were still low (SI Appendix, Fig.
1B). Nonreducing SDS/PAGE with autoradiography (Fig. 1C) S3). The insoluble inclusion bodies were then refolded by using
indicated no significant accumulation of soluble monomer, and the dialysis refolding protocol at pH 10.5 (SI Appendix, SI
the higher molecular weight species indicated the formation of Materials and Methods), and the products were evaluated us-
incorrect intermolecular disulfide bonds. ing size-exclusion HPLC (Fig. 2B). Mutants M3 and M5 produced
Refolding of the inclusion bodies was then attempted, using much fewer aggregates than the wild-type or other variants. The
a dialysis refolding method (25–27) under different pH con- most influential mutations appeared to be F164D + L174D.
ditions. Eight molar urea was used to solubilize the inclusion Therefore, mutant M5 (F164D + L174D) was used for further
bodies and the HA stem protein was purified by using the His6 development.
tag. Oxidized (GSSG) and reduced glutathione (GSH) (molar
ratio 1:4) were added to establish a sulfhydryl/disulfide redox Improving Protein Refolding at pH 8.0. Although the high pH (10.5)
environment for the formation and potential isomerization of folding conditions and the M5 mutation increased the soluble
correct disulfide bonds. L-arginine (0.5 M) and the detergent, fraction after refolding, when the pH was then lowered to 8.0,
Tween 20 (0.05% wt/vol), were also added to assist in refolding. the product still became insoluble. We therefore focused on
Three different pH values (6.0, 8.0, and 10.5) were tested (Fig. adjusting components in the dialysis buffer to improve folding at
1D). At pH 6.0, most of the protein was lost, apparently adhering pH 8.0, thereby avoiding the need for subsequent pH change.
to the dialysis membrane. At pH 8.0, most reaggregated, whereas Many factors can affect protein refolding, such as pH, chaotropic
at pH 10.5, less aggregation occurred. However, the soluble agent choice, -S-H/S-S redox ratio, ionic strength, buffer choice,
fraction from the pH 10.5 procedure also aggregated when the and low molecular weight additives such as arginine and deter-
pH was reduced to 8.0. These observations prompted us to ex- gents. Detergent Brij 35 was more effective than Tween 20 as
amine regional isoelectric point (pI) values for the polypeptide judged by SDS/PAGE and size-exclusion HPLC analysis (SI
calculated by using the program ProtParam (28) (Fig. 1E). Appendix, Fig. S4). Because the HA stem domain protein con-
The theoretical pI values of Stem-HA1-Fragment1 (pI 8.85), tains four disulfide bonds, the choice of sulfhydryl/disulfide redox
Stem-HA1-Fragment2 (pI 9.10), and Stem-HA2-Fragment1 buffer was next examined. The cystamine/cysteamine system was
(pI 8.60) were above 8.5, whereas the pI values of other stem more effective than GSSG/GSH (SI Appendix, Fig. S5). Cyst-
domain fragments were less than 5. The divergent pIs suggested amine and cysteamine are smaller molecules than GSSG and
the potential for significant interdomain ionic attractions at GSH and possibly can more readily diffuse into the protein
neutral pH that would largely be avoided at pH 10.5. Fur- structure to assist in disulfide formation and interchange.
thermore, because the detergent, Tween 20, reduced aggre- To identify the best refolding conditions while evaluating
gation, inappropriate hydrophobic interactions were also seven different factors, an orthogonal (design of experiment)
suggested. Finally, significant aggregation resulted at all three procedure was designed with the seven factors evaluated at two
pHs with the CAT trimerization fusion and this folding partner levels (SI Appendix, Table S2). The seven factors are buffer
was abandoned. choice (50 mM Tris or 50 mM histidine), [Brij 35] (0% or 0.03%),

APPLIED BIOLOGICAL
SCIENCES

Fig. 2. (A) HA stem constructs with different mutations in the exposed hydrophobic region. (B) Size-exclusion HPLC profiles of refolded proteins (wild-type,
mutants M1, M2, M3, M4, and M5). Dotted red traces indicated 14C radioactivity after 14C leucine incorporation during CFPS. (C) Three-dimensional structures
of mutants M5 and M6. (D) Autoradiogram of nonreducing SDS/PAGE gel of refolded HA stem domain mutants M5 and M6.

Lu et al. PNAS | January 7, 2014 | vol. 111 | no. 1 | 127


Fig. 3. Introduction of disulfide bonds between foldon monomers. (A) Three-dimensional structures of foldon trimer with six possible disulfide bond
positions. Autoradiogram of a nonreducing SDS/PAGE gel (B) and size-exclusion HPLC profiles (C) of refolded HA stem construct mutant M6 (O) and constructs
with the six different foldon mutations.

[arginine] (0 mM or 500 mM), [NaCl] (0 mM or 150 mM), [su- To further confirm the formation of intermolecular S-S bonds
crose] (0% or 10%), [glycerol] (0% or 10%) and [urea] (0 M or between foldon monomers, we used the TEV protease cleavage
0.5 M). The data were analyzed by using IBM SPSS Statistics site between the HA stem domain and foldon domain to release
software. The results showed that Brij 35 was the most important the foldon trimer after the refolding procedure. Nonreducing
factor for improving the refolding yield (SI Appendix, Table S3). SDS/PAGE (SI Appendix, Fig. S7) of the released foldon
Arginine (500 mM) was also marginally beneficial for suppres- domains also showed that version E enabled the best formation
sion of protein aggregation in the refolding process and was in- of disulfide-linked trimers. This mutant was used in subsequent
cluded in the refolding buffer. development.
Furthermore, two different refolding protocols, dialysis refold-
ing and dilution refolding, were compared. These two protocols Further Polypeptide and Folding Protocol Adjustments. To increase
produced similar refolding results (SI Appendix, Fig. S6). The di- the proportion of trimer and the completeness of intermolecular
lution refolding protocol is much faster and more convenient and disulfide bond formation within the foldon trimer, the refolding
was used in the subsequent experiments. Although the preceding conditions were further examined. Using cystamine and cyste-
changes improved protein refolding results, SDS/PAGE analysis amine as the “oxido-shuffling” reagent, we compared several
indicated that intermolecular disulfide bonds were still impeding ratios, 1:4, 1:10, and 1:20. During such procedures, oxygen-
proper folding. driven oxidation can modify the sulfhydryl/disulfide redox po-
tential over time, and a higher fraction of trimer was obtained
Targeted Deletions for Further Improvement. To avoid the forma- anaerobically. We therefore conducted refolding procedures in
tion of undesired intermolecular S-S bonds, further reduce sur- an anaerobic glove box (SI Appendix, Fig. S8), and the cystamine
face hydrophobicity and pI, and avoid regions with possibly to cysteamine 1:20 ratio was best (SI Appendix, Fig. S9). The final
disordered structure, mutant M5 was further modified by de- composition of the modified refolding buffer was as follows: 50
leting two polypeptide regions (H38 to C43 and C49 to N61) mM Tris·HCl, 600 mM arginine, 2 mM EDTA, 0.5 mM cyst-
containing hydrophobic residues, two positively charged lysines, amine/10 mM cysteamine, and 0.05% Brij35 at pH 8.
and three cysteines (Fig. 2C). Cysteine 77 was also mutated to Although a higher proportion of trimer was obtained after
threonine to remove a cysteine that was then unpaired. Again, improving refolding conditions with the M6E mutant, we still
the deleted regions are far away from the neutralizing antibody observed dimer and monomer bands. In the initial design of the
binding region. The number of disulfide bonds in each monomer HA stem domain antigen, the foldon domain is separated from
was thereby decreased from 4 to 2. This mutant was named M6. the HA stem sequence by a cleavable TEV protease site and the
SDS/PAGE results (Fig. 2D) showed that these modifications fusion protein also has a C-terminal His6 tag. We hypothesized
greatly decreased the formation of undesired intermolecular that the TEV protease site and the direct linkage to the His6 tag
disulfide bonds. could have impeded the trimer formation or formation of the
intermolecular S-S bonds in the foldon trimer. To structurally
Stabilization of the Stem-Foldon Trimer. The HA stem trimer may isolate the foldon domain, stem trimer mutant M6 with foldon
be used as an independent vaccine antigen or could possibly be mutation E (M6E) was modified by replacing the TEV protease
conjugated to virus-like particles (VLPs) or other carriers. In ei-
ther case, it will be important for the trimeric form to be stable
during manufacturing, storage, and administration. Reliable sta-
bility is particularly important during the Cu(I) catalyzed click
conjugation reaction. Although the foldon domain proved effec-
tive in forming the trimers, we reasoned that the limited number
of intermolecular hydrogen bonding attractions in the foldon may
not sufficiently stabilize this unnatural ectodomain trimer.
To stabilize the HA stem trimer, we introduced new cysteine
residues into the foldon domain to potentially form intermolecular
disulfide bonds between foldon monomers. Beginning with mu-
tant M6, we introduced cysteines at the six following locations to
evaluate which pairs could most effectively form intermolecular
disulfide bonds (A: G12-D19; B: R10-D19; C: R10-G20; D: P9-
R20; E: A14-K18; and F: Y15-R17), as shown in Fig. 3A. SDS/ Fig. 4. Newly designed HA stem construct (M6EL) based on mutant M6 with
PAGE results (Fig. 3B) showed that mutant E formed the most foldon mutation E (M6E). (A) Schematic diagram of new HA stem construct
disulfide-linked trimer. Size-exclusion HPLC results were con- design. (B) Autoradiogram of nonreducing SDS/PAGE gel of refolded M6EL
sistent with this interpretation (Fig. 3C). HA stem trimer.

128 | www.pnas.org/cgi/doi/10.1073/pnas.1308701110 Lu et al.


Fig. 5. Introduction of intermolecular disulfide bonds between stem monomers in the stem trimer. (A) The position of newly introduced intermolecular S-S
bonds between stem monomers. (B) Protein yields and overall recovery yields of stem construct mutants (M6L and M6EL with the S-S mutations SS1, SS2, SS3,
or SS4) after protein expression, purification, and refolding. (C) Autoradiogram of a nonreducing SDS/PAGE gel of refolded HA stem construct mutants. (D)
Size-exclusion HPLC profiles of refolded HA stem construct mutant M6EL and M6EL(SS2).

target sequence with a (GS)3 linker, and another (GS)3G linker Conclusion
was added between the foldon domain and the His6 tag, as shown In this work, we describe an E. coli-based CFPS method for
in Fig. 4A. This HA stem design was named “M6EL.” Non- producing an HA stem domain trimer assembly derived from the
reducing SDS/PAGE results (Fig. 4B) showed that the addition 2009 pandemic H1N1 strain. The CFPS production of this antigen
of the (GS)n linkers increased the proportion of trimer to ap- potentially provides a rapid, scalable, and cost-effective means to
proximately 80%. Repeated production batches indicated ex- combat pandemic and epidemic threats when a rapid response
cellent reproducibility (SI Appendix, Table S4). and large quantities of vaccine are required. Although the HA
stem antigen by itself may not be an effective vaccine, it can now
Additional Stabilization of the Stem Trimer. Although the HA stem
be tested with adjuvants or by attaching it to virus-like particles to
trimer had been stabilized with disulfide bonds between foldon
enhance its immunogenicity and increase protective efficacy.
monomers (Fig. 3) at the membrane proximal end of the mole-
We also developed an effective refolding method to produce
cule, we reasoned that stability could be further improved by also
the correctly folded trimeric HA stem domain protein. We suggest
introducing disulfide bonds between stem monomers at the distal
this stepwise procedure may provide a valuable paradigm for pro-
end of the protein structure. Cysteines were introduced in the
Stem-HA2-Fragment3 (Fig. 1E) at the following locations (SS1: ducing properly folded multimeric domains of membrane-associated
I192C-W193C; SS2: G188C-F189C; SS3: K184C-V185C; SS4: cell and viral surface proteins. To reduce the aggregation and the
L181C-N182C), as shown in Fig. 5A. Cysteine adjacency was formation of incorrect intermolecular disulfide bonds, several
chosen to discourage intramolecular bonds in favor of bonds newly exposed hydrophobic residues were replaced, the isoelectric
between monomers. Sites were also chosen based on natural points of newly exposed protein fragments were decreased, and the
intermolecular proximity. Disulfide bond formation at these sites number of cysteines in the HA stem domain was reduced from eight
could both encourage and help confirm proper trimer formation. to four. The HA stem trimer was further stabilized by introducing
SDS/PAGE results (Fig. 5C) showed that all four designs formed intermolecular disulfide bonds between foldon monomers and
intermolecular disulfide bonds between stem monomers in the between stem domain monomers. The refolded and stabilized tri-
absence of foldon mutation E. The SS2 mutant of M6EL showed meric HA stem domain protein was recognized identically com-
the highest protein yield (363 μg/mL CFPS reaction volume; pared with full-length HA ectodomain by neutralizing antibodies
27% higher than M6EL) after protein expression, purification, CR6261, C179, and FI6, suggesting that the neutralizable epitope is
and refolding (Fig. 5B). The mutant M6EL(SS2) was used in
subsequent studies. Size-exclusion HPLC results (Fig. 5D) fur-

APPLIED BIOLOGICAL
ther confirmed better trimer formation for M6EL(SS2).

SCIENCES
Confirming Neutralizing Antigen in Stabilized HA Stem Trimer. The
correct immunogenic conformation of this stabilized HA stem
trimer M6EL(SS2) was confirmed by using antibodies CR8020,
CR6261, C179, and FI6 known to be specific for neutralizing
epitopes located in the HA stem domain. An ELISA was used
with green fluorescence protein (GFP) as the negative control.
The full-length H1 HA monomer [A/California/07/2009(H1N1)]
and the full-length H1 HA trimer [A/South Carolina/1/1918
(H1N1)] (with foldon trimerization domain) were used as posi-
tive controls. ELISA binding results (Fig. 6) showed that the new
H1 HA stem trimer and the full-length H1 HA trimer control
were recognized with nearly identical affinity by all four anti-
bodies. As expected, recognition of the full-length HA monomer
differed significantly from that of the trimers. The FI6 antibody
bound HA monomer more strongly than the trimer, whereas
Fig. 6. ELISA of HA stem construct M6EL(SS2) using antibody CR8020 (A),
CR6261 and C179 bound the trimer more strongly. CR6261,
CR6261 (B), C179 (C), and FI6 (D). Green fluorescence protein (GFP) protein
C179, and FI6 all bind HAs belonging to the H1 subtype, but with a C-terminal His6 tag was used as the negative control. The full-length
CR8020 only can bind HAs belonging to H3 and H7 subtypes H1 HA monomer [A/California/07/2009(H1N1)] (with His6 tag) and the full-
(29). Together these results demonstrate that the M6EL(SS2) length H1 HA trimer [A/South Carolina/1/1918(H1N1)] (with foldon trimeri-
HA stem antigen is properly folded. zation domain and His6 tag) were used as the positive controls.

Lu et al. PNAS | January 7, 2014 | vol. 111 | no. 1 | 129


properly folded. This influenza HA stem trimer therefore is an cation (or EDTA), and mild detergents (Tween 20, Brij 35),
excellent candidate for the design and production of broadly pro- using a design of experiment approach.
tective influenza vaccines. Based on the insights gained, we believe
that similar procedures can be rapidly developed for the other in-
Materials and Methods
fluenza HA stem domains needed for universal vaccines.
Design and Construction of HA Stem Domain Construct. The HA stem domain
As mentioned, this overall protocol may provide a paradigm for
protein was from the influenza virus A/California/05/2009 (H1N1) hemag-
the rational design and evaluation of stable properly folded
glutinin (accession no. ACP41926). The sequences of all HA stem constructs
domains of other multimeric proteins. The protocol begins with
are shown in SI Appendix, SI Materials and Methods.
cell-free protein expression and then involves four main steps. Of
course, the development of any successful folding protocol ben- CFPS. CFPS was conducted by using the PANOx-SP (phosphoenolpyruvate,
efits when the starting material is relatively pure, and a rapid and amino acids, nicotinamide adenine dinucleotide, oxalic acid, spermidine, and
precise assessment method for protein folding is essential. putrescine) cell-free system as described (30, 31) (SI Appendix).
i) Analyze the pI of protein fragments particularly on newly
exposed protein surfaces. Significant pI differences between Protein Purification and Refolding. After CFPS reactions, the denatured pro-
fragments could result in significant intermolecular ionic teins were purified by a Ni-NTA column (Qiagen). Two different refolding
methods were used and compared: dialysis refolding and dilution refolding
attractions and protein aggregation at neutral pH, which
(SI Appendix, SI Materials and Methods).
could inhibit folding. Mutations to reduce or alter charges
on new surfaces may be beneficial.
Size Exclusion HPLC. HA stem domain proteins (mutants M1, M2, M3, M4, and
ii) Analyze the hydrophobicity at the newly exposed protein surfa-
M5) refolded at pH 10.5 were initially analyzed by using a Discovery Bio GFC
ces. Newly exposed hydrophobic regions could result in unde- 500 HPLC column (Sigma-Aldrich). Unless specifically stated otherwise, other
sired protein aggregation in the protein folding process. Mutating refolded HA stem domain proteins were tested by using an Ultrahydrogel
these exposed hydrophobic residues to hydrophilic residues 500 HPLC column (Waters). See SI Appendix for details.
could simultaneously reduce surface hydrophobicity and the pI
differences of polypeptide domains. If possible, targeted muta- ELISA Binding of HA Stem Constructs. Antibodies CR8020, CR6261, C179, or FI6
tions should be distant from the biologically relevant regions were coated on 96-well ELISA plates (NUNC MaxiSorp) for ELISA tests. Full-
of protein. length HA monomer [A/California/07/2009(H1N1) and full-length HA trimer
iii) Introduce intermolecular covalent disulfide bonds between (A/South Carolina/1/1918(H1N1)] were used as the positive controls. GFP
monomers to stabilize the multimeric protein. The disulfide protein was used as the negative control. See SI Appendix for details.
bonds can form where the distance between the two residues
is expected to be less than 8.5 Å. Again, if possible, these ACKNOWLEDGMENTS. We thank Prof. Mark M. Davis for help in providing
modifications should be introduced at points distant from the the initial funding during the 2009 influenza pandemic and Dr. Jeffrey
Boyington and the National Institutes of Health (NIH) Vaccine Research
region of the protein with important functional properties. Center for providing antibodies and HA proteins. This work was supported
iv) Optimize the refolding conditions, including pH, disulfide through the NIH-funded Stanford U19 consortium Grants AI057229
exchange reagents, buffer additives (arginine, salts), divalent and AI090019.

1. Bommakanti G, et al. (2010) Design of an HA2-based Escherichia coli expressed in- 16. Biesova Z, et al. (2009) Preparation, characterization, and immunogenicity in mice of
fluenza immunogen that protects mice from pathogenic challenge. Proc Natl Acad Sci a recombinant influenza H5 hemagglutinin vaccine against the avian H5N1 A/Viet-
USA 107(31):13701–13706. nam/1203/2004 influenza virus. Vaccine 27(44):6234–6238.
2. Extance A (2011) Cell-based flu vaccines ready for US prime time. Nat Rev Drug Discov 17. Harris AK, et al. (2013) Structure and accessibility of HA trimers on intact 2009 H1N1
10(4):246–247. pandemic influenza virus to stem region-specific neutralizing antibodies. Proc Natl
3. Yang LPH (2013) Recombinant trivalent influenza vaccine (flublok(Ò)): A review of its Acad Sci USA 110(12):4592–4597.
use in the prevention of seasonal influenza in adults. Drugs 73(12):1357–1366. 18. Tao YZ, Strelkov SV, Mesyanzhinov VV, Rossmann MG (1997) Structure of bacterio-
4. Wright PF (2008) Vaccine preparedness—are we ready for the next influenza pan- phage T4 fibritin: A segmented coiled coil and the role of the C-terminal domain.
demic? N Engl J Med 358(24):2540–2543. Structure 5(6):789–798.
5. Zhirnov OP, Ikizler MR, Wright PF (2002) Cleavage of influenza a virus hemagglutinin 19. Stevens J, et al. (2004) Structure of the uncleaved human H1 hemagglutinin from the
in human respiratory epithelium is cell associated and sensitive to exogenous anti- extinct 1918 influenza virus. Science 303(5665):1866–1870.
proteases. J Virol 76(17):8682–8689. 20. Leslie AGW (1990) Refined crystal structure of type III chloramphenicol acetyl-
6. Chen GL, Subbarao K (2009) Attacking the flu: Neutralizing antibodies may lead to transferase at 1.75 A resolution. J Mol Biol 213(1):167–186.
‘universal’ vaccine. Nat Med 15(11):1251–1252. 21. Carlson ED, Gan R, Hodgman CE, Jewett MC (2012) Cell-free protein synthesis: Ap-
7. Ikonen N, et al. (2010) High frequency of cross-reacting antibodies against 2009 plications come of age. Biotechnol Adv 30(5):1185–1194.
22. Yin G, et al. (2012) Aglycosylated antibodies and antibody fragments produced in
pandemic influenza A(H1N1) virus among the elderly in Finland. Euro Surveill 15(5):
a scalable in vitro transcription-translation system. MAbs 4(2).
16–23.
23. Zawada JF, et al. (2011) Microscale to manufacturing scale-up of cell-free cytokine
8. Gaudin Y, Ruigrok RWH, Brunner J (1995) Low-pH induced conformational changes in
production—a new approach for shortening protein production development time-
viral fusion proteins: Implications for the fusion mechanism. J Gen Virol 76(Pt 7):
lines. Biotechnol Bioeng 108(7):1570–1578.
1541–1556.
24. Goerke AR, Swartz JR (2008) Development of cell-free protein synthesis platforms for
9. Steel J, et al. (2010) Influenza virus vaccine based on the conserved hemagglutinin
disulfide bonded proteins. Biotechnol Bioeng 99(2):351–367.
stalk domain. MBio 1(1):e00018–e10.
25. Rudolph R, Lilie H (1996) In vitro folding of inclusion body proteins. FASEB J 10(1):
10. Wang TT, et al. (2010) Vaccination with a synthetic peptide from the influenza virus
49–56.
hemagglutinin provides protection against distinct viral subtypes. Proc Natl Acad Sci 26. Lilie H, Schwarz E, Rudolph R (1998) Advances in refolding of proteins produced in
USA 107(44):18979–18984. E. coli. Curr Opin Biotechnol 9(5):497–501.
11. Li GM, et al. (2012) Pandemic H1N1 influenza vaccine induces a recall response in 27. Chang JY, Swartz JR (1993) Single-step solubilization and folding of Igf-1 aggregates
humans that favors broadly cross-reactive memory B cells. Proc Natl Acad Sci USA from Escherichia-coli. ACS Symp Ser 526:178–188.
109(23):9047–9052. 28. Wilkins MR, et al. (1999) Protein identification and analysis tools in the ExPASy server.
12. Ekiert DC, et al. (2009) Antibody recognition of a highly conserved influenza virus Methods Mol Biol 112:531–552.
epitope. Science 324(5924):246–251. 29. Corti D, Lanzavecchia A (2013) Broadly neutralizing antiviral antibodies. Annu Rev
13. Wei CJ, et al. (2010) Induction of broadly neutralizing H1N1 influenza antibodies by Immunol 31:705–742.
vaccination. Science 329(5995):1060–1064. 30. Jewett MC, Swartz JR (2004) Mimicking the Escherichia coli cytoplasmic environment
14. Swalley SE, et al. (2004) Full-length influenza hemagglutinin HA(2) refolds into the activates long-lived and efficient cell-free protein synthesis. Biotechnol Bioeng 86(1):
trimeric low-pH-induced conformation. Biochemistry-Us 43(19):5902–5911. 19–26.
15. Curtis-Fisk J, Spencer RM, Weliky DP (2008) Isotopically labeled expression in E. coli, 31. Lu Y, Welsh JP, Chan W, Swartz JR (2013) Escherichia coli-based cell free production
purification, and refolding of the full ectodomain of the influenza virus membrane of flagellin and ordered flagellin display on virus-like particles. Biotechnol Bioeng
fusion protein. Protein Expr Purif 61(2):212–219. 110(8):2073–2085.

130 | www.pnas.org/cgi/doi/10.1073/pnas.1308701110 Lu et al.

You might also like