Hemagglutination Inhibition (HAI) Antibody Landscapes after Vaccination with diverse H7 hemagglutinin (HA) proteins

Background A systemic evaluation of the antigenic differences of the H7 influenza hemagglutinin (HA) proteins, especially for the viruses isolated after 2016, are limited. The purpose of this study was to investigate the antigenic differences of major H7 strains with an ultimate aim to discover H7 HA proteins that can elicit protective receptor-blocking antibodies against co-circulating H7 influenza strains. Method A panel of nine H7 influenza strains were selected from 3,633 H7 HA amino acid sequences identified over the past two decades (2000-2018). The sequences were expressed on the surface of virus like particles (VLPs) and used to vaccinate C57BL/6 mice. Serum samples were collected and tested for hemagglutination-inhibition (HAI) activity. The vaccinated mice were challenged with lethal dose of H7N9 virus, A/Anhui/1/2013. Results VLPs expressing the H7 HA antigens elicited broadly reactive antibodies each of the selected H7 HAs, except the A/Turkey/Italy/589/2000 (Italy/00) H7 HA. A putative glycosylation due to an A169T substitution in antigenic site B was identified as a unique antigenic profile of Italy/00. Introduction of the putative glycosylation site (H7 HA-A169T) significantly altered the antigenic profile of HA of the A/Anhui/1/2013 (H7N9) strain. Conclusion This study identified key amino acid mutations that result in severe vaccine mismatches for future H7 epidemics. Future universal influenza vaccine candidates will need to focus on viral variants with these key mutations.


Abstract (maximum 250 words)
23 Background: A systemic evaluation of the antigenic differences of the H7 influenza hemagglutinin (HA) 24 proteins, especially for the viruses isolated after 2016, are limited. The purpose of this study was to 25 investigate the antigenic differences of major H7 strains with an ultimate aim to discover H7 HA proteins 26 that can elicit protective receptor-blocking antibodies against co-circulating H7 influenza strains.

27
Method: A panel of nine H7 influenza strains were selected from 3,633 H7 HA amino acid sequences 28 identified over the past two decades (2000)(2001)(2002)(2003)(2004)(2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015)(2016)(2017)(2018). The sequences were expressed on the surface of virus 29 like particles (VLPs) and used to vaccinate C57BL/6 mice. Serum samples were collected and tested for Avian-origin influenza A hemagglutinin subtype 7 viruses (H7 AI viruses) circulate primarily in 3 avian hosts. Humans are dead-end hosts for these virus infections and the H7 epidemics rarely persist 4 among humans. However, some H7 influenza viruses may mutate in the human respiratory track and cause 5 severe recurring epidemics (1). There have been six epidemics caused by Asian H7N9 influenza viruses 6 between 2013-2018 and this raises concern that this subtype may have the potential to cause influenza virus 7 pandemics (2-4). H7N2 influenza viruses caused epidemics in 2002 and 2003 and silently circulated among 8 feline species and/or unknown reservoirs for fourteen years (5). In the northeastern U.S., H7N2 influenza 9 viruses have high affinity for the mammalian respiratory tract and are highly adapted to mammalian species 10 with increased affinity toward α2-6 linked sialic acid (6). In 2016, the feline H7N2 influenza viruses 11 resulted the transmission from shelter cats to an attending veterinarian (7). Even without adaptation, H7 between humans to initiate the next H7 influenza virus pandemic.

18
For prompt production and distribution of vaccines during a pandemic emergency, the World

19
Health Organization (WHO) has stockpiled candidate vaccine viruses (CVVs) for all H7 influenza viruses 20 (9). However, the antigenic differences of stockpiled CVVs have not been investigated, especially for the 21 H7N9 viruses isolated after 2016 (10). To prepare for the next H7 influenza virus epidemics, it is imperative 22 to identify the antigenic differences of co-circulating H7 HA proteins and clarify the target coverage by the 23 antigen.

24
There have been a small number of studies that investigated the antigenic differences of multiple H7 strains.

39
and subjected for the antigenic landscape analysis. Since it was not plausible to conduct crosssacrificed to harvest lung tissue ( Figure 2). Remaining mice were monitored for the weight loss and euthanized at 14 days post-challenge ( Figure 2). Weight loss more than 25% was used as a primary

132
Determination of HAI cut-off to predict protection against challenge.

133
The receiver operating characteristic (ROC) curve analysis between HAI titer and protection 134 against Anhui/13 challenge, as previously described (17). The protection was defined when the mouse could 135 maintain 95% of the original body weight during entire challenge study. The sensitivity and specificity of 136 four cut-off values (VLP HAI titer=40, 80,160, and 320) to predict protection were analyzed. The sensitivity 137 was calculated as "number of mouse which showed hemagglutination inhibition (HAI) titer ≥ cut-off and 138 was protected from the challenge study/ number of all protected mice". The Specificity was calculated as 139 "number of mouse which showed hemagglutination inhibition (HAI) titer < cut-off and unprotected from 140 the challenge study/ number of all unprotected mice". The ROC curve was generated by connecting plots 141 of sensitivity% versus 100-specificity% (false positive). The area under the curve (AUC) and Youden's 142 index (Sensitivity + Specificity -1) was calculated by Prism (Graphpad software). The optimal cut-off was 143 determined based on highest AUC or Youden's index to be used as a surrogate of protection.

146
The H7 HA numbering was based on a previous report (18

155
We measured the antigenic breath of the antisera collected at week 8. At week 8, all mice were challenged 156 with Anhui/13 H7N9 wild type virus, as described above, and looked for weight loss, survival, and lung 157 viral titer at 4 days-post-challenge.

160
The difference in serum HAI titer and lung viral titer among groups was analyzed by ordinary one-161 way ANOVA, followed by Tukey's multiple comparison test. The difference in body weight loss of each 162 time point was tested by Repeated Measures one-way ANOVA followed by Tukey's multiple comparison 163 test. All statistical analysis was performed using Prism GraphPad Software. Results

168
uploaded amino acid sequences were biased to isolates from Asian H7N9 epidemics between 2013-2017.

169
Since the Anhui/13-like sequences skewed the overall phylogenetic analysis, the sequences were separately  The majority of viral sequences isolated from 2013-2020 were Anhui/13-like H7N9 influenza 198 viruses (Fig. 2B). Approximately 5.12% of the HA1 sequences had 3-5% difference in the amino acid 199 sequence and represented as a separate clusters from Anhui/13-like HA sequences (Fig. 2B). This small

203
The Shanghai/13 was one of the earliest human H7N9 isolates in spring 2013, which evolved into a separate 204 phylogenetic cluster from Anhui/13-like viruses (19,20). In this sequence analysis, the Shanghai/13 virus 205 itself belonged to Anhui/13-like virus due to high homology (98.39%) of the HA amino acid sequences.

206
However, the derivatives of Shanghai/13 had divergent sequences to form a separate cluster that occupies 207~1% of the overall HA sequences (Fig. 2B).

208
The majority of non-Asian H7N9 influenza strain sequences uploaded on GSAID database between 209 2013 and 2020 were North-American H7N3 influenza virus derivatives, which represented ~26% of the 210 HA amino acid sequences prior to the 2013 Asian H7N9 influenza virus outbreaks (Fig. 2C) (Table   229 2). Of note, the hallmark mutation that causes N-linked glycosylation in antigenic site B was observed from 230 Italy/00 (Table 2, blue-color coded and asteroid).   (Fig. 4A). The HAI titer against live Anhui/13 virus showed 248 similar pattern, albiet with lower titers (Fig 4B). The level of cross-HAI reactivity did not directly correlate 249 with the antigenic similarity (Table 1 and Fig. 4). Following challenge with Anhui/13, mice were observed for clinical signs and mortality (Fig. 5).

256
To determine the protection, average body weight loss 5% or less was considered as minimal body weight 257 loss (Dotted line in Figure 5A). Mock vaccinated mice lost greater than 15% body weight by day 7 post- The ROC curve analysis was conducted between HAI titer and protection (body weight loss less 272 than 5%) data following Anhui/13 challenge study (Suppl. Fig. 1). The selection of the cut-off was 273 determined by two criteria: maximizing sensitivity (AUC of the curve) and maximized the summation of 274 sensitivity and specificity (Youden's indect) (21). The highest sensitivity of the prediction was observed as 275 the maximum area under the curve when the VLP HAI cut-off was 1:80 (Suppl. Fig. 1B). The Youden's 276 index (specificity + sensitivity -1) was highest when the HAI cut-off was 1:160 (Suppl. Fig. 1C). Thus, we 277 used the range 1:80 as the cut-off of HAI titer that can provide protection against a stringent challenge by 278 each H7 influenza virus in panel. The absolute protection is expected if the VLP HAI titer is higher than 279 160, while HAI titer between 80-160 is expected to provide marginal protection. When applying the cut-280 offs determined by the ROC analyses, the pre-challenge HAI titer appears to correctly predict the level of 281 protection in weight loss ( Fig. 4A and Fig 5A) in a stringent Anhui/13 challenge. For a comparison of cross-reactive HAI activity, the cut-off 80 was also applied. The HAI 293 antibodies elicited by each H7N9 VLPs had a broad-range of cross-reactive antibodies (Fig 6). The cross-294 reactivity of each antisera did not correlate with the amino acid sequence similarity of the HA (Table 1 and 295 Fig. 6). Mice vaccinated with the four Asian H7N9 strains (Anhui/13, Shanghai/13, Guangdong/16, and 296 Hunan/16) had cross-reactivity to each other (Fig. 6A-D), but did not recognize Jiangxi/09, Italy/00 or 297 Ohio/04 (Fig. 6E-G). Antisera to the Jiangxi/09 or Ohio/04 showed broad cross-reactive HAI activity 298 against all the H7 viruses in the panel, except to Italy/00 (Fig. 6). In contrast, anti-Italy/00 sera had broad

299
HAI activity against all the viruses in the panel, except against Jiangxi/09 and Ohio/00 (Fig 6). Mice 300 vaccinated with NY/02 VLPs elicited antibodies with HAI activity against the homologous NY/02 virus, 301 but did not recognize any of the other H7 viruses (Fig. 6).
302 Figure 6. Cross-reactiveness among H7 panel strains The week 8 sera was tested for the cross-303 reactivity to H7 VLPs expressing HA from all eight panel strains. Individual titer was plotted.

304
Interquartile range, median, minimum and maximum values were presented as box, middle line,

305
upper and lower whiskers, respectively . Dotted line indicates the cut-off for the protection (80 HAI 306 unit).

308
With regard to the unique antigenic profile of Italy/00, we found that there was a putative 309 glycosylation site at HA 169 (H7 numbering from our own sequence alignment) ( Table 2). Since the location 310 of putative N-linked glycosylation was located in antigenic site B, we hypothesized that glycosylation at 311 this location may be responsible for the unique antigenic profile of Italy/00. To test the hypothesis, we 312 introduced a mutation into the HA nucleotide sequence of Anhui/13 (HA A169T) (Fig. 7A)and looked for 313 the change in reactivity elicited antisera by each VLP vaccine (Fig. 7B). Interestingly, the reactivity of VLP 314 expressing the Anhui/13 HA A169T mutation elicited antibodies with a significant decrease in HAI activity 315 against Anhui/13 and Hunan/13, but no change against the other 6 viruses (Fig. 6B). According to the 316 predicted trimeric structure (Protein data base number=4N5J), the glycosylation site appear to be located 317 on the antigenic site B, and next to the receptor binding site (Fig. 7C).

331
We also immunized C57B/L6 mice with the Anhui/13 A169T VLPs and looked for the antigenic 332 breath of the antisera and protection efficacy against Anhui/13 WT H7N9 challenge (Fig. 8). Interestingly,

333
the HAI titer to the Anhui/13 A169T VLPs (homologous antigen) was significantly lower and showed 334 bigger standard deviation than the HAI titer to the Anhui/13 WT ( Fig. 7C and 8A). The HAI activity to the 335 Shanghai/13 VLPs was similar with the titer to the Anhui/13 A169T VLPs (Fig. 8A). High reactivity to the 336 New York/02 VLPs (Fig. 8C), which was also observed from other antisera for all 8 panel strains (Fig. 6).

337
The HAI reactivity to the Hunan/16, Guangdong/16, Jiangxi/09, Italy/00, and Ohio/03 H7 VLPs was 338 significantly lower than the titer to the Anhui/13 WT and New York/02 H7 VLPs. In consistent with the 339 high HAI titer to the Anhui/13 WT H7 VLPs, the mice were completely protected from weight loss and 340 onset of any clinical symptom by the lethal challenge with the Anhui/13 WT H7N9 virus (Fig. 8C&D).

341
There was no detectable infectious viral titer in the lung collected at day 4 post challenge, which was clearly 342 contrasted with the naïve control mouse (Fig. 8B).

353
This study investigated the antigenic differences of selected H7 panel influenza HA proteins. Since 354 most available H7 HA sequences originated from major human infections, the selected H7 panel strains

355
were similar with the list of candidate vaccine viruses (CVVs) from the WHO (10). There was a high 356 similarity of amino acid sequences in the putative HA antigenic sites (Table 2). In addition, antibodies 357 elicited by these HA antigens had HAI activity to most of these H7 viruses (Fig. 6) figure 2). The unique structure of HA appear to ease the binding of antibodies 373 from other antisera, while the antisera for the H7N2 was lack of major epitope. Meanwhile, The HAI titer 374 against Italy/00 and Ohio/04 VLPs was observed low from all antisera, even to the homologous antisera.

375
Only anti-Italy/00 antibodies against Italy/00 VLP were above the cut-off, and only anti-Ohio/04 antibodies 376 against Ohio/04 VLPs were above cut-offs. It seemed that in comparison to other VLPs, the access to the two VLPs were much restricted. The presence of glycosylation on the receptor binding site also 378 significantly impair the reactivity to the homologous antisera; even the antisera collected from mice 379 vaccinated with the Anhui/13 A169T H7 VLPs detected the Anhui/13 WT H7 VLPs better (Fig. 8A). We

380
can explain that the Italy/00 has glycosylation site near the receptor binding site, so even homologous 381 antisera showed relatively lower access to the VLP. We could not find plausible explanation for the Ohio/04

382
VLPs, but suspect that the structure of Ohio/04 expressing VLP might have hindered the access of the 383 antibodies.

384
The level of cross-HAI activity among H7 HA proteins did not follow phylogenetic similarity or 385 geographic origin. Instead, mutations that altered the glycosylation pattern around the receptor binding site

386
(RBS) played a critical role in shaping the antigenic profile. A single amino acid substitution (HA A169T) 387 caused a significantly reduce the reactivity to antisera specific for Asian H7N9 strains. The mutation did 388 not significantly influence on reactivity to other anti-sera, which suggests that such antigenic site was not  Table 2). The putative location of the N-glycosylation is adjacent to the receptor binding site of 392 the trimeric form of HAs (Fig. 6C). Spontaneous occurrence of the N-linked glycosylation sites at the same 393 location in H7 HA proteins was previously reported during the H7N1 epidemics in Italy in the early 2000's

394
(24). The study used reverse genetics to generate virus which has the corresponding mutation A149T

395
(A169T by our numbering) and showed that the single mutation alone resulted in glycosylation by 396 electrophoresis(24). Also, the mutation was spontaneous and stable during the passage of the H7N1 viruses 397 in turkeys, which suggests that the mutation can naturally occur during circulation in poultry species (24).

398
There was no significant influence of the glycosylation site on host tropism, however, the potential change 399 in antigenicity was not investigated (24). The latest study published in 2020 also verified that the 400 corresponding mutation A151T (A169T by our numbering) occurred in one of the escaping mutants and 401 proved that the mutation results in glycosylation (25). But both studies did not investigate its influence on to thank the University of Georgia Animal Resource staff, technicians, and veterinarians for animal care and the staff 452 of the Animal Health Research Center (AHRC) Biosafety Level 3 laboratories for providing biosafety and animal 453 care. Also, the authors thank the members of the CVI protein production core, Jeffrey Ecker, Spencer Pierce, and