key: cord-0984482-6os5iy5w
authors: Lingwood, Daniel; McTamney, Patrick M.; Yassine, Hadi M.; Whittle, James R. R.; Guo, Xiaoti; Boyington, Jeffrey C.; Wei, Chih-Jen; Nabel, Gary J.
title: Structural and genetic basis for development of broadly neutralizing influenza antibodies
date: 2012-08-29
journal: Nature
DOI: 10.1038/nature11371
sha: 7a7bfe6bea6e18ea0e4011e4a8407d7628c32d26
doc_id: 984482
cord_uid: 6os5iy5w

Influenza viruses take a yearly toll on human life despite efforts to contain them with seasonal vaccines. These viruses evade human immunity through the evolution of variants that resist neutralization. The identification of antibodies that recognize invariant structures on the influenza haemagglutinin (HA) protein have invigorated efforts to develop universal influenza vaccines. Specifically, antibodies to the highly conserved stem region of HA neutralize diverse viral subtypes. These antibodies largely derive from a specific antibody gene, heavy-chain variable region IGHV1-69, after limited affinity maturation from their germline ancestors(1,2), but how HA stimulates naive B cells to mature and induce protective immunity is unknown. To address this question, we analysed the structural and genetic basis for their engagement and maturation into broadly neutralizing antibodies. Here we show that the germline-encoded precursors of these antibodies act as functional B-cell antigen receptors (BCRs) that initiate subsequent affinity maturation. Neither the germline precursor of a prototypic antibody, CR6261 (ref. 3), nor those of two other natural human IGHV1-69 antibodies, bound HA as soluble immunoglobulin-G (IgG). However, all three IGHV1-69 precursors engaged HA when the antibody was expressed as cell surface IgM. HA triggered BCR-associated tyrosine kinase signalling by germline transmembrane IgM. Recognition and virus neutralization was dependent solely on the heavy chain, and affinity maturation of CR6261 required only seven amino acids in the complementarity-determining region (CDR) H1 and framework region 3 (FR3) to restore full activity. These findings provide insight into the initial events that lead to the generation of broadly neutralizing antibodies to influenza, informing the rational design of vaccines to elicit such antibodies and providing a model relevant to other infectious diseases, including human immunodeficiency virus/AIDS. The data further suggest that selected immunoglobulin genes recognize specific protein structural ‘patterns’ that provide a substrate for further affinity maturation. SUPPLEMENTARY INFORMATION: The online version of this article (doi:10.1038/nature11371) contains supplementary material, which is available to authorized users.

Influenza viruses take a yearly toll on human life despite efforts to contain them with seasonal vaccines. These viruses evade human immunity through the evolution of variants that resist neutralization. The identification of antibodies that recognize invariant structures on the influenza haemagglutinin (HA) protein have invigorated efforts to develop universal influenza vaccines. Specifically, antibodies to the highly conserved stem region of HA neutralize diverse viral subtypes. These antibodies largely derive from a specific antibody gene, heavy-chain variable region IGHV1-69, after limited affinity maturation from their germline ancestors 1,2 , but how HA stimulates naive B cells to mature and induce protective immunity is unknown. To address this question, we analysed the structural and genetic basis for their engagement and maturation into broadly neutralizing antibodies. Here we show that the germline-encoded precursors of these antibodies act as functional B-cell antigen receptors (BCRs) that initiate subsequent affinity maturation. Neither the germline precursor of a prototypic antibody, CR6261 (ref. 3) , nor those of two other natural human IGHV1-69 antibodies, bound HA as soluble immunoglobulin-G (IgG). However, all three IGHV1-69 precursors engaged HA when the antibody was expressed as cell surface IgM. HA triggered BCR-associated tyrosine kinase signalling by germline transmembrane IgM. Recognition and virus neutralization was dependent solely on the heavy chain, and affinity maturation of CR6261 required only seven amino acids in the complementaritydetermining region (CDR) H1 and framework region 3 (FR3) to restore full activity. These findings provide insight into the initial events that lead to the generation of broadly neutralizing antibodies to influenza, informing the rational design of vaccines to elicit such antibodies and providing a model relevant to other infectious diseases, including human immunodeficiency virus/AIDS. The data further suggest that selected immunoglobulin genes recognize specific protein structural 'patterns' that provide a substrate for further affinity maturation.

Antibodies to the conserved stem region of HA block membrane fusion and prevent productive infection by diverse influenza viruses. The structural basis of HA stem recognition of two such monoclonal antibodies, CR6261 and F10, has been defined 3, 4 . These antibodies bind with nanomolar affinity to a highly conserved hydrophobic groove at the interface of HA1 and HA2, the two polypeptides that constitute HA, and neutralize several influenza group 1 subtypes including H1, H5, H6, H8 and H9 (ref. 3 ). Among the anti-stem antibodies isolated so far, the IGHV1-69 gene segment is observed more frequently than expected by chance 2, 5 . To understand the development of these antibodies, we studied the prototypic IGHV1-69-derived broadly neutralizing antibody CR6261, isolated by phage display of human immunoglobulin genes, as well as two others cloned from single human cells, FE53 and 1009-3B05 (refs 1-3). Influenza IGHV1-69-based broadly neutralizing antibodies undergo a relatively low degree of somatic mutation (an average of 14 amino acids in the heavy chain, n 5 9) 1-4 (Fig. 1a) . We first asked whether their germline antibody precursors might recognize HA with measurable affinity. Notably, the IGHV1-69 germline ancestors of CR6261, FE53 and 1009-3B05 failed to bind HA as soluble IgG, even at concentrations as high as 100 mg ml 21 (Fig. 1b) .

To define the molecular basis for affinity maturation of these antibodies, we analysed the respective contributions of heavy and light chains to antigen recognition. We compared chimaeric antibodies that consisted of somatic heavy (sH) and germline light (gL) chains to the mature antibody (sHsL). The chimaeric sHgL of all three antibodies bound to a recombinant H1 HA with affinities similar to their respective matured sHsL (Fig. 1b) . Maturation of the light chain thus does not affect binding to H1 HA. Rather, somatic mutation of the IGHV1-69 heavy chain gene alone mediates the increase in binding affinity. This finding is consistent with the lack of light-chain interaction observed in the crystal structures of both CR6261-HA and F10-HA complexes 3, 4 , suggesting that IGHV1-69 light-chain somatic mutation may be incidental and is not required for heavy-light-chain pairing or improved neutralization function.

We next investigated the minimum requirements of heavy-chain loop maturation that lead to somatic activity. We chose to focus this analysis on CR6261, the first IGHV1-69 anti-stem broadly neutralizing antibody discovered, for which the structure is known. The sH of CR6261 differs from the germline heavy (gH) chain by only 14 amino acids (11.6%) (Fig. 1a) . Four of these amino acids (Pro 28, Arg 30, Lys 58 and Phe 74) contact the HA stem in the crystal structure (Fig. 2a) . We generated several CR6261 germline variants by introducing residues from sH, including CDR H1, CDR H2, CDR H3 or FR3, into gH, singly or in combination. The resulting antibodies were analysed for their ability to bind HA and neutralize virus in an HApseudotyped lentiviral system 6 ( Table 1) . Tested individually, only the somatic CDR H1 (sCDR H1: Thr28Pro/Ser30Arg) increased binding, but it was more than 100-fold less than sH. Furthermore, sCDR H1 alone did not confer detectable activity in our neutralization assay ( Table 1) . Although sFR3 alone did not improve the potency of germline CR6261, sCDR H1 and sFR3 together restored full activity for both binding and neutralization against H1N1 and H5N1 viruses ( Table 1 , Fig. 2b and Supplementary Fig. 1 ). This finding suggests that the mutation of only a small number of germline residues enables potent neutralization.

Because CDR H1 and FR3 sit adjacent to one another in the folded protein, we examined the protein structure further. In most IGHV1-69 antibodies with no CDR H1 mutations, Phe 29 is buried in the 'canonical' conformation of the CDR H1 (ref. 7) . In contrast, the somatically mutated CDR H1 of CR6261 flips the hydrophobic residue Phe 29 out, placing this residue in contact with HA. To determine whether the position of Phe 29 in CR6261 is due to maturation or binding, we solved the crystal structure of unliganded CR6261(sHgL), and compared it to both a IGHV1-69 antibody with no mutations in CDR H1, and HA-liganded CR6261 (Fig. 2c) . In the unliganded structure, Phe 29 is exposed on the surface, suggesting that the somatic mutations Thr28Pro and Ser30Arg lead to its placement there-a hydrophobic residue from the interior of the antibody is made available to contribute to the interface by mutation of adjacent residues. The CDR H1 of IGHV1-69 in both germline and affinity-matured states is not well ordered unless bound to antigen. However, the CDR H1 loop in CR6261 favours the non-canonical conformation, with Phe 29 exposed, whether or not HA is bound. The solved structure suggests that the main consequence of somatic CDR H1 mutation is to favour this non-canonical state. Comparison of these structures also explains the synergistic effect of mutations in CDR H1 and FR3; somatic mutation in FR3 introduces Phe 74 on the surface of CR6261 adjacent to Phe 29 in CDR H1, and these two sets of mutations increase the hydrophobicity of the contact surface more than either alone.

The minimal somatic mutation of the CR6261 germ line required to confer full activity made it surprising that the germline antibody did not recognize HA with any measurable affinity in solution. We proposed that antibody recognition by the low-affinity germline IGHV1-69 revertant requires a more physiological presentation, for example, on the cell surface where such antibodies would normally be expressed. A naturally bivalent transmembrane IgM form of the CR6261 germ line was transfected into human embryonic kidney 293F cells. Cell surface expression was confirmed by staining with an anti-lambdachain antibody. Fluorescent-labelled HA stained CR6261-germlineexpressing cells but not mock transfected cells, as measured by flow cytometry (Fig. 3a) . In contrast, no binding was seen in solution with CR6261 germline Fab monomer, bivalent IgG, or decameric IgM antibody derivatives ( Supplementary Fig. 2 ). The specificity of CR6261 germline binding to the HA stem was further confirmed by its minimum reactivity to a mutant HA probe that blocks anti-stemantibody binding (Supplementary Fig. 3 ). FE53 and 1009-3B05 each showed similar membrane-dependent recognition (Fig. 3a) . Our results suggest that membrane presentation of antibody provides a mechanism by which low-affinity germline B cells achieve sufficient binding to recognize antigens before affinity maturation, and are consistent with experiments showing that two-dimensional confinement and clustering at the plasma membrane can increase the apparent affinity of cell surface receptors 8,9 . Two amino acids at the tip of CDR H2, Ile 53 and Phe 54, seem to be an anchor by which germline IGHV1-69 might attach to HA (Fig. 2a) . Indeed, mutation of these two amino acids to alanines abolished HA germline binding for all VH-1-69 antibodies (Fig. 3a) . Individual mutation of Ile 53 and Phe 54 also abolished binding of unmodified/native HA trimer (binding in the presence of 69-sialyllactose to prevent sialic acid mediated cell adhesion; Fig. 3b and Supplementary Fig. 4 ). In the CR6261-HA co-crystal structure, CDR H2 inserts into a hydrophobic pocket between HA1 and HA2 (ref.

3) (Fig. 3c) . These data suggest that this interaction of the germline CDR H2 has a central role in recognition and engagement of the HA stem.

To determine whether binding of antigen to germline antibodies displayed on the cell surface could induce BCR activation and signalling, the transmembrane IgM version of the CR6261 germline was transfected into a transformed human B cell capable of expressing a functional BCR 10,11 , a Ramos cell clone whose endogenous IgM is not 

expressed (Methods). We found that proteoliposome-arrayed HA ( Supplementary Fig. 5 ) selectively triggered tyrosine phosphorylation of BCR effector proteins HS1 and SLP-65 (ref. 12) (Fig. 3d) . Signalling by HA was comparable to that induced by IgM cross-linking. Furthermore, mutation of Ile53Ala/Phe54Ala in CDR H2 abolished the response to HA stimulation, consistent with the binding data and confirming the importance of the germline CDR H2 structure in naive B-cell activation.

These findings indicate that engagement of low-affinity germline IGHV1-69 antibody can lead to BCR activation, thus triggering further maturation and the subsequent humoral immune response. 

We have shown in this study that IGHV1-69 antibodies, with no mutation of the germline-encoded VH sequence, engage influenza HA with sufficient affinity to trigger B-cell activation. In all cases, engagement depends on membrane presentation of antibodies with the same structural determinant-two specific hydrophobic residues in the CDR H2. Mutational analysis further revealed that just a few mutations convert a germline IGHV1-69 into an antibody with the full activity of CR6261. Taken together, these results suggest that the IGHV1-69 germ line is poised to form, with a varied set of CDR H3 sequences, broadly neutralized antibodies directed against the highly conserved stem of influenza HA.

Having an antibody with the inherent potential to recognize a common feature of influenza virus would seem to offer obvious evolutionary advantages. Although neutralizing antibodies against other viruses such as human immunodeficiency virus (HIV), SARS and hepatitis C virus also use the IGHV1-69 gene for recognition [13] [14] [15] [16] [17] [18] , these interactions differ. For example, 17b antibody binding to the CD4induced site on HIV-1 Env 14 and 8066 antibody binding to HIV-1 gp41 (ref. 19 ) orient the CDR H2 into a hydrophobic cleft whereas others such as the SARS neutralizing antibody M396 orient CDR H2 into a more hydrophilic site 16 . In each case, the actual fold recognized by the CDR H2 is distinct from that recognized on the HA stem. Together, the data suggest that the IGHV1-69 CDR H2 motif is particularly well adapted to recognize the specific protein fold that is highly conserved on the stem of diverse influenza virus HAs.

This germline VH gene, although expressed as part of the adaptive immune response, may therefore serve as a primordial pattern recognition receptor, structurally adapted to participate in recognition of such hydrophobic grooves 3 . As no other heavy-chain V genes have a hydrophobic CDR H2 (ref. 14) , we speculate that influenza may have exerted selection pressure leading to retention of this VH gene. It is also conceivable that the adaptive immune system retains this innate capacity for hydrophobic contact to facilitate the generation of highaffinity antibodies after only a limited degree of somatic mutation. This situation contrasts with the considerable somatic mutation (31%) observed in broadly neutralizing antibodies directed to the CD4binding site of HIV 20 , a more recently evolved virus for which there has been less opportunity to select for protective antiviral genes.

The majority of current vaccine-elicited influenza antibodies are directed to the globular head region of viral surface glycoprotein HA, which undergoes considerable antigenic drift to evade the human immune system. Thus, influenza vaccination requires annual assessment of the strains likely to circulate in the coming year to generate protective immunity for the world population. Consequently, there is great interest in the development of a universal influenza vaccine 21, 22 . Antibodies targeting the highly conserved HA stem epitope represent the vast majority of all broadly neutralized antibodies isolated against influenza so far [1] [2] [3] [4] . We have previously demonstrated that it is possible to elicit CR6261-like stem-directed antibodies by vaccination 22, 23 . Despite sequence variations in the CDRs and FR regions, IGHV1-69 antibodies evolve with a relatively small number of somatic mutations. At the same time, robust induction of these neutralizing antibodies in humans remains a challenge. Analysis of current genomics data (1000 Genome Project) reveals a polymorphism in the IGHV1-69 allele at the contact site in CDR H2 where Leu replaces Phe 54 at a predicted homozygous frequency of 13% in the general population. This substitution abolishes HA recognition by all three germline antibodies ( Supplementary Fig. 6 ). This finding underscores the remarkable specificity in the CDR H2 contact and raises the possibility that some vaccine recipients may not mount a IGHV1-69 response. At the same time, ,30% of stem antibodies isolated until now derive from other VH gene segments 1,2 , and it remains possible that such individuals might develop responses from other IgG loci. Taken together, knowledge of the initial engagement of the HA stem with the IGHV1-69 V gene, its subsequent BCR stimulation and somatic mutation will aid in the rational design of universal influenza vaccines.

The genes encoding wild-type HA proteins (H1 A/New Caledonia/20/1999 (1999 NC) and H5 A/Indonesia/05/2005 (2005 IND)) and somatically mutated and inferred germline antibodies of CR6261, FE53, and 1009-3B05 were synthesized. Somatically mutated CR6261 germline revertants and germline CDR H2 Ala mutations of Ile and Phe were constructed by introducing mutations to germline heavy chains by site-directed mutagenesis. Plasmids encoding these proteins were transfected into the human embryonic kidney cell line 293F and isolated from expression supernatants 72-96 h after transfection. HA trimeric proteins were purified as previously described 24 . Soluble Fab, IgG and IgM antibodies were purified using HisTrap, Protein G and IgM affinity columns, respectively, with additional gel-exclusion chromatography performed on IgM samples (GE Healthcare).

The purified antibody variants (1.7 3 10 24 -100 mg ml 21 ) were assayed for binding to H1 1999 NC and in some cases to H5 2005 IND by enzyme-linked immunosorbent assay (ELISA) with purified trimeric HA proteins. The various antibodies were detected by peroxidase-conjugated goat anti-human IgG unless otherwise noted. Endpoint dilutions were determined from nonlinear fit doseresponse curves using a detection limit of 43 background absorbance. CR6261 variants (0.39-25 mg ml 21 ) were also assayed for neutralization of pseudotyped recombinant lentiviruses expressing wild-type HA with the corresponding neuraminidase (NA) with a luciferase reporter gene as previously described 22 .

The crystal structure of CR6261 sHgL Fab at 2.85 Å resolution was determined by molecular replacement (Methods).

For membrane presentation of antibody, IGHV1-69 germ lines were expressed in membrane IgM format in 293F cells, which were then subject to a FACS-based HA cell surface binding assay 72 h after transfection (Methods). For BCR triggering, membrane IgM germ lines were transiently expressed in an IgM-negative Ramos B cell line and exposed to proteoliposome-arrayed H1 1999 NC (Methods). Activation was assessed by extent of tyrosine phosphorylation using the 4G10 pY antibody as described 12 .

Full Methods and any associated references are available in the online version of the paper.

Heterosubtypic neutralizing antibodies are produced by individuals immunized with a seasonal influenza vaccine

Broadly cross-reactive antibodies dominate the human B cell response against 2009 pandemic H1N1 influenza virus infection

Antibody recognition of a highly conserved influenza virus epitope

Structural and functional bases for broad-spectrum neutralization of avian and human influenza A viruses

Wide prevalence of heterosubtypic broadly neutralizing human antiinfluenza A antibodies

Immunization by avian H5 influenza hemagglutinin mutants with altered receptor binding specificity

Standard conformations for the canonical structures of immunoglobulins

Increased TCR avidity after T cell activation: a mechanism for sensing low-density antigen

TCR-peptide-MHC interactions in situ show accelerated kinetics and increased affinity

An EBV-genomenegative cell line established from an American Burkitt lymphoma; receptor characteristics. EBV infectibility and permanent conversion into EBV-positive sublines by in vitro infection

Immunoglobulin secretion by cell lines derived from African and American undifferentiated lymphomas of Burkitt's and non-Burkitt's type

The serine and threonine residues in the Ig-a cytoplasmic tail negatively regulate immunoreceptor tyrosine-based activation motif-mediated signal transduction

IGHV1-69 gene is preferentially used by hepatitis C virus-associated B cell lymphomas and by normal B cells responding to the E2 viral antigen

Structural basis of tyrosine sulfation and V H -gene usage in antibodies that recognize the HIV type 1 coreceptor-binding site on gp120

Structural basis for HIV-1 neutralization by a gp41 fusion intermediate-directed antibody

LETTER RESEARCH

Structure of severe acute respiratory syndrome coronavirus receptor-binding domain complexed with neutralizing antibody

Low pH-dependent hepatitis C virus membrane fusion depends on E2 integrity, target lipid composition, and density of virus particles

Germline-like predecessors of broadly neutralizing antibodies lack measurable binding to HIV-1 envelope glycoproteins: implications for evasion of immune responses and design of vaccine immunogens

Structural basis of HIV-1 neutralization by affinity matured Fabs directed against the internal trimeric coiled-coil of gp41

Rational design of envelope identifies broadly neutralizing human monoclonal antibodies to HIV-1

Induction of unnatural immunity: prospects for a broadly protective universal influenza vaccine

Induction of broadly neutralizing H1N1 influenza antibodies by vaccination

DNA priming and influenza vaccine immunogenicity: two phase 1 open label randomised clinical trials

Comparative efficacy of neutralizing antibodies elicited by recombinant hemagglutinin proteins from avian H5N1 influenza virus

Characterization of the human Ig heavy chain antigen binding complementarity determining region 3 using a newly developed software algorithm

Sequences of Proteins of Immunological Interest (US Dept. of Health and Human Services

Genes encoding HA proteins A/New Caledonia/20/1999 (H1 1999 NC

The germ lines were predicted using JOINSOLVER 25 and were created from the following segments: IGHV1-69*01

For the HA ligand, an N-linked glycan (HA1 Arg192Thr, H3 numbering) was introduced into the receptor binding site (DRBS) of wild-type or HA stem binding mutant (Ile45Arg/Thr49Arg mutations in HA2, Dstem) H1 1999 NC to prevent sialic-acid-mediated cell binding. The resulting avi-tagged trimer was biotinylated, purified by size-exclusion chromatography, and labelled with a streptavidin-linked PE fluorophore 20 . Two days after transfection, 2 3 10 6 cells expressing membrane IgM versions of CR6261 germ line (gl), FE53 gl, 1009-3B05 gl (all 6 CDR H2 mutations Ile53Ala and Phe54Ala), or VRC01 (mock) were placed on ice, stained with violet fluorescent reactive dye (ViViD, catalogue no. L34955, Invitrogen) for 30

PE-anti-human kappa chain (catalogue no. 12-9970-42, eBioscience) (FE53 gl, 1009-3B05 gl)

For unmodified/native HA binding, cells were incubated with 4 mg ml 21 trimer along with 30 mM 69-sialyllactose to prevent sialic-acid-mediated cell binding (Supplementary Fig. 4) (1 h, 4 uC, in PBS). Cells were washed twice and then fixed in PBS containing 0.5% PFA. ViViD-negative cell surface PE intensity was quantified by flow cytometry (ViViD excitation at 407 nm, bandpass filter 450/50 nm

One gram of 1,2-dioleoyl-sn-glycero-3-phosphocholine:1,2-dioleoyl-sn-glycero-3-[(N-(5-amino-1-carboxypentyl)iminodiacetic acid)] succinyl (Avanti Polar Lipids) in a 1:1 molar ratio was evaporated under a stream of nitrogen for 1 h. The dry lipid film was then rehydrated in 1,000 ml of liposome buffer (50 mM HEPES, 150 mM NaCl, pH 7.25 (HBS)) and shaken for 40 min, all the time being heated above the membrane melting temperature of the lipid mixture. The resulting homogeneous suspension was subjected to ten freeze-thaw cycles and then extruded 21 times through a 100-nm pore polycarbonate membrane using the Avanti mini-extruder (Avanti Polar Lipids). HA proteoliposomes were produced by incubating the resultant liposomes with His-tagged DRBS H1 1999 NC trimer for 1 h at room temperature (HA trimer:lipid molar ratio of 1:900). The sample was then adjusted to 15% iodixanol (in 1.25 ml HBS) and overlaid with 1.75 ml, 0.5 ml and 0.5 ml of 10%, 2.5% and 0% iodixanol in HBS, respectively. Samples were then centrifuged at 200,000g in a TH660 rotor (Sorvall) for 2 h. The proteoliposome fraction, which concentrated at the 2.5-0% iodixanol interface (Supplementary Fig. 5), was collected and dialysed overnight (Slide-A-Lyzer Dialysis Cassette, 10000 MWCO, catalogue no. 66380, Thermo Scientific) to remove density gradient material

B-cell activation. Ramos B lymphocytes (ATCC) were initially stained with fluorescently conjugated anti-lambda chain and anti-IgM monoclonal antibodies (mouse PE-anti-human lambda chain

After SDS-PAGE, BCR activation was assessed by western blot analysis of the cell lysate: 4G10 pY (catalogue no. 05-0321, Millipore) reactivity to phosphotyrosine-SLP-65 (p65) and phosphotyrosine HS1 (p75) as described 12,30,31 . Total phosphotyrosine intensity from three independent experiments was measured by densitometry (Image Processing in Java (Image J) software with curve area density calculation performed in Microsoft Excel). Phosphotyrosine intensity was standardized to the level of cell lysate actin (monoclonal anti-b-actin, catalogue no. A5316, Sigma), with the background value for BCR activation being defined as the extent of stimulation in untransfected IgM-negative Ramos exposed to 0.5 mg ml 21 mouse anti-human IgM F(ab9) 2 . Crystallography. The Fab fragment of CR6261 sHgL was concentrated to 7 mg ml 21 in PBS and crystallized in a hanging drop over a reservoir containing 100 mM imidazole pH 6.5, 17.5% polyethylene glycol 8,000, and 3% 2-methyl-2,4-pentanediol. Crystals were cryoprotected by transfer through paratone-N and vitrified in liquid nitrogen. Data to 2.85 Å resolution were collected at Southeast Regional Collaborative Access Team (SER-CAT) 22-ID beamline at the Advanced Photon Source, Argonne National Laboratory. The structure was determined by molecular replacement using the program PHASER with the structure of CR6261 bound to influenza HA (3GBN) as the search model, and refined using the program PHENIX (Supplementary Table 1). The asymmetric unit contains eight copies of the Fab. The CDR H1 was deleted from the model before refinement. An electron density map calculated before modelling the CDR H1

Protective immunity to lethal challenge of the 1918 pandemic influenza virus by vaccination

Transmembrane signaling: the joy of aggregation

It's all about change: the antigendriven initiation of B-cell receptor signaling

Evidence for a preformed transducer complex organized by the B cell antigen receptor

Syk is a dual-specificity kinase that selfregulates the signal output from the B-cell antigen receptor

Acknowledgements We thank N. Longo for help with JOINSOLVER analysis, Z.-Y. Yang for the design of membrane-bound IgM antibody, and X. Chen for technical support. We also thank A. Tislerics for manuscript preparation, B. Hartman and J. 

Supplementary Information is linked to the online version of the paper at www.nature.com/nature.