We have previously described broadly reactive antibodies that were protective in vivo against group 1 influenza A viruses (CR6261) (12, 13), antigenically diverse influenza B viruses (CR8033 and CR8071), and both group 1 and group 2 influenza A viruses as well as influenza B viruses (CR9114) (20). CR6261 binds the stem region of HA and neutralizes the virus by preventing the conformational changes of this protein that are required for the viral fusion process (12). In contrast, CR8033 and CR8071 bind non-overlapping epitopes in the head region of influenza B HA and neutralize in vitro. While CR8033 disrupts virus attachment to sialic acids on the surface of host cells – as can be measured by the hemagglutination inhibition (HAI) assay (HAI⁺) – CR8071 inhibits the release of progeny viruses from infected cells (19, 20). Interestingly, CR9114 binds to an epitope in the stem region of HA that is conserved among all influenza A and B viruses but only inhibits the viral fusion process of influenza A viruses and thus only has in vitro neutralizing activity against these viruses (20).

To further explore the molecular mechanism by which these broadly reactive antibodies provide protection in vivo, we characterized the antibodies, together with antibody 2D1, previously shown to inhibit the interaction between the HA and the sialic acids of several different H1N1 strains (29), in terms of their ability to induce effector functions as measured by a bio assay that measures antigen-specific antibody-dependent FcγRIIIa activation (30, 31).

The HAI⁺ antibodies CR8033 and 2D1 showed virtually no FcγRIIIa activation, while CR6261, CR8071, and CR9114 were able to induce robust FcγRIIIa activation to HA-transfected target cells (Figure 1).

The distinction between FcγRIIIa activation mediated by CR6261, CR9114, and CR8071 on one hand and CR8033 and 2D1 on the other appears to coincide with the ability to block HA–sialic acid interaction. This suggests that HA binding to sialic acid on immune cells may play a role in FcγRIIIa activation mediated by anti-HA antibodies. To investigate the role of HA–sialic acid interactions, a mutation in the HA of A/California/07/09 (L195F) and B/Florida/04/06 (L201F) was introduced preventing the HAs from interacting with sialic acid (non-SA binding HA) (Figure S1 in Supplementary Material). The FcγRIIIa activation to non-SA-binding HA mediated by CR9114 was much lower as compared to the FcγRIIIa activation to the wild-type HA (Figures 2A,B), while the antibody binding was comparable between the non-SA binding and wild-type HA (Figure S1 in Supplementary Material).

The HAI⁺ antibodies 2D1 and CR8033 prevent the HA-expressing target cells from interacting with sialic acids on immune cells and may therefore be able to inhibit FcγRIIIa activation mediated by the HAI-negative antibodies CR6261, CR9114, and CR8071. In agreement with the dependency of potent FcγRIIIa activation on sialic acid interactions, 2D1 and CR8033 inhibit in a dose-dependent manner the FcγRIIIa activation induced by CR9114 binding to pH1N1 (A/California/07/09) HA and influenza B (B/Florida/04/06) HA, respectively (Figures 2C,D). Likewise, the FcγRIIIa activation mediated by CR6261 binding to pH1N1 HA and CR8071 binding to influenza B HA was inhibited by 2D1 and CR8033, respectively (Figure S2 in Supplementary Material). These results are consistent with a model in which interactions between HA and sialic acid residues on immune cells, in addition to the established antibody–FcγR interactions, are required for optimal FcγRIIIa activation by anti-influenza HA antibodies (Figure 3).

Upon exposure to influenza by infection or vaccination, humans develop both HAI⁺ and HAI⁻ antibodies (32). To assess whether the HAI⁺ antibodies inhibit FcγRIIIa activation in humans, we determined the HAI against H5N1 (A/Vietnam/1194/2004) and pH1N1 and FcγRIIIa activation in plasma from participants of a phase I dose-escalation study evaluating an H5N1 virosomal vaccine adjuvanted with Matrix-M (EudraCT no. 2008-006940-20). From the subjects who were vaccinated two times with a 30-μg HA dose of Matrix-M (50 μg/dose) adjuvanted virosomal vaccine, plasma samples were collected at day 0 (pre-vaccination), day 21 (post-prime), and day 42 (post-boost) for analysis. High levels of stem-binding antibodies were detected in these subjects after prime immunization. These levels were not boosted after the second immunization (Figure S4A in Supplementary Material).

HAI⁺ antibodies against the vaccine homologous H5N1 strain, but not against the heterosubtypic pH1N1 strain, were induced, in particular, after the second vaccination (Figures 4A,D). In order to evaluate the inhibiting effect of HAI⁺ antibodies, we assessed the H5N1 HA-specific FcγRIIIa activation in plasma samples taken before (pre-vaccination) and after one (post-prime) and two (post-boost) vaccinations.

To quantify the amount of inhibition of FcγRIIIa activation, we used a modified four-parameter logistic curve (Eq. 2) to model the observed dose–response curve. Using this model, we estimated an inhibition corrected for the FcγRIIIa activation dose response and determined the area between the estimated and the observed dose–response curves as measure for FcγRIIIa activity inhibition (for an illustration, see Figure S3 in Supplementary Material).

H5N1 HA-specific FcγRIIIa activation was inhibited, in particular, in plasma taken after the boost immunization (Figure 4E). The inhibition of FcγRIIIa activation strongly correlated (Kendall’s tau b = 0.697, p < 0.001) with the HAI titers to H5N1 (Figure 4F). In line with our proposed model, pH1N1-specific FcγRIIIa activation in the absence of detectable HAI antibodies was barely inhibited (Figures 4A–C).

We next assessed pH1N1-specific FcγRIIIa activation in pools of individual plasma samples that were complemented with four different concentrations of the pH1N1 HAI⁺ antibody 2D1 or a control antibody. The levels of stem-binding antibodies were unaffected by the addition of 2D1 (Figure S4B in Supplementary Material), while both the HAI titers and the inhibition of FcγRIIIa activation increased in a 2D1 dose-dependent manner (Figures 4G,H). The levels of inhibition of FcγRIIIa activation and HAI titers were unaffected when control antibody was used (Figures S4C,D in Supplementary Material). The inhibition of FcγRIIIa activation strongly correlated with the HAI response against pH1N1 (Kendall’s tau b = 0.795, p < 0.001, Figure 4I), thereby confirming our results obtained with the monoclonal antibodies.

Human IgG1 antibodies, CR6261, CR9114, CR8033, and CR8071, were constructed by cloning the heavy (VH) and light (VL) chain variable regions into a single expression vector containing either the wild-type IgG1 constant regions or an L234A + L235A double mutant (for CR8033 LALA) that abolishes the ability of the Fc part of the antibody to interact with FcγR. The variable heavy and light chain of 2D1 (29) was also cloned into the same human IgG expression vector. The resulting IgG1 had identical specificity as the published antibodies (data not shown). The antibodies were produced on HEK293-F cells that were transfected with the IgG expression constructs, and the expressed antibodies were purified from serum-free culture supernatants using protein A chromatography (HiTrap, GE Healthcare, Buckinghamshire, United Kingdom) followed by a desalting step (HiPrep 26/10, GE Healthcare).

Sixty healthy volunteers (22 males and 28 females) aged 20–49 years (mean age 31.2 years) were recruited into this clinical trial to evaluate a virosomal H5N1 vaccine that was performed between March and June 2009 at Haukeland University Hospital, Bergen, Norway. The open label phase I dose-escalating study was approved by the Regional Ethics Committee [Regional Committee for Medical Research Ethics, Northern Norway (REK Nord)] and the Norwegian Medicines Agency and registered in the European Clinical Trials Database (EudraCT no. 2008-006940-20). Informed consent was obtained from subjects prior to inclusion in the study. The subjects had never received an H5-based vaccine. Participants were randomly divided into four groups (n = 15) and vaccinated two times with 30-μg HA dose of non-adjuvanted vaccine or 1.5-μg HA, 7.5-μg HA, or 30-μg HA Matrix-M (50 μg/dose) adjuvanted virosomal vaccine, respectively, at days 0 and 21. At days 0, 21, 42, and for 6 and 12 months, blood samples were collected for analysis (43). For the current study, plasma samples collected at day 0 (unvaccinated; visit 1), 21 (after prime; visit 2), and 42 (after boost; visit 3) for the participants who received 30-μg HA adjuvanted with Matrix-M were used.

Influenza virus H1N1 A/California/07/09 reassortant (NYMC X-181), B/Florida/04/06 (wild-type), or H5N1 A/Hong Kong/156/97 reassortant (four HA units per well) were combined in quadruplicate wells with an equal volume of antibody serially diluted in PBS. Plates were incubated for 1 h at 37°C in 96-well V-bottom plates. Fifty microliters of 1% turkey red blood cells was then added to each well and incubated for 1 h at room temperature. Button formation was scored as evidence of HAI. Titers were determined using the Spearman–Kärber formula.

Binding of antibodies to wild-type and mutant influenza A and B HAs was tested by transfecting A549 cells with full-length recombinant HAs using Lipofectamin 2000 (Invitrogen, Carlsbad, CA, USA) to produce surface-expressed HA. On the following day, the cells were transferred to 96-well plates (PDL-coated black clear bottom) at a density of 1.5 × 10⁴ cells per well and incubated. Twenty-four hours later, cells were fixed (3% PFA, 15 min), washed, and incubated for 1 h with PBS/2%BSA at room temperature. Next, independent duplicates of serial antibody dilutions starting at 10 μg/ml were added to the cells (50 μl/well) and incubated for 3 h at room temperature, followed by wash steps and 30-min incubation with the secondary antibody [goat-anti-human IgG F(c) IRDye^® 800CW Conjugated] (Rockland, Limerick, ME, USA) diluted 1/500 in 5 PBS/2%BSA. After an additional wash step and fixation (1% PFA), fluorescence was measured on an Odyssey classic imager (Licor, Lincoln, NE, USA).

To determine stem-binding antibody responses, we measured CR9114-competing antibody responses as described before (44). Briefly, Maxisorp 96-well plates (Merck) were coated with purified polyclonal rabbit anti His-Tag IgG (Genscript, NJ, USA) O/N at 4°C followed by washing. After blocking and washing, plates were incubated with a titrated amount of His-Tagged HA of A/California/07/09 (in house produced and purified) for 2 h at RT. Plates were washed, and plasma was added to the plate in duplicate, serially diluted in block buffer, and incubated for 1 h at RT, followed by addition of a titrated amount of biotinylated human IgG1 CR9114 (produced and purified in house) and incubation for another hour at RT. After washing, streptavidin–HRP was added and incubated for 1 h at RT, followed by washing and OPD (Thermo Scientific, Bremen, Germany) development. The colorimetric reaction was stopped after 10 min by adding 1M H₂SO₄. The optical density (OD) was measured at 492 nm, and standard curves were created using a four-parameter logistic curve. The CR9114 competition of each sample was quantified as the slope of the linear regression of OD value on the log 10 dilution for the duplicate series.

Human lung carcinoma-derived A549 epithelial cells (ATCC CCL-185) were maintained in Dulbecco’s modified eagle medium (DMEM, Invitrogen) medium supplemented with 10% heat-inactivated fetal calf serum at 37°C. The A549 cells were transfected with wild-type B/Florida/04/2006 or A/California/07/09 (pH1N1) HA expression vectors or corresponding (non-sialic acid-binding mutants), using Lipofectamine 2000 (Invitrogen). The next day, 1.5 × 10⁴ A549 cells were seeded in white opaque flat bottom 96-wells plates. Eighteen hours later, antibodies or human plasma samples were serially diluted in assay buffer [4% ultra-low IgG FBS (Invitrogen) in RPMI 1640 (Invitrogen)]. Sample (antibody or human plasma) dilutions and ADCC Bioassay Jurkat effector (immune) cells expressing human FcγRIIIa and an NFAT-response element regulating a luciferase reporter gene (V158 variant provided by Promega, Fitchburg, MA, USA) were added to the cells and incubated for 6 h at 37°C. Cells were equilibrated at room temperature for 15 min before Bright-Glo luciferase substrate (Promega) was added. Luminescence was read out after 5 min on a Trilux Microbeta plate reader.

The interference experiments with monoclonal antibodies were performed in a similar fashion; however, instead of transfecting the A549 target cells with HA-expressing vectors, virus infection was used to obtain HA-expressing target cells. Therefore, 2 days before the experiment, 1.5 × 10⁴ A549 cells were seeded in white opaque flat bottom 96-well plates. One day before, the experiment cells were infected with 12,000 TCID₅₀ per well influenza virus. The next day, independent duplicates of 2D1 or CR8033 LALA mutant (because wild-type CR8033 applied in very high concentrations did induce some FcγRIIIa activation) were eightfold serially diluted in assay buffer (4% ultra-low IgG FBS in RPMI 1640), and a fixed concentration of CR6261, CR9114, CR8071, or control antibody was added to each dilution. FcγRIIIa activity was normalized to the FcγRIIIa activity obtained with control antibody.

The interference by different concentrations of 2D1 or control antibody (4.1, 52.3, and 209 μg) in pooled plasma samples (an equal amount of the individual plasma samples per time point were pooled) was performed on A/California/07/09 (pH1N1) HA expression target cells as described above.

FcγRIIIa activity dose–response curves are routinely fitted using a four-parameter logistic curve (Eq. 1). Here, the dose–response curve is summarized into four parameters: a lower asymptote (A), an upper asymptote (D), the inflection point (C), and a slope factor (B). (1)RLUlog=A+(D−A)×11+10(C−logX)×B

which is equivalent to fitting a linear effect in logit-log space.

An inhibition of FcγRIIIa activity, reflected by a hook effect in the dose–response curve (Figure S4B in Supplementary Material, black dots), can be seen as a combination of an active and an opposite inhibition dose–response curve (Figure S4B in Supplementary Material, respectively, curves 1 and 2). When assuming that (i) the non-inhibited maximum FcγRIIIa activation (parameter D) and the minimum FcγRIIIa activation (parameter A) are shared between the two dose–response curves and that the two dose–response curves share a common (but opposite) slope factor, the four-parameter logistic curve can be modified to model 2 opposite linear effects simultaneously: (2)RLUlog=A+(D−A)×11+10(C−logX)×B×11+10(IC−logX)×−1×B

where B represents the slope factor for the active dose–response curve, C and IC represents the inflection point of the active and the inhibiting dose–response curve, respectively (Figure S4B in Supplementary Material, curve 3).

For each subject, the amount of inhibition of FcγRIIIa activity at day 0 (unvaccinated; pre), at day 21 (post prime), and at day 42 (post boost) was determined by using Eq. 2 with time point-specific parameter estimates for C, IC, B, and D, whereas the estimate for A was shared between all three time points. In addition, to accommodate a possible vaccination effect on the FcγRIIIa activation (i.e., increase in FcγRIIIa activity at post prime or post boost), the model was constrained in such a way that the upper asymptote of the dose–response curves post prime or post boost had to be equal to, or higher than, the upper asymptote at pre-vaccination (i.e., dotted line in Figure S4A in Supplementary Material). For the data where antibody was spiked in a pool of sera, a similar approach was followed, without the constraint for the upper asymptote. The inhibition was then quantified as the area between the predicted (inhibition corrected) FcγRIIIa activity dose–response curve and observed dose–response curve (Figure S4C in Supplementary Material, area between, respectively, curves 1 and 3, depicted in green). Correlation between HAI titer and the amount of FcγRIIIa activity inhibition (i.e., area between curves) was assessed using Kendall’s tau-b.

Lentiviral pseudoparticles were generated by co-transfecting 293T cells using Lipo2000CD (Invitrogen) with pLP1, pLP2, and pLenti6-luciferase (Invitrogen), expression plasmids coding for influenza HA (wild-type and L201F mutant B/Florida/04/2006), neuraminidase (N1 HK/156/1997), and human airway trypsin-like (HAT, accession number NP_004253.1) (45). One day after the transfection, medium was replaced for culture medium (DMEM supplemented with l-glutamine, 5-mM MgCl, and 5% FBS; Invitrogen), and 96 h after transfection, the supernatants were passed through 0.45-μm filters and stored in aliquots at −80°C. HA content of pseudoparticles was determined by ELISA using anti-HA antibodies as described before (20).

Maxisorp ELISA plates (Merck, Darmstadt, Germany) were coated overnight at 4°C with 30 μg/ml Fetuin (Merck) in PBS. Plates are blocked with 1% BSA in wash buffer (PBS + 0.05% Tween-20). Twofold dilutions of wild-type or L195F mutant A/California/07/09 HA or L201F mutant B/Florida/04/06 HA pseudotyped particles were added to the wells, and plates were incubated for 1.5 h at room temperature. Human CR9114 (2 μg/ml) and peroxidase-conjugated affinipure mouse anti-human IgG (Fcγ fragment specific, 1:1000; Jackson, West Grove, PA, USA) were used for detection. Plates were incubated with SigmaFast OPD (Merck) for 20 min; reactions were stopped by adding 1M H₂SO₄ and read using a Biotek plate reader.