The pre-existing human antibody repertoire to computationally optimized influenza H1 hemagglutinin vaccines

The computationally optimized broadly reactive antigen (COBRA) approach has previously been used to generate hemagglutinin (HA) immunogens for several influenza subtypes that expand vaccine-elicited antibody breadth. As nearly all individuals have pre-existing immunity to influenza viruses, influenza-specific memory B cells will likely be recalled upon COBRA HA vaccination. We determined the epitope specificity and repertoire characteristics of pre-existing human B cells to H1 COBRA HA antigens. Cross-reactivity between wild type HA and H1 COBRA HA proteins were observed at both the oligoclonal B cell level and for a subset of isolated monoclonal antibodies (mAbs). The mAbs bound five distinct epitopes on the pandemic A/California/04/2009 head and stem domains, and the majority of the mAbs had HAI and neutralizing activity against pandemic H1 strains. Two head-directed mAbs, CA09-26 and CA09-45, had HAI and neutralizing activity against a pre-pandemic H1 strain. One mAb, P1-05, targets the stem region of H1 HA proteins, but does not compete with known stem-targeting H1 mAbs. We determined that mAb P1-05 recognizes a recently discovered membrane proximal epitope on HA, the anchor epitope, and we identified similar mAbs using B cell repertoire sequencing. In addition, the trimerization domain distance from HA was critical to recognition of this epitope by P1-05. Overall, these data indicate that seasonally vaccinated individuals possess a population of functional H1 COBRA HA- reactive B cells that target head, central stalk, and anchor epitopes, and demonstrate the importance of structure-based assessment of subunit protein vaccine candidates to ensure accessibility of optimal protein epitopes. Significance Influenza imposes significant human and economic costs every year. The current seasonal vaccine elicits primarily strain-specific antibodies, and year to year vaccine effectiveness is variable. The COBRA approach could provide longer protection and obviate the requirement for annual vaccination. Whereas COBRA HAs have previously been evaluated in animal models, the pre-existing COBRA HA-reactive human B cell population has yet to be elucidated, and is important to identify specific B cells that may be recalled by H1 HA COBRA vaccination. This work demonstrates that seasonally vaccinated individuals possess a functional B cell population targeting both head and stem domains that could be recalled with COBRA HA immunogens.

Introduction toward clinical trials. Here, we identify epitope and repertoire characteristics of the recalled antigen-reactive IgGs from the human plasma IgG standard curve. The EC50 value for each mAb The HAI titer for each mAb was determined as previously described (16). Influenza viruses were to achieve >95% confluency. Cells were washed twice with PBS, then 200 µL of virus, diluted in 231 VGM to an MOI=0.01, was added. Cells were returned to 37 °C, 5% CO2. Infection proceeded for 1 hr. Viral inoculum was then removed and VGM containing mAb at 1x IC50 was added. Cells 233 were returned to 37 °C, 5% CO2 for 24 hr. For the subsequent passage, MDCK cells at >95% 234 confluency were washed with PBS twice, and a 1:10 dilution of the media from the first passage 235 in 200 µL virus growth media was used to infect MDCK cells, followed by a 1 hr incubation. Viral 236 inoculum was removed and virus growth media containing mAb at 1x IC50 was added, and cells 237 were incubated at 24 hr at 37 °C, 5% CO2. This was done for five passages with mAb at 1x IC50, 238 then one passage with mAb at 2x IC50, followed by two passages with mAb at 3x IC50 for eight 239 passages total. After the eighth passage, media containing escape mutant virus was stored, and 240 cells were pelleted by centrifugation at 3000 rpm. Cells were lysed using SPRI beads and RNA 241 was extracted. vRNA encoding the HA gene was amplified using the Uni12 primer specific to all

251
FluoSpheres were then centrifuged at 5000 rpm for 15 minutes, 900 µL supernatant was removed, 252 and the FluoSpheres were resuspended with 900 µL of 0.1% PBS. This process was repeated 253 for a second wash, then the FluoSpheres were resuspended with 20 µg of biotinylated Y2 protein.
rocking. Next, HA-specific antibodies were diluted in complete RPMI media (cRPMI, RPMI + 10% with the antibody followed by a 2 hr incubation at 37 °C for opsonization. After 1. 5

311
Biolayer interferometry kinetic assays were performed in triplicate on the Octet® Red384 system 312 (Sartorius) with a buffer containing PBS, 1% BSA, and 0.05% Tween. Anti-penta-HIS biosensors 313 were immersed in buffer for 120 s, then loaded with 10 µg/mL Y2 for 300 s. The biosensors were 314 then dipped into buffer for 120 s to obtain a baseline, dipped into buffer containing P1-05 Fab in 315 a dilution series ranging from 54 nM to 0.67 nM for 300 s, and buffer for 600 s to measure 316 dissociation. The data were processed in the Octet Data Analysis HT software v7 (Sartorius).

317
Each curve was reference subtracted, aligned to the baseline, and aligned for inter-step correction 318 through the dissociation step for each curve. Each replicate was fit globally for well-resolved 319 curves in the dilution series using a 1:1 binding model. Parameters were optimized based on the 4k camera (FEI). Micrographs were collected with Leginon and particles were picked using a difference of Gaussians particle picker and processed with Appion (31-33). Particles were Single-cell V(D)J sequencing and analysis.
reactive IgGs indicated that binding to A/California/04/2009 HA protein was consistently higher.

362
The majority of subjects demonstrated significant P1 HA-reactive IgG titers that, while lower than

447
To determine the epitopes bound by the panel of 27 mAbs isolated from these vaccinated subjects, 448 biolayer interferometry-based epitope binning was utilized as previously described (24, 41).

449
Biosensors were loaded with A/California/04/2009 HA protein, associated with one mAb, and then

481
which may lead to shielding of the antibody epitope (42). It is possible that this epitope is present 482 between epitopes 1 and 3 on the HA head domain, and that this K129N mutation may be a method 483 of virus escape from mAbs that bind in different orientations within these distinct epitopes. No escape mutations were found for any mAbs in epitope 3. Epitopes 4 and 5, which are defined by mAbs CA09-38 and P1-05, respectively, did not generate any escape mutations.

487
Epitope 5 is located on the stem of the H1 HA protein yet does not overlap with the conserved 488 stem epitope of mAb CR6261 (Figure 3A, 6). To determine the epitope of mAb P1-05, we 489 generated a complex of Y2 HA bound to P1-05 Fab fragments and evaluated its structure by