A public broadly neutralizing antibody class targets a membrane-proximal anchor epitope of influenza virus hemagglutinin

Broadly neutralizing antibodies against influenza virus hemagglutinin (HA) have the potential to provide universal protection against influenza virus infections. Here, we report a distinct class of broadly neutralizing antibodies targeting an epitope toward the bottom of the HA stalk domain where HA is “anchored” to the viral membrane. Antibodies targeting this membrane-proximal anchor epitope utilized a highly restricted repertoire, which encode for two conserved motifs responsible for HA binding. Anchor targeting B cells were common in the human memory B cell repertoire across subjects, indicating pre-existing immunity against this epitope. Antibodies against the anchor epitope at both the serological and monoclonal antibody levels were potently induced in humans by a chimeric HA vaccine, a potential universal influenza virus vaccine. Altogether, this study reveals an underappreciated class of broadly neutralizing antibodies against H1-expressing viruses that can be robustly recalled by a candidate universal influenza virus vaccine.


31
Broadly neutralizing antibodies against influenza virus hemagglutinin (HA) have the potential to 32 provide universal protection against influenza virus infections. Here, we report a distinct class of 33 broadly neutralizing antibodies targeting an epitope toward the bottom of the HA stalk domain 34 where HA is "anchored" to the viral membrane. Antibodies targeting this membrane-proximal 35 anchor epitope utilized a highly restricted repertoire, which encode for two conserved motifs 36 responsible for HA binding. Anchor targeting B cells were common in the human memory B cell 37 repertoire across subjects, indicating pre-existing immunity against this epitope. Antibodies 38 against the anchor epitope at both the serological and monoclonal antibody levels were potently

105
Identification of antibodies targeting the anchor epitope

106
To dissect conserved HA stalk domain epitopes, we generated mAbs from acutely activated plasmablasts 107 isolated from subjects who received licensed or experimental influenza virus vaccines or were naturally 108 infected with pH1N1 during the 2009 pandemic (Table S1). Notably, plasmablasts found in the blood of 109 subjects after infection or vaccination derive from pre-existing MBCs (Andrews et al., 2015), and generation 110 of mAbs from plasmablasts allows for the dissection of how distinct influenza viruses recall pre-existing 111 immunity. We also generated mAbs from sorted HA + B cells one month following vaccination with an antibodies. MAbs that bound the cHA and that lacked HAI activity were classified as those binding the HA 120 stalk domain. Of all mAbs tested, nearly 49% targeted the HA stalk domain, whereas 40% targeted the HA 121 head domain ( Figure S1A). To investigate what proportion were binding the BN stalk domain epitope, we 122 competed the stalk binding mAbs from our cohorts with CR9114, a well-defined antibody targeting the BN 123 stalk epitope (Dreyfus et al., 2012). We identified that only 21% of mAbs targeting the stalk domain had 124 greater than 80% competition with CR9114 ( Figure S1B), indicating most H1 stalk domain targeting 125 antibodies were binding other epitopes of the HA stalk.

126
To investigate which epitopes the remaining 79% of mAbs were targeting on the stalk domain, we 127 performed negative stain electron microscopy with two stalk domain binding mAbs. Both mAbs bound an 128 epitope near the anchor of the HA stalk, towards the lower portion of the HA protomer ( Figure 1A-B; Figure   129 S1C-D). Both mAbs were oriented at an upward angle towards the epitope ( Figure 1A-B), suggesting this 130 epitope may be partially obstructed by the lipid membrane and may only be exposed for antibody binding To understand what proportion of stalk binding mAbs were binding to the anchor epitope, we 137 competed 047-09 4F04 mAb with the remaining stalk binding mAbs that did not compete with CR9114. In 138 total, we identified 50 distinct mAbs that competed for binding to the anchor epitope from a total of 21 139 subjects (Table S2) and accounted for 28% of all stalk mAbs identified ( Figure 1E-F). Together, these data 140 indicate that the anchor epitope is a common target of antibodies binding the H1 stalk domain.

141
Antibodies binding the anchor are broadly reactive amongst H1 viruses

142
As the stalk domain is conserved amongst influenza viruses, we next determined the viral binding breadth 143 of antibodies targeting the anchor epitope. Anchor mAbs were broadly reactive amongst H1-expressing 144 viruses, including a swine origin H1N2 virus, but rarely cross-reacted with other influenza subtypes ( Figure   145 2A; Figure S2A-B), as often occurs for antibodies targeting the BN stalk epitope (Figure 2A; Figure S2B).

146
While highly conserved amongst H1 viruses, the anchor epitope was poorly conserved across divergent 147 group 1 viral subtypes ( Figure S2C). Anchor epitope targeting mAbs had nearly a 2-fold higher affinity for 148 pH1N1 virus than mAbs targeting the BN stalk epitope ( Figure 2B). Because the anchor epitope is partially 149 obstructed by the lipid membrane, we next tested whether anchor binding mAbs had reduced affinity for 150 whole virus relative to recombinant HA (rHA). MAbs binding the anchor epitope and the BN stalk epitope 151 both exhibited reduced affinity for the whole virus (A/California/7/2009) relative to rHA from the same virus, 152 whereas mAbs targeting the HA head domain had similar affinity for whole virus and rHA ( Figure S2D).
These data indicate that antibodies targeting the anchor epitope are broadly reactive amongst H1-154 expressing viruses.

155
Antibodies targeting the anchor epitope maintain binding to HA mutants in the stalk domain 156 H1N1 viruses have acquired several mutations within the HA stalk domain, likely due to antibody 157 selective pressures or to increase stability (Cotter et al., 2014). To understand whether these mutations 158 have affected antibody binding to the anchor epitope, we screened mAbs against naturally occurring 159 mutants and experimentally identified viral escape mutants of BN stalk epitope binding mAbs ( Figure Table S3). Anchor epitope binding mAbs were mostly unaffected by the mutants tested, 161 whereas most of the mAbs targeting the BN stalk epitope were affected by at least one mutant, notably 162 Q42E mutation in HA2 ( Figure 2D). Regardless of mAb specificity, most antibodies had reduced binding 163 to A44V of HA2, which was recently shown to preferentially grow in the presence of mAbs against the BN 164 stalk epitope (Park et al., 2020). While A44 is distant from the anchor epitope, the A44V mutation was 165 shown to affect the conformation of the HA stalk (Park et al., 2020) and could explain the broad reduction

182
To test whether mAbs targeting the anchor epitope were protective in vivo, we prophylactically 183 administered a cocktail of 5 mAbs targeting the anchor epitope or the BN stalk epitope to mimic a polyclonal 184 response against these epitopes and infected mice with a lethal dose of a mouse-adapted pH1N1 virus

209
To confirm that anchor epitope targeting mAbs generated from plasmablasts were representative   S5C-D). Furthermore, we identified and generated anchor epitope targeting mAbs from acutely activated 249 plasmablasts isolated from one subject that received the cH8/1 IIV+AS03 prime and one subject that 250 received the cH5/1 IIV+AS03 boost (Table S1 and Table S2). Together, these data indicate that the cHA 251 vaccine strategy can robustly induce antibodies against the anchor epitope.   (Table S2) Figure S6C). Anchor epitope 291 targeting mAbs were mutated to a similar extent as mAbs targeting the BN stalk epitope ( Figure S6D). The

292
K-CDR3 length of anchor epitope binding mAbs was highly restricted, with all K-CDR3s being ten amino 293 acids in length ( Figure S6E). Together, these data indicate that anchor epitope targeting mAbs utilize a 294 highly restricted repertoire, particularly for the light chain.

360
The anchor targeting mAbs utilized a highly restricted repertoire that were public clonotypes across

374
Our study showed that humans have pre-existing immunity against the anchor epitope and

516
There are restrictions to the availability of mAbs from this study due to the lack of an external centralized 517 repository for its distribution and our need to maintain the stock. We are glad to share mAbs with 518 reasonable compensation by requestor for its processing and shipping.

529
Human PBMCs were obtained from multiple subjects from multiple cohorts, which is outlined in Table S1.

530
All studies were performed with the approval of the University of Chicago Institutional Review Board (ID 531 #09-043-A). The chimeric HA vaccine study cohort is identified as clinical trial NCT03300050.

701
were washed 2 times and run on a BD LSRFortessa X-20. Data were analyzed using FlowJo v10.