The cohort, including 203 children 5–17 years old and 413 adults 18–59 years old, was enrolled in the pre– and post–flu season of 2018/2019. The overall infection incidence based on serological tests was reported in our previous work (Xu et al., 2021). The research protocol was approved by the institutional review board of the National Institute for Viral Disease Control and Prevention, Chinese Centers for Disease Control and Prevention, and written informed consent was obtained from the participants. Among the 203 children 5–17 years old and 413 adults 18 to 59 years old, there were 19 children and 17 adults with a ≥ 4-fold rise in MN Abs against A/Singapore/INFIMH-16-0019/2016(H3N2, SN16/16). The titer of < 10 will be assigned as a value of 5 for statistical purposes.

MDCK (Madin–Darby canine kidney)–SIAT1 cells stably transfected with human CMP-N-acetylneuraminate:β-galactose α-2-6-sialyltransferase overexpress the α-2-6–linked sialic acid receptor compared with MDCKs (Hatakeyama et al., 2005). The MDCK-SIAT1 cells are kindly gifted from the US Centers for Disease Control and Prevention. MDCK and human embryonic kidney (293T) cells were obtained from the American Type Culture Collection. The cells were cultured in Dulbecco modified eagle medium (Invitrogen, Carlsbad, CA, United States) supplemented with 10% fetal bovine serum (Invitrogen) and glutamine (2 mM; Invitrogen).

The influenza virus gene segments were amplified by reverse transcription–polymerase chain reaction and cloned into a modified version of the bidirectional expression plasmid pHW2000. The reassortant H6N2 viruses bear the HA of A/Taiwan/1/2013(H6N1) and NAs of the tested viruses, including SN16/16(H3N2, 3C.2a1), A/Hongkong/2671/2019 (H3N2, 3C.2a1b1b, HK2671/19), A/Kansas/14/2017(H3N2, 3C.3a, KS14/17), A/Brisbane/10/2007(H3N2, clade2, Br10/07), A/Switzerland/9715293/2013(H3N2, 3C.3a, SWZ/13), A/Hongkong/1968(H3N2, HK/68), or A/Hongkong/33982/2009(H9N2, HK/09), respectively, and the internal genes of A/Puerto Rico/8/1934(H1N1) (PR8). The H6N2 viruses were rescued by the reverse-genetic (RG) technology with the cotransfection of eight bidirectional plasmids. The viruses were propagated in 9–11-day-old embryonated chicken eggs and stored at −80°C until use. The single or two mutations, N329T, E344K, or T329N, K344E, were introduced into the corresponding pHW2000 plasmid containing the NA segment using appropriate primers and QuikChange™ Site-Directed Mutagenesis Kit (Stratagene, La Jolla, CA, United States), following the manufacturer’s instructions. All plasmids were sequenced to ensure the presence of the introduced mutations and the absence of undesired mutations.

One hundred TCID50 SN16/16 H3N2 virus was precultured with serially diluted sera for 1 h at 37°C and then transferred to the MDCK-SIAT1 monolayer cells in 96-well plates for 18 h at 37°C in a 5% CO₂ incubator. As reported previously, the cells were processed to be stained by anti-NP (Millipore, Burlington, MA, United States) (Rowe et al., 1999). The IC₅₀ of the tested sera was defined as the reciprocal of the last dilution that caused a 50% reduction of OD compared with the virus control.

As previously reported (Eichelberger et al., 2016; Yu et al., 2017), the tested RG H6N2 viruses were on fetuin-coated 96-well plates for 18 h at 37°C. After the plate was washed, horseradish peroxidase (HRP)–conjugated peanut agglutinin (Sigma–Aldrich, St. Louis, MO, United States) was added. The galactose was exposed by removing the sialic acids, and the fetuin plate was detected by tetramethylbenzidine and measured at 450 nm. The maximum signal within the linear range of the titration curve was selected. The heat-treated sera were serially diluted and then incubated with the tested RG viruses and were performed as the above steps. The IC₅₀ of the serum was analyzed by using GraphPad Prism software, version 5.

Recombinant H3 of SN16/16 and N2 of A/Babol/36/05(H3N2, Babol36/05) and KS14/17 were purchased from Sino Biological Inc., China. The 96-well plates were coated with 50 ng per well of the proteins and incubated with serially diluted sera. The bounded Abs were detected using the rabbit anti-human immunoglobulin G–HRP (Sino Biological Inc.), developed by tetramethylbenzidine, and stopped by 2 M H₂SO₄, and OD was measured at 450 nm. The titer of the tested sera was defined as the reciprocal of last dilution with an OD value of threefold compared with the blank control.

The HA and NA gene sequences of the H3N2 viruses from January 2014 to December 2021 in the Global Initiative on Sharing All Influenza Data (GISAID) database were downloaded. The Bioedit 7 program was used to align and analyze amino acid residues. The clade was classified as HA clade.

All graphing and statistical analyses were carried out using R software (version 4.1.2)¹ and GraphPad Prism software (version 5). The Ab titer and the fold-increase values were used for the response descriptions. Data are expressed as mean ± 95% confidence interval (CI). Mann–Whitney U-test was used for comparing differences between the values of children and adult groups. Wilcoxon signed rank test was used for the paired data of children or adult group. Spearman rank correlation analysis was performed on the fold increase of Ab titers. The probability p < 0.05 was considered to indicate statistical significance and marked with one asterisk; p < 0.01 was marked with two asterisks.

We performed a seroepidemiological investigation on the paired sera using the vaccine strains (Supplementary Table 1). The traditional HI is for H1N1 and type B influenza virus. Besides the data obtained by optimized HI in the presence of oseltamivir carboxylate, we also used MN by anti-NP staining for the H3N2 virus. The assay was based on the infection and 1–2-life-cycle viral replication in which the surface glycoproteins, HA and NA, and viral RNP complex were all involved. The numbers of MN seroconverts were 17 adults (21–59 years old) and 19 children (7–17 years old), respectively. The Ab titer and fold increase of postseason sera showed no difference (136.8 vs. 111.8; 10.32 vs. 9.30) between children and adults (Figures 1A,B). As some seroconverts against SN16/16 were also seroconverting to type B or A(H1N1) pdm09, we assayed the specific HA-recognizing of the antisera. A significant correlation between the MN Abs and H3-binding Abs was found in children (r = 0.54, *p = 0.02; Figure 1C). In contrast, no correlation was found in adults (r = 0.30, p = 0.24; Figure 1D). Those coincident data support the serodiagnosis for H3N2 infection in children cases.

The NA of influenza virus cleaved sialic acid linkage and aided virus release during infection. Moreover, NA was another important antigen for host immunity. We expected the possible use of anti-NA Abs for the serological tests of the H3N2 virus other than the modified HI assay. Because of the steric interrupting of anti-HA Abs (Chen et al., 2019; Kosik et al., 2019), a panel of recombinant H6N2 viruses bearing the HA of group 1 HA and the tested N2 was used in our experiments. As expected, NI seroconversion targeting SN16/16 N2 was detected in both adults and children infection cases. In children, the Ab titer of NI was approximately 129 and was significantly higher than in adults (73.47, **p < 0.01; Figure 2A). The conversion rates in children were 68.42 and 41.17% in adults (Figure 2B). To explore the role of NI Abs, we analyzed the correlations between MN Abs and NI. A significant correlation was found in children (r = 0.59, **p < 0.01; Figure 2C), not in adults (r = 0.11, p > 0.05; Figure 2D). A possible role of protecting NI Abs in children suggested that NI assay could be applied to diagnose H3N2 infection and functional evaluation in children.

We compared the phylogenetic tree of N2 (Supplementary Figure 1). Supplementary Figure 1 (the phylogenetic tree) suggests that N2 subtype influenza viruses could be classified into two prominent clades, corresponding to avian and mammalian (human/swine) influenza viruses. The phylogenetic relationships were largely consistent with the previous study (Fanning et al., 2000). The genetic distance of N2 of 1999–2004 showed a minor difference (Sandbulte et al., 2011). Clade 3 H3N2 viruses have been prevalent since 2010 and cocirculated with (H1N1) pdm09 (Falchi et al., 2011). The clade 3C.2a was dominant and chronologically evolved into 3C.2a1b2a of 2021. Then, we compared the epitopes of the NAs from 1968 to 2021 (Supplementary Table 2). The antigenic epitopes at the head of N2, including D’197–199, F’329–339, G’344–347, I’368–369, and K’400–403, as well as the regions of E’302–308, H’356–358, J’384–386, and L’463–468, were analyzed. Several amino acid substitutions were found in the earlier 3C.2a H3N2 virus from those circulated subsequently. A striking difference was found in N2 of SN16/16 (3C.2a1) and KS14/17 (3C.3a), N-glycosylation at 329, and E at 344 of SN16/16 versus non-glycosylation at 329 and K at 344 of KS/14/17. A detailed analysis was performed for the 3C.2a and 3C.3a H3N2 viruses from January 2014 to December 2021 (Table 1). The percentage of viruses with the epitopes of N-glycosylation and E344 decreased pronouncedly, although the vaccine strain HK2671/19 of 2020/21 flu season bore non-glycans at 329 and E344 (Supplementary Table 2). Then, the epitopes of glycan loss in 329 and E344K predominated in 3C.2a1b2a, and its representative virus was A/Darwin/9/2021 (H3N2). Unexpectedly, the N-glycosylation on 329 in the 3C.2a1b2a H3N2 viruses with K344 reemerged after September 2021, implying those mutations related to viral fitness and evolution in humans.

Herein, we tested the cross-reactivities of the sera to a serial of N2. We measured the NA-binding and NI Abs. Similar trends were found in children and adults (Figure 3 and Supplementary Figure 2, respectively). The antisera cross-bound with 2005N2(Babol36/05) and 2017N2(KS14/17) and cross-inhibited the N2 activity of 2007N2(Br10/2007) and 2013N2(SWZ/13), even the avian-origin N2 such as 1968N2 and 2009H9N2. We discovered a significant correlation between serum anti-SN16/16 MN Abs and KS14/17-binding Abs in children (r = 0.52, *p = 0.02; Supplementary Figure 3). Furthermore, the associations exist between anti-SN16/16 NI and Babol36/05-binding (r = 0.74, ***p < 0.001) or KS14/17-binding Abs (r = 0.67, ***p < 0.001) but not between anti-KS14/17 NI and KS14/17-binding Abs (r = 0.3, P = 0.26) (Figure 3). The discrepancy between anti-KS14/17 NI and N2-binding Abs suggests the conformational structure of NA enzyme was recognized by Abs in NI assay, which is a functional evaluation of their interactions. The information generally confirmed the contribution of anti-NA Abs to host protective immunity, and the wide spectrum of response to older NAs was related to the conserved property of NA. For instance, the D’197-199 was highly conserved, as shown in Supplementary Table 2. We speculated that the varied phenotype of NI against SN16/16 and KS14/17 is due to the genetic mutations mentioned previously.

Then, we proceeded with the in vivo testing using recombinant H6N2 variant (SN16/16 + E344K) with a single E344K mutation in SN16/16 N2. This variant had an epitope of N-glycosylation at 329 and E344K. The antisera from children’s cases displayed a significantly lower reactivity to the variant than SN16/16 (Figure 4A, ***p < 0.001). On the other hand, we generated N2 mutants with the KS14/17 NA backbone with one or two combinational mutations that had an epitope of N-glycosylation at 329 or E344 or both. Only a single mutation, 329 glycans or K344E, could upregulate the reactivation of children’s antisera to KS14/17, while the effects of the combinational mutations showed a moderate rise, implying a possible local fitness in the antigenic region (Figure 4B, ***p < 0.001). The NI activities of children antisera to KS17 variants were enhanced, although the NI titers did not reach the significantly comparable level as that to SN16/16 (Supplementary Table 4). We also discovered the children’s antisera demonstrated similar NI reactivity to HK2671/19 virus, bearing 329 non-glycans and K344E, as to SN16/16. Those differential NI activities suggest that other mechanisms might be involved. We also tested those variants using the ferret antisera challenged by the wild-type H3N2 viruses (Supplementary Table 3). The NA antigen ratio of SN16/16 and KS14/17 was greater than 1.5 and demonstrated antigenicity difference between the two NAs. Limited by the available volume of human sera, we performed the NI tests on SN16/16 and KS14/17 variants using ferret antisera. One-way interaction data suggest some differences among the NA variants compared with wild-type NA. Similar enhanced NI activities to KS14/17 variants bearing T329N and/or K344E were found. Of interest, the mutation N329T of SN16/16 also enhanced the NI activity of ferret sera and reversed the downregulation effects of E344K, implying complex interactions presented among antigenic sites and host Ab profile. The findings support that the amino acids in sites 329 and 344 are the key determinants for NA antigenicity.