For our study, we enrolled 141 persons who had close contact with poultry at 5 chicken farms that had influenza A(H5N2) outbreaks in chickens during January–March 2012. These contacts were 15 farm workers, 90 animal health officials, and 36 temporary employees who participated in culling of infected chickens; no symptoms of influenza-like illness occurred in these persons within 1 week after culling. All 15 poultry workers had been working at their poultry farms for >6 years, and most of the animal health officials had experience in stamping out infected poultry. However, for the 36 temporary employees, previous contact histories with infected chickens were unknown.

Throat swab specimens were collected from all contacts for virus detection within 7 days from the beginning of exposure to the virus, and paired serum samples were collected 21 days apart for serologic testing. Complete testing methods are described in the Technical Appendix. Participants were offered an inactivated influenza A(H5N1) vaccine on a voluntary basis on the day the first serum specimens were collected, and vaccination histories within 1 year before the specimen collection date were recorded through oral questionnaires. A total of 102 (72.3%) of the 141 participants were vaccinated: 22 (15.6%) received influenza A(H5N1) only; 39 (27.7%) received seasonal influenza vaccine only; 41 (29.0%) received both vaccines; and 39 (27.7%) received neither (Table 1).

We found all swab specimens were negative for influenza viruses by real-time reverse transcription PCR. However, hemagglutination inhibition (HI) and/or microneutralization (MN) test results showed 7 persons had antibody titers >80 for subtype H5N2 virus; 6 of these persons showed seroconversion for the virus (Table 2). Elevated antibodies against subtype H3N2 or H5N1 viruses were detected in some of the 6 persons who showed seroconversion (Table 2), which suggests that positive results for subtype H5N2 could be the result of cross-reactive antibodies from previous influenza vaccinations or infections. All 6 persons who showed seroconversion for influenza A(H5N2) virus had received vaccinations for influenza A(H5N1) and seasonal influenza (Table 2). Further, persons who received both influenza vaccinations showed a significant (p = 0.001) geometric mean titer increase in HI antibody against influenza A(H5N2) virus in the second samples, whereas those who did not receive both vaccinations did not show a similar increase. This finding indicates these heterologous vaccinations may influence HI antibody titers against influenza A(H5N2) virus.

To investigate whether the influenza A(H5N2) antibodies were elevated as a result of exposure to that virus or because of vaccination with heterologous influenza viruses, we determined antibody levels to influenza A(H5N2) nonstructural protein 1 (NS1) (12–15). The NS1 protein is not readily incorporated into virions used to make inactivated influenza vaccine, so a response to NS1 protein would indicate active influenza A(H5N2) infection. Paired serum samples were analyzed by ELISA plates coated with 2 peptides, NS1^36–48 (LRRDQKSLRGRGS, NS1-pA) and NS1^204–225 (RSSNENGGPPLTPKQKREMART, NS1-pB), synthesized on the basis of the NS1 protein sequences of influenza A(H5N2) virus (NS1-pA, A/chicken/Taiwan/1209/2003) and influenza A(H3N2) virus (NS1-pB, A/Taiwan/4055/2009), respectively. The NS1-pA of the 2012 influenza A(H5N2) outbreak strain has an S48N substitution that is not found in the 2003 strain.

For controls, we simultaneously analyzed 3 groups of paired serum samples with seroconversion (data not shown): 1) samples from 7 ferrets infected with different influenza virus strains (H1N1, n = 3; H3N2, n = 1; H5N1 [A/Vietnam/1194/2004], n = 1; and A[H1N1]pdm09, n = 2); 2) samples from 8 persons infected with influenza A(H1N1)pdm09 virus; and 3) samples from 9 persons who received vaccinations against influenza A(H5N1) virus. The resulting NS1 antibody responses were plotted (Figure). Five (71.4%) of 7 ferrets showed positive NS1 response against NS1-pA and all against NS1-pB (Figure, panel A), which indicates that influenza virus infection can cause a measurable anti-NS1 response after virus challenge. For influenza virus–infected persons, 3 (37.5%) of 8 showed responses against NS1-pA and NS1-pB (Figure, panel B), but for vaccinated persons, 1 (11.1%) of 9 showed responses against NS1-pA and none against NS1-pB (Figure, panel C). These patterns suggest that anti-NS1 response elicited by natural infection is stronger than that induced by vaccination.

For the group of 7 contacts we identified who had elevated influenza A(H5N2) antibodies, 2 (contacts 1 and 3) had positive NS1 antibody response against both peptides; the remaining 5 did not (Figure, panel D). These results suggest that contact 1, a poultry worker, and contact 3, an animal health official, may have experienced recent influenza infections.

To better establish the validity of using NS1 to distinguish infected from vaccinated persons, we analyzed paired serum samples for the 134 persons who did not show seroconversion for influenza A(H5N2) virus. Of these, 5 (3.7%) showed positive NS1 antibody response against NS1-pA and 12 (9.0%) against NS1-pB (Figure, panel E). This result suggests that an NS1-ELISA should not be used alone to determine influenza infection but can provide additional data to validate the results of protein-based serologic assays.

In this study, we sampled 141 persons exposed to poultry infected with influenza A(H5N2) virus to assess virus shedding and used multiple serologic assays (including a novel NS1 ELISA) to determine seroconversion status. We found that 6 (4.3%) persons had elevated HA antibodies detected by HI and/or MN assays; a lower percentage (1.4%, 2/141) of subclinical infections was suspected after validation by NS1 antibody assays. The NS1-peptide B was designed on the basis of influenza A(H3N2) virus; however, it also reacted with antibodies elicited by viruses of different subtypes, which suggests that consensus residues may play an essential role in forming the epitope of NS1 protein.

Our study has limitations. Patient histories of exposure to avian influenza viruses and influenza vaccination were given orally and thus may not be accurate, and mismatching between circulating viruses and antigens used in the study may have occurred. Also, recent seasonal influenza infection may interfere with the determination of subclinical infection with influenza A(H5N2) virus because the NS1 protein is remarkably conserved in type A influenza viruses.

Cross-reactive antibodies in humans elicited from heterologous influenza viruses can complicate serologic, HA-based identification of influenza subtype. The NS1-ELISA method we describe may help determine the type more readily and improve diagnosis of subclinical infection in humans. Further, our findings indicate that occupational exposure to infected poultry may pose a risk for infection in humans.

Expanded materials and methods for study of influenza A(H5N2) virus antibody seroconversion in humans after contact with infected poultry, Taiwan, 2012.