Influenza H3 and H1 hemagglutinins have different genetic barriers for resistance to broadly neutralizing stem antibodies

In the past decade, the discovery and characterization of broadly neutralizing antibodies (bnAbs) to the highly conserved stem region of influenza hemagglutinin (HA) have provided valuable insights for development of a universal influenza vaccine. However, the genetic barrier for resistance to stem bnAbs has not been thoroughly evaluated. Here, we performed a series of deep mutational scanning experiments to probe for resistance mutations. We found that the genetic barrier to resistance to stem bnAbs is generally very low for the H3 subtype but substantially higher for the H1 subtype. Several resistance mutations in H3 cannot be neutralized by stem bnAbs at the highest concentration tested, do not reduce in vitro viral fitness and in vivo pathogenicity, and are often present in circulating strains as minor variants. Thus, H3 HAs have a higher propensity than H1 HAs to escape major stem bnAbs and creates a potential challenge in the development of a bona fide universal influenza vaccine. ONE SENTENCE SUMMARY Acquisition of resistance by influenza virus to broadly neutralizing hemagglutinin stem antibodies varies tremendously depending on subtype.


38
The major surface antigen of influenza virus, the hemagglutinin (HA), is composed of a highly 39 variable globular head domain that houses the receptor binding site and a conserved stem 40 domain that is responsible for membrane fusion (1). All of the major antigenic sites on HA are 41 located on the HA globular head (2-5), which is immunodominant over the stem (6). However, 42 most antibodies to the globular head domain are strain-specific. In contrast, although harder to 43 elicit during natural infection or vaccination, many HA stem antibodies have impressive cross-44 reactive breadth (7,8). The isolation, characterization and structure determination of broadly 45 neutralizing antibodies (bnAbs) to the HA stem over the past decade have provided tremendous 46 insights into antiviral and vaccine development against influenza virus (9), including immunogen 47 design towards a universal influenza vaccine (10)(11)(12). Several stem bnAbs are also currently in 48 clinical trials as therapeutics (13). Stem bnAbs have also provided templates for design of small 49 proteins, peptides and small molecules against influenza virus (14-18). Therefore, while 50 influenza virus remains a major global health concern, stem bnAbs open up multiple promising 51 avenues to tackle this challenging problem.

53
However, emergence of resistance mutations can be a major obstacle for antiviral and vaccine 54 development. Several studies have reported difficulty in selecting strong resistance mutations to 55 stem bnAbs even after extensive passaging of the viruses (19-21), or through deep mutational 56 scanning (22), which is a comprehensive and unbiased approach (23). Nonetheless, strong 57 resistance mutations have been reported in other studies through virus passaging (20,24,25).

58
It is unclear then why some studies were able to identify strong resistance mutations while 59 others could not. Here we systematically compare how readily resistance can emerge to stem 60 bnAbs in H3 and H1 HAs, and find that there are major differences between the subtypes.

78
We quantified the in vitro fitness of 147 out of 152 possible single viral mutants and 6,234 out of 79 10,108 possible double viral mutants across the eight residues of interest in H3/HK68 HA2 80 under five different conditions: no antibody, 2 µg/mL CR9114 IgG, 10 µg/mL CR9114 IgG, 0.3 81 µg/mL FI6v3 IgG, and 2.5 µg/mL FI6v3 IgG ( fig. S1). In the absence of antibody, many viral 82 mutants have a relative fitness [proxy for replication fitness (30)], similar to wild type (WT), 83 which was set as 1 (Fig. 2, A and B), and indicate that the HA stem region can tolerate many 84 mutations.

86
We further quantified the relative resistance of each viral mutant by normalizing their relative  103 against mutants I45Y/S/N/F/W were all >100 and ≥20 µg mL -1 , respectively, compared to 3.1 104 and 0.2 µg mL -1 for WT. This validation experiment substantiates our finding that strong 105 resistance mutations are prevalent in H3/HK68.

107
Natural occurrence of resistance mutations 108 Next, we explored whether these resistance mutations were found in naturally circulating strains 109 While most strong resistance mutations have not yet been observed in naturally circulating 110 strains, it is important to note that a few could be identified at low frequency in natural human 111 H3N2 isolates (33), including I45T, I45M and N49D (Fig. 4A). I45T is also observed in human 112 H3N2 isolates sequenced without any passaging ( fig. S4A), implying that its presence was not 113 due to a passaging artifact (34). Moreover, the strong cross-resistance mutation I45F was found 114 in all human H2N2 viruses that circulated from 1957 to 1968 (Fig. 4B, fig. S4B), while almost all 6 avian H2N2 viruses have Ile45 (Fig. 4C), and explains why it is more difficult for human H2N2 116 viruses to be bound or neutralized by some stem bnAbs compared to other subtypes (24, 26, 117 35). Thus, these findings suggest that some resistance mutations to stem bnAbs already occur 118 in circulating strains.

125
I45F were able to escape in vivo prophylactic protection. While mice infected with WT were 126 completely protected by CR9114 IgG at all tested doses (1, 4, and 10 mg kg -1 ), mutants I45T, 127 I45M, and I45F were lethal even at the highest dose of CR9114 IgG (Fig. 4, D to G).

130
To dissect the resistance mechanism, we tested the binding of H3/HK68 I45T, I45M, and I45F 131 recombinant HAs to CR9114 and FI6v3, and also to another stem bnAb 27F3 (35), which 132 utilizes the same V H 1-69 germline as CR9114 and similarily neutralizes group 1 and 2 influenza 133 A viruses. The binding (K d ) of CR9114 Fab, CR9114 IgG, 27F3 Fab, 27F3 IgG, and FI6v3 IgG 134 was all diminished against the HA mutants compared to WT (Table 1 and (Table 1). As a control, we also tested binding of bnAb S139/1 that 139 targets the receptor-binding site far from the stem epitope (36, 37). S139/1 IgG affinities against 7 those HA mutants (K d = 1.8 nM to 3.1 nM) were similar to WT (K d = 2.1 nM). Thus, virus 141 resistance to stem bnAbs correlated with a decrease in binding affinity to the mutant HAs.

143
To understand the structural basis of the resistance, we determined crystal structures of HAs 144 with HA2 mutations I45T, I45M, and I45F to 2.1 to 2.5 Å resolutions (table S1 and fig. S7A).

145
Compared to WT (Ile45), the shorter side chain of I45T would create a void when CR9114 is 146 bound ( fig. S7B) that would be energetically unfavorable. In contrast, the longer flexible side 147 chain of I45M would likely clash with CR9114 ( fig. S7B), but CR9114 is still able to bind the 148 I45M mutant, albeit with much lower affinity than WT (Table 1). The I45F mutant, however, 149 makes a more severe clash with CR9114 and no binding was detected (Table 1,

194
A and B). In H3 HA, mutation of HA2 residue 53 would abolish a hydrogen bond to the 195 complementarity-determining region (CDR) H3 of FI6v3 ( fig. S11B). Together, these results

196
suggest that the prevalence of resistance mutations to stem bnAbs is a general phenomenon for 197 the H3 subtype, but not the H1 subtype.  S9, A and B) show that many mutations can be 205 tolerated in the HA stem, similar to H3/HK68 ( Fig. 2A, fig. S9, C to F). Thus, the difference in 206 genetic barrier to resistance to stem bnAbs between H1 and H3 subtypes cannot be fully 207 explained by their ability to tolerate mutations (i.e. fitness cost of mutations).

209
We therefore further compared the structures of CR9114 in complex with H3 HA and in complex 210 with H5 HA (Fig. 5G) (26). Since the structure of CR9114 with H1 HA is not available, CR9114 211 with H5 HA was used instead, as it also belongs to group 1 HAs and is therefore more similar to 212 H1 than to H3 HA (group 2). Structural comparison indicates that CR9114 packs tighter to the 213 helix A of H3 HA than to H5 HA. Specifically, there is ~1 Å difference in the position of the Cα of 214 HA2 Ile45. Subsequently, a bulkier substitution at HA2 Ile45, such as I45M, would create a 215 larger disruption of the CR9114-HA binding interface in the context of H3 subtype. Thus, subtle 216 differences in the binding of bnAbs to different HA subtypes may lead to differences in how 217 antibodies are affected by mutations in or near the epitope.

218
10 219 Similar observations can be made for FI6v3. The orientation of Tyr100c on CDR H3 of FI6v3 220 differs when binding to H1 or H3 HAs (Fig. 5H) (27). The position of HA2 Ile45 also differs 221 between H1 and H3 HAs when FI6v3 is bound. As a result, Tyr100c of FI6v3 packs tighter to 222 HA2 Ile45 of H3 than to H1 HA. Thus, a bulkier substitution at HA2 Ile45 will disrupt binding 223 between FI6v3 and H3 HA to a greater extent than FI6v3 to H1 HA. Therefore, the low genetic 224 barrier to resistance to stem bnAbs in the H3 subtype can be at least partly attributed to both 225 high mutational tolerance in the HA stem and subtype-specific structural features. While a 226 number of subtype-specific structural features are known in the stem region (40)  focused on targeting the HA stem (8,9). If stem bnAbs begin to be distributed on a global scale, 244 11 the immunological pressure on the HA stem will certainly surge to a level not previously seen.

245
Our findings here indicate that resistance mutations could emerge, at least in H3 subtype.

247
Although resistance mutations to stem bnAbs are still rare in currently circulating influenza 248 strains (Fig. 4A), it is important to evaluate the potential impact of such mutations since many 249 vaccine strategies aim to elicit anti-stem antibodies. In fact, we were not able to overcome some 250 key resistance mutations (I45T, I45M, and I45F) by in vitro evolution of CR9114 ( fig. S12).

251
Nonetheless, the best strategy to prevent or overcome such resistance may involve delivery or 252 elicitation of a combination of antibodies with different resistance profiles. In addition, it remains 253 to be explored whether stem bnAbs exist or can be generated that are difficult to escape from 254 the H3 subtype. The discovery and characterization of bnAbs with different escape profiles will 255 therefore continue to be key to broaden our arsenal against influenza virus. For example, 256 human H2N2 virus, which carries a Phe at HA2 residue 45, often has low reactivity with stem 257 bnAbs (24,26,27,35,43), although a very few can have high potency against human H2N2 (44)(45)(46). Future studies on anti-stem responses against human H2N2 and emerging viruses, 259 such as H5N1 and H7N9, may provide further insights into how to overcome potential 260 resistance when immune pressure is transferred to the HA stem.