Cases of H5 highly pathogenic avian influenzas (HPAI) are on the rise. Although mammalian spillover events are rare, H5N1 viruses have an estimated mortality rate in humans of 60%. No human cases of H5 infection have been reported in Malaysian Borneo, but HPAI has circulated in poultry and migratory avian species transiting through the region. Recent deforestation in Malaysian Borneo may increase the proximity between humans and migratory birds. We hypothesise that higher rates of human-animal contact, caused by this habitat destruction, will increase the likelihood of potential zoonotic spillover events. In 2015, an environmentally stratified cross-sectional survey was conducted collecting geolocated questionnaire data in 10,100 individuals. A serological survey of these individuals reveals evidence of H5 neutralisation that persisted following depletion of seasonal H1/H3 binding antibodies from the plasma. The presence of these antibodies suggests that some individuals living near migratory sites may have been exposed to H5. There is a spatial and environmental overlap between individuals displaying high H5 binding and the distribution of migratory birds. We have developed a novel surveillance approach including both spatial and serological data to detect potential spillover events, highlighting the urgent need to study cross-species pathogen transmission in migratory zones.
Article
Serological analysis in humans in Malaysian Borneo suggests prior exposure to H5 avian influenza
https://doi.org/10.21203/rs.3.rs-4021361/v1
This work is licensed under a CC BY 4.0 License
Journal Publication
published 17 Oct, 2024
Version 1
posted
You are reading this latest preprint version
Cases of highly pathogenic avian influenza (HPAI) are on the rise in poultry and wild bird populations. Of particular concern is the influenza A(H5N1) clade 2.3.4.4b, which has been responsible for a striking increase in deaths in wild birds and domesticated poultry since 20201. The World Organisation for Animal Health estimated that in 2022 at least 131 million domestic poultry succumbed to the virus or were culled due to exposure to the virus2.
Migratory shorebirds and waterfowl are known hosts for avian influenza viruses (AIVs), including H5N13,4. The global movements of these species can facilitate the dispersion and increased genetic diversity of the virus. Climate change and extreme weather events may be one cause of the recent increased pervasiveness of H5N14,5. Habitat destruction resulting from these environmental changes influences the distribution of migratory species, which in turn impacts the dynamics of viral spread4–6. As habitats and food sources become increasingly scarce, there is potential for higher contact rates amongst different migratory avian species and between wild and domesticated birds. This increased contact creates more opportunities for viral reassortment and for the virus to jump between hosts.
In 2022 and 2023, several notable spillover events into mammals were caused by 2.3.4.4b H5N1 viruses7, including seals8, mink9, and sea lions10. There is no evidence of sustained community transmission amongst humans or other mammals, with the number of human cases remaining limited2,11. As human cases of H5N1 are often clinically severe with a high mortality rate, there are concerns that the virus will reassort to enable community transmission1. To prevent further viral spillover, it is crucial to understand current rates of HPAI H5N1 exposure. If there are subclinical or non-severe cases of H5N1, the symptoms of such cases would likely be indistinguishable from that of other influenza subtypes, so exposure to the virus would remain undetected unless an infected individual receives medical care. Therefore, the global prevalence of H5N1 exposure in humans is presently unknown and is likely to be underestimated12. Consequently, there is an urgent need to study serological exposure to these HPAI H5N1 viruses in humans.
We investigated H5N1 in humans in Malaysian Borneo, due to a unique confluence of environmental and ornithological factors in the region. Borneo is an important migratory stop for wild shorebirds transiting along the East-Asian Australasian flyway, which spans from Russia to Australia. Malaysia contains 13 important stopover sites for 20 species of migratory shorebirds, with four such sites in Malaysian Borneo13. Although Malaysian Borneo has extensive coastline, it is also a region which has experienced much environmental change in the last 50 years14,15. Habitats which would usually be home to these migratory birds including freshwater swamp forests16, mangrove forests17, and peat swamp forests18 have been cleared and logged to make room for industrial agricultural efforts13.
Several of the migratory species known to transit through Malaysian Borneo are carriers of various influenza subtypes, including H5 viruses19–21. However, the only recorded outbreak of HPAI H5N1 in Malaysian Borneo to date was a clade 2.3.2.1c virus outbreak in a commercial poultry farm in 2018 in the Sabah state22. An estimated 30,000 birds were either culled or succumbed to the virus during a month-long period23. Although this poultry outbreak was severe, there have been no human cases of H5N1 reported in Malaysian Borneo.
Here, we perform a serological survey of human influenza exposure in Sabah, Malaysian Borneo to examine the immunological footprint of H5N1 in the region. We present a method for minimizing the impact of influenza subtype cross-reactivity on these serological results. We define species distributions of domesticated poultry and migratory wild birds and demonstrate that environmental covariates can be used as proxy to model wild bird contact. We additionally identify shared spatial distributions and environmental risk factors between the presence of migratory shorebirds and clade-specific H5N1 seroprevalence using a Bayesian framework. This study highlights the need to increase surveillance for rare zoonotic diseases at migratory sites and presents an approach for modelling the distributions of serological results and reservoir species.
Identification of heterogeneous patterns of influenza binding and pseudo-neutralisation
In 2015, 10,100 individuals from 2,650 households in Northern Sabah were sampled24. The surveyed population age ranged from 3 months to 105 years and was 47% male (n = 4776) and 53% female (n = 5324)25. The locations of the sampled villages and shorebird sightings throughout Sabah are displayed in Fig. 1. 14% (n = 359) of these households were situated within 10km of wild bird sightings. 51% of the population are poultry owners (n = 5137), mostly evenly distributed throughout the entire sample region25.
2,000 of these individuals were initially randomly selected for serological testing of influenza exposure by enzyme-linked immunosorbent assays (ELISAs) which measure antibody (IgG) binding to HA. 500 individuals were selected from each of the four geographic regions sampled in the study: Kudat, Pitas, Ranau, and Kota Marudu. The population was exposed to two potential sources of avian influenza spillover: domesticated poultry and wild birds. 21% (n = 419) of the randomly selected individuals lived within 10km of a migratory shorebird and 48% (n = 951) were poultry owners. These categories were used to examine differences in the binding, and later, neutralisation of H5 viruses.
We used four hemagglutinins (HAs) as our viral antigens for ELISAs. Hemagglutinin is a glycoprotein in influenza and the dominant immunogen. Antibody binding was reported in absorbance units (AU), where higher AU indicates increased antibody binding to the HA.
Two HAs were used in our ELISAs to capture binding to the subtypes of seasonal viruses which would have been circulating prior to sample collection in 2015: H1N1 (A/California/07/pdm2009) and H3N2 (A/New York/55/2004). Two H5 antigens were used to examine binding to avian viruses. A/Duck/Laos/3295/2005(H5N1) belongs to the H5 clade 2.3.4. This clade was circulating in Asia from 2003 to 2013 in avian species before breaking off into sub-lineages26. A/snow goose/Missouri/CC15-84A/2015(H5N2) belongs to clade 2.3.4.4 which did not become the predominant avian flu clade in Asia until 201827. There are 38 amino acid differences between these two viral HAs (approximately 7% heterogeneity) (Supplementary Table 1).
Seasonal influenza binding was evenly distributed throughout the population (Fig. 2a). Mean H3N2 binding was 1.437(± 0.035) AU and there were no differences when the population was categorized by poultry ownership or proximity to wild birds. There were a wide variety of H3N2 responses with 23 above 3 AU, 517 above 2 AU, and 1414 above 1 AU. H1N1 binding was not as strong in the study population (mean of 0.3287(± 0.0109) AU). Poultry owners within the population had a slightly higher mean binding to the H1N1 HA than non-owners, 0.31 versus 0.29 AU (p = 0.0382) by a two-tailed Mann-Whitney test. Two responses were above 2 AU and 22 were above 1 AU.
There were no significant differences (p < 0.05) in H5 binding to either hemagglutinin when the population was sorted by poultry ownership. However, those living within 10km of wild birds had a statistically significantly higher mean than those living further away for both H5 antigens, 0.25 versus 0.21 AU (p = 0.0019) for the 2.3.4 pre-2015 virus and 0.12 versus 0.11 AU (p = 0.0204) for the post-2015 2.3.4.4 virus. Responses to A/snow goose/Missouri/CC15-84A/2015(H5N2) were modest with one response above 2 AU and only 10 above 1 AU. A/Duck/Laos/3295/2005(H5N1) was a more reactive antigen with one response above 3 AU, 20 above 2 AU, and 161 responses above 1 AU. High binding to one antigen did not correlate with similar binding to a different antigen. The Spearman’s rank correlation coefficient (r) of binding to H5 antigens was r = 0.22 and the other correlations ranged from r = -0.03 to 0.13 (Supplemental Fig. 1). These Spearman r values indicate that the binding to one ELISA antigen was not linearly related to binding of a different antigen.
204 samples were selected for follow-up pseudotyped microneutralisation assays (PNA) against two viruses from pre-2015: A/Indonesia/05/2005(H5N1) and A/Bar headed goose/Qinghai/1A/2005(H5N1). We generated pseudotyped lentiviruses which displayed hemagglutinin on their surface but were replication-deficient28. This allowed us to study the ability of our cohort plasma to inhibit viral growth by binding to hemagglutinin without the risk of working with live virus.
The 20 samples which exhibited binding above 2 AU to A/Duck/Laos/3295/2005(H5N1) were chosen for inclusion in these follow-up experiments, and the rest were randomly selected based on the quantity of available pseudovirus. 52% (n = 108) of the population tested for neutralisation were poultry owners and 34% (n = 69) lived within 10km of a shorebird sighting. The neutralising abilities of a sample were evaluated based on NT50 value, which is the 50% inhibitory concentration (i.e. the concentration of serum or plasma necessary to achieve 50% virus inhibition). Previous studies have used 1:100 as a cut-off for seropositivity28, but we opted to take the highest NT50 from a cohort of negative control samples (1:1173 for A/Indonesia/05/2005 and 1: 1015 for A/Bar headed goose/Qinghai/1A/2005) to create a cut-off, as in Thompson et al.29 (Supplemental Fig. 2).
Three samples were considered seropositive in their neutralising ability (1.5% seropositivity rate) against A/Indonesia/05/2005(H5N1), of which two (67%) were poultry owners and all three (100%) lived within 10km of wild bird migratory sites (Fig. 2b). The samples were 6.8% seropositive (n = 14) against A/Bar headed goose/Qinghai/1A/2005(H5N1). Six of these seropositive individuals owned poultry and all 14 lived in proximity to wild birds. When comparing differences in the mean NT50, there were no significant differences based on poultry ownership to either virus. Individuals with proximity to wild birds did have a statistically significant increase in mean NT50 over those living further away, 210.4 versus 59.24 (p < 0.0001) by a Mann-Whitney test.
Persistence of pseudo-neutralisation upon depletion of cross-reactive antibodies
We developed a bead-based antibody depletion assay to determine the role of cross-reactivity in these neutralising responses. We conjugated magnetic non-fluorescent beads to several HA for seasonal influenza strains (Supplemental Fig. 3). The beads were then incubated with the plasma. Using a magnet to pull down seasonal influenza antibody-bound beads, unbound antibody can be extracted. This plasma extract is depleted of cross-reactive seasonal influenza responses and was tested in neutralisation assays for a maintained H5 response. Neutralisation assays were repeated with the treated plasma against A/Bar headed goose/Qinghai/1A/2005(H5N1). There were nearly five times as many neutralizing samples against A/Bar headed goose/Qinghai/1A/2005(H5N1) than A/Indonesia/05/2005(H5N1), so only the first virus was repeated to preserve plasma volume.
Pooled positive control serum was obtained from a vaccination trial (vaccinated with A/Indonesia/05/2005(H5N1)) from BEI Resources (NIH, NIAID). The two negative control samples were low and high titre human H1N1 convalescent sera and were assumed not to have been exposed (i.e. the observed neutralisation is likely due to cross-reactive antibodies). The high and low tire responses are relative to each other according to the supplier (BEI Resources). Neither positive pool experienced a statistically significant change in NT50 before and after treatment (plasma dilutions of 1:1915 vs. 1:1857 and 1:2436 vs. 1:2166) (Fig. 3a). The first negative sample (low titre) experienced a 72% reduction in NT50 after treatment (1:595.1 reduced to 1:163.7), which was considered statistically significant (p = 0.0010) via a sum-of-squares F Test comparing the logNT50 values. The NT50 for negative control 2 (high titre) experiences a 74% reduction upon treatment with the beads, from 1:1850 to 1:486.8, which was also statistically significant (p < 0.0001).
Only two of the fourteen original high neutralisers (NT50 > 1000) from our sample cohort could be tested via this method (due to limited plasma volume), along with six low neutralisers where the NT50 was above 100 but below 1000. Neither the high nor low neutralisers experienced a statistically significant reduction in NT50 after treatment with the beads (Fig. 3b). This suggests that the neutralising response measured in these Malaysian Borneo samples may be H5-specific and not a cross-reactive neutralising response stemming from seasonal influenza exposure.
Spatial and environmental overlap of H5 binding and wild bird distributions
From July 1979 to August 2023, there were 86,502 observations of the 20 species of shorebirds known to migrate through Malaysian Borneo in the Sabah region13,30. These observations were submitted to eBird from 7,636 unique entries. The cumulative abundance of each species ranged from 2 to 28,631 observations (Table 1). The distributions of shorebirds were included across the whole-time frame of data availability (1979–2023), rather than just the year of sample collection (2015), to model species distribution and reduce year-to-year variability.
There are pockets of high and low domesticated poultry ownership distributed throughout the Northern Sabah region, while wild bird contact and high H5 binding are confined to several distinct areas (Fig. 4a, Supplemental Fig. 5). There is a clear section of overlap between wild bird contact and high H5N1 binding in the north-western portion of our sample region, in a district known as Kudat. This high H5N1 binding cohort is n = 20 and have an AU > 2. An AU > 2 was considered high binding as the highest response in the control cohorts had AUs of 1.28 (urban Malaysia cohort) and 0.47 (Scottish control cohort) (Supplemental Fig. 4). ELISA binding was chosen for this spatial analysis over neutralisation, as this allowed us to include data on binding at 2,000 locations.
Distributions of wild bird contact and H5N1 binding (by ELISA to A/Duck/Laos/3295/2006) were then plotted as probabilities of species distribution or antibody binding (Fig. 4a). We identified environmental risk factors for shorebird distribution and ELISA binding, reported as adjusted odds ratios based on generalized linear mixed effects modelling (Fig. 4b). The model of H5 binding (Supplemental Table 2) and of the wild bird distribution (Supplemental Table 3) had several shared environmental risk factors (Fig. 4b). Closeness to the sea increases the odds of wild bird contact by nearly 10 times and the odds of high H5 binding over 800 times. Both H5 binding and wild bird sightings were more likely in areas with lower elevation and lower precipitation. Proximity to forested areas increased the odds of wild bird sightings along with Euclidean distance from roads (a measure of remoteness). Increased distances from irrigated farmland also increased the odds of H5 binding. The risk factors which associated with H5 binding and wild bird contact are indicative of remote habitats and low populous areas. Inclusion of poultry ownership, within the household or surrounding villages, did not improve model fit for H5 binding (Supplemental Table 4).
Using these environmental risk factors, we fit Bayesian models for wild bird species distributions and H5 ELISA data. All models had statistically significant spatial autocorrelation detected by Moran’s I. The wild bird distribution indicates clustered data, with a positive Moran’s I value of 0.0190591 (expected value of -0.0001059547, p < 2.2e-16). Similar results were observed for H5 binding, in which Iobserved= 0.1235021, Iexpected= -0.0005376344, and p < 2.2e-16. Based on Moran’s I, we then fit geostatistical models including spatially structured random effects. H1 (p = 0.6692209), H3 (p ≈ 1), and H5N2 (p = 0.1640969) binding did not show spatial autocorrelation via Moran’s I, so the results of these assays were not included in further geospatial analysis.
We subsequently fit joint spatial models of wild shorebird distribution and H5 binding. Incorporating a shared spatial random effect between shorebird distribution and H5 binding improved model performance, suggesting common spatial processes between H5 exposure and migratory bird distribution (Supplemental Table 5, 6).
Our results show evidence of heterogenous serological responses to avian influenza in Sabah. These results are spatially correlated and follow the distribution and habitats of migratory wild birds over domesticated poultry in the region. As contact with avian species is currently necessary for a spillover event, it is critical to consider these migratory sights as key interfaces in stopping the viral spread.
Infection or vaccination with one influenza subtype is thought to generate memory B cells which could produce antibodies that bind to shared epitopes on other hemagglutinins31,32. The extent and pervasiveness to which these so-called cross-reactive antibodies exist across multiple HA subtypes and virus strains is not fully understood33. In this study, we reported a lack of correlations in the ELISA binding to different antigens (Fig. 2a and Supplemental Fig. 1). This may serve as evidence that the responses we see are not driven by the presence of cross-reactive antibodies. A study by Nachbagauer et al. showed that cross-reactive antibodies against H5 HAs were present after H1 infection in mice and that these antibodies could be detected by ELISAs33. The cross-reactive binding responses shown in Nachbagauer et al. to H5 after H3 infection were limited when compared to the H1 results, as H1 and H5 belong to the same HA phylogenetic group32,34. Therefore, if our H5 binding responses were the result of cross-reactive antibodies, we would expect a much stronger correlation and spatial relationship between the H1 and H5 results (instead we noted correlations near zero). The highest correlation in our study is between the H5 antigens. As these are from the same HA subtype, some level of cross-reactivity is likely31–33. This evidence supports our hypothesis that the binding to H5 is not due to some general influenza reactivity or recent seasonal infection/vaccination and is indeed H5-specific.
H5 phylogeny adds further credibility to our results. A/Duck/Laos/3295/2005(H5N1) and 2.3.4 clade viruses were circulating in Asia for at least 10 years prior to our sample collection26. If there was exposure to an H5 virus in our study cohort, it is possible that it could have been to a 2.3.4-like virus due to the timings of circulation. While we saw a range of binding responses to the HA from 2.3.4, we hardly see any responses above AU 1 (n = 11, 0.55%) to our A/snow goose/Missouri/CC15-84A/2015(H5N2) 2.3.4.4 virus (Fig. 2a). As this second viral HA clade did not become dominant until three years after our sample collection27, it is highly unlikely that members of our study cohort could have had an opportunity to be exposed to a 2.3.4.4 virus. The phylogenetic specificity of our binding results amongst H5 responses would not be observed if these responses were exclusively cross-reactive.
The observed trends in neutralisation are also supported by phylogenetic evidence. Our 2.3.4 H5N1 ELISA antigen has more phylogenetic similarity with A/Bar headed goose/Qinghai/1A/2005(H5N1) (clade 2.2) than A/Indonesia/05/2005(H5N1) (clade 2.1.3.2), according to a phylogenetic analysis run in Mega X, reported as pairwise distance (Supplementary Table 1). None of our high binders to 2.3.4 by ELISA neutralised A/Indonesia/05/2005, while 20% had NT50 > 1000 (n = 4) and 16 (80%) had NT50 > 100. This shows that the serological responses against H5 observed in this cohort are not general H5 responses. In fact, the responses are temporally and phylogenetically specific, which suggests that the individuals in our cohort may have encountered a virus closely related to the H5 clades 2.3.4 or 2.2. A limitation of this work is the inability to obtain matching ELISA and pseudoneutralisation hemagglutinins, but the observed binding and neutralisation are in accordance with the phylogenetic distance.
The persistence of these neutralising responses upon treatment with seasonal flu-coated beads also illustrates that the observed responses behave more like true H5 exposures than cross-reactive neutralisers (Fig. 2b and Fig. 3). Our study is limited in that a lack of PCR and health data makes it impossible to know if this population had genuine symptomatic cases of H5 avian influenza in this population. However, attributing the results of this study to seasonal influenza alone would not fully explain the unusual responses we have observed. Previous work has highlighted the strong correlations between pseudovirus neutralisation and live virus neutralisation results for H5N1 IAV28,35, further validating these results even without access to health data. We propose that our depletion assay could be useful to other researchers seeking to contextualise the role of cross-reactivity in their influenza serology results when clinical data is not available.
The geospatial distributions of our responses add further support to our hypothesis. Binding to seasonal influenza antigens exhibits a wide range of responses regardless of exposure to reservoirs, and this data is not spatially correlated (Fig. 2a). As seasonal influenza viruses are endemic, it is reassuring that we see spatial homogeneity. Indeed, if the seasonal influenza binding shared trends with that of the avian influenza, it would be far more likely that the H5 results would be the result of cross-reactive antibodies.
We also did not see a spatial correlation in H5 2.3.4.4 binding. As members of our sample population would likely not have had the chance to be exposed to this post-2015 virus, spatially correlated responses would have been highly unexpected. Instead, we see that only H5 2.3.4 binding, only, was higher amongst those living near migratory shorebird locations and that this data is highly spatially correlated (Fig. 2a and Fig. 4a). These apparent spatial patterns suggest that the H5 2.3.4 results are distinct from the other ELISA responses and may represent a footprint of some genuine previous exposure.
There are shared environmental risk factors between the wild bird distributions and H5 binding (Fig. 4b). The identification of proximity to the sea, low elevation (i.e. sea level) areas, closeness to the forest, and remote areas indicate that we are not only reporting wild migratory shorebirds in their natural habitats16–18, but that these habitats are also areas where binding to the H5 2.3.4 antigen was highest. Shared environmental risk factors and the overlapping spatial distributions between wild birds and H5 binding, suggest that some kind of contact may be occurring between humans and wild birds in these locations. Future studies could collect contemporaneous data on wild bird movements and pathogen presence.
One mechanism of H5 spillover is the spread of the virus from wild birds to domesticated poultry and then into humans36. We can surmise that H5 2.3.4 binding is not related to poultry contact in this study, the inclusion of poultry ownership as a fixed effect did not approve model fit. Although poultry farmers can be at risk of encountering avian influenza36, our findings suggest a need to consider individuals living close to these migratory sites as a part of regular surveillance efforts (Supplemental Fig. 5).
Migratory shorebirds transiting between countries may carry influenza and expose individuals in the surrounding areas to avian viruses. Our results suggest that individuals living within 10km of known migratory locations may have had previously unknown exposure to avian influenza of the 2.3.4 or similar clade. As shorebird habits are being destroyed due to rising sea levels and land use changes, there is an urgent need to consider how this may force zoonotic reservoirs including migratory wild birds into closer contact with humans and increasing the risk of HPAI spillover.
Samples collection and ethical approval
Scottish Blood Donors
A total of 63 serum samples from blood donors aged 17–80 collected by the Scottish National Blood Transfusion Service (SNBTS) in 2020 were used for this study. Ethical approval was obtained for the SNBTS anonymous archive - IRAS project number 18005. SNBTS blood donors gave fully informed consent to virological testing; donation was made under the SNBTS Blood Establishment Authorisation, and the study was approved by the SNBTS Research and Sample Governance Committee.
Cross-sectional survey of rural Malaysian Borneo (Sabah)
Samples were collected, as in Fornace et al.38 in 2015, from individuals in the Pitas, Kudat, Ranau, and Kota Marudu districts of the Sabah region in Malaysian Borneo. All individuals in households selected for the environmental survey responses were asked to donate blood unless they were younger than 3 months old or could not be reached after three attempts. Whole blood was collected into precoated EDTA tubes (Becton-Dickinson, Franklin Lakes, USA). Samples were tested by ELISA and pseudoneutralisation assays after processing.
Written informed consent was obtained from all study participants or a legal guardian. The Medical Research Sub-Committee of the Malaysian Ministry of Health (NMRR-14-713-21117), the Research Ethics Committee of the London School of Hygiene & Tropical Medicine (8340), and the Oxford Tropical Research Ethics Committee (560 − 22) have approved this study.
Urban Malaysian samples (Kota Kinabalu)
678 samples were collected from blood donors in Kota Kinabalu, Malaysian Borneo from 2017–2020 for a cross-sectional study on Leptospirosis run in 2017 by Jeffree et al.39. These samples were tested by ELISA in Kota Kinabalu, Malaysia at the Borneo Medical Health Research Centre. Ethical clearance was obtained from the Ethics Committee of the Faculty of Medicine and Health Sciences, Universiti Malaysia Sabah, Kota Kinabalu, Sabah, Malaysia (UMS).
Sample selection
The 2000 samples used for ELISAs in this study were chosen at random from the 10,100 samples in the rural cohort, including 500 samples each from the four districts in the study (Pitas, Kudat, Ranau, and Kota Marudu).
Viral proteins, antibodies, & antiserum
All viral proteins were obtained through BEI Resources, NIAD, NIH from various contributors (Supplemental Table 7). All HAs were produced in Sf9 insect cells using a baculovirus expression system with a variety of purification methods and tags. All antibodies and antiserum for IAV HAs were sourced externally (Supplemental Table 8).
Detection of IgG binding by ELISA
The binding of antibodies to a variety of seasonal and avian haemagglutinin was measured by IgG ELISAs in a protocol adapted from Thom et al40.
MaxiSORP™ NUNC-Immuno plates were coated with 0.125µg/mL of A/Duck/Laos/3295/2006(H5N1), A/snow goose/Missouri/CC15-84A/2015 (H5N1), and A/New York/55/2004(H3N2) or 1µg/mL for A/California/07/pdm2009(H1N1) in 50µL 1X PBS. Plates were incubated overnight at 4°C. The following day, plates were blocked with 200µL Blocker™ casein in PBS for 1hr on a shaking incubator at RT. The plates were then washed six times with 200µL of 0.05% PBST and tapped dry. All subsequent wash steps follow the same procedure. Serum and positive controls were added in duplicate at a 1:50 dilution in 50µL Blocker™ casein in PBS and incubated for 1hr, shaking at RT (each plate contained a human and goat antiserum control for intra-assay validation). Plates were washed and tapped dry.
The detection/secondary antibody (either anti-human or anti-goat, Supplemental Table 8) was then added to the plate at 1:2000 in 100µL, as recommended by the suppliers, and incubated for 30min at RT. The plates were washed and tapped dry, and 100µL of BioFX® TMB One Component HRP Microwell Substrate was added. The plates were allowed to develop for 15minutes before stopping the reaction with 100uL of BioFX® 450nm Stop Reagent for TMB Microwell Substrates to prevent over-saturation. Absorbance was read at an optical density of 450nm with a GloMax® Explorer Multimode Microplate Reader. Results were analysed manually (background subtraction and averaging of replicates) and plotted in R Studio41 with ggplot242, ggstatsplot43, and R 4.2.044 and GraphPad Prism 10.0.3.
Polyclonal goat antiserum, post-H5 vaccination pooled human plasma, and H1N1 convalescent sera were used as positive controls (Supplemental Table 8).
Pseudotyped influenza virus production
Pseudotyped lentiviruses displaying IAV H5 haemagglutinins were produced via transfection of human embryonic kidney (HEK) 293T cells (Sigma, 12022001), as described in Thompson et al.45 and Temperton et al.28 1.0µg of gag/pol construct (p8.91), 1.5µg of a luciferase reporter carrying construct (PCSFLW), and 1.0µg of HA glycoprotein expressing construct were combined with 200µL OptiMEM (Gibco™, 31985062) and 35µL of 1 mg/mL polyethyleneimine (PEI) branched (Sigma, 408727). The p8.91, pCSFLW, and the HA-expressing plasmids were obtained from CPT.
Transfections were performed in 10mL of Dulbecco’s Modified Eagle Medium (DMEM) 1X, high glucose, GlutaMAX™ (Gibco™, 10569010) with 10% fetal bovine serum (FBS) (Gibco™, 16000044) and 1% penicillin-streptomycin (Sigma-Aldrich, P4333) and left for 24hrs. One unit of neuraminidase from Clostridium perfringens (Sigma-Aldrich, N2876) was added to fresh media to induce virus budding. After 48hrs, pseudotyped viruses were removed and filtered with a sterile 0.45-µm Millex®-HA filter unit (Millipore, SLHA033SB). The HA glycoprotein constructs used for production of the pseudotyped H5N1 lentiviruses were A/Indonesia/05/2005 (H5N1) and A/Bar headed goose/Qinghai/1A/2005 (H5N1).
The pseudotyped influenza viruses were titrated for use in microneutralisation assays. Pseudotyped influenza viruses with initial titres above 1x105 relative light units (RLU) were be used (Supplemental Fig. 6).
Pseudotyped microneutralisation assay
Neutralisation of the pseudotyped influenza viruses was quantified using a microneutralisation assay. Sera was added in duplicate at a 1:20 dilution in 100uL DMEM to a white Nunc™ MicroWell™ 96-Well, flat-bottom microplate. Each serum sample was then serially diluted down the plate (Row A to Row H) 1:1 in 50µL DMEM. The plates were incubated for 2hrs at 37°C with 50µL 1x105 RLU pseudotyped influenza virus per well. One row per plate contained virus, cells, and no serum, and one row is a cell-only control.
After 2hrs, 50µL of 1x105 HEK 293T cells were added to each well and incubated for a further 48hrs at 37°C. The cells were then lysed with 50µL per well of a 1:1 mixture of Bright-Glo™ Luciferase Assay System and 1X PBS. The RLU of the cell lysate was determined using a GloMax® Explorer Multimode Microplate Reader.
Microneutralisation assay data was analysed in GraphPad Prism 10.0.3. The reduction of infectivity from the microneutralisation assay was determined by comparing the RLU in the presence and absence of serum and expressed as percentage neutralisation. The data was normalised using the virus and cell-only control wells, and a neutralisation curve was fit with non-linear regression (log-inhibitor versus normalized response). The 50%-neutralising titre (NT50) is here defined as the sample dilution at which viral RLU were reduced by 50% compared with control wells.
A sample was considered neutralising if the dilution factor of serum needed to reach NT50 was greater than or equal to 1:100, as established by Temperton et al.28 To account for cross-reactive neutralisation, we used the highest NT50 from the cohort of negative control Scottish blood donor samples to create cut-offs for seropositivity (1:1173 for A/Indonesia/05/2005 and 1: 1015 for A/Bar headed goose/Qinghai/1A/2005) as in Thompson et al29 (Supplemental Fig. 2). Where applicable, NT50s were compared in Prism using the Extra sum-of-squares F-test, with p < 0.05 as the cut-off for statistically significant differences in NT50 values.
Microsphere (bead) protein conjugation
Carboxyl magnetic particles (beads) (Spherotech, SPHERO™ 4.2µm) were conjugated with IAV HAs as described in Brown et al.46, Barrett et al.47, and Tomic et al.48 for plasma depletion assays. Beads were coated with HAs from a variety of seasonal influenzas obtained from BEI Resources: A/Wisconsin/67/2005(H3N2),and A/California/07/2009(H1N1), A/New York/18/2009(H1N1). A subset of beads prepared were not coated with any HA to test the effect of the beads themselves in assays (referred to as non-coated beads).
All previously prepared beads were diluted to 200beads/µL in Assay Buffer (buffers described in Supplemental Table 9). 50uL of each solution of beads was added to a 5mL sterile, round bottom, polystyrene Falcon® test tube (Corning, 352054). Four conditions were prepared to test each bead: a blank tube with no secondary antibody, a positive control serum tube, a negative control serum tube, and an IgG isotype tube to identify any non-specific binding of the secondary antibody. The positive control serum was an individual who also exhibited high H1 and H3 antibody binding responses by ELISA relative to the negative control. The negative control serum exhibited lower antibody binding to the seasonal HA antigens by ELISA. Responses by ELISA were used to establish controls, as past influenza subtype exposure is often unknown.
Serum samples were diluted 1:100 (50uL) in Assay Buffer and incubated with beads for 1hr at RT, shaking at 700rpm. Beads were pelleted on the EasyEights™ EasySep™ magnet (STEMCELL™ Technologies, 18103) for 1min and supernatant was gently aspirated without disturbing the bead pellet. Beads were washed three times using 1mL Wash Buffer per tube. Beads were resuspended in 200µL Assay Buffer. 100µL of 1µg/mL phycoerythrin (PE)-conjugated secondary antibody was added to the tubes, either the isotype control antibody (BioLegend #400112) or anti-human IgG (Southern Biotech #9040-09) depending on the condition. No secondary antibody was added to the blank condition. All tubes were then left to incubate for 1hr at RT in the dark, shaking. Beads were pelleted on a magnet and washed three times in 1mL Wash Buffer. Samples were resuspended in 500µL of Storage Buffer and vortexed.
Samples were then acquired on the flow cytometer by CB to determine if the conjugation of viral proteins to the beads was successful. Side scatter versus forward scatter plots were used to draw gate 1 to exclude debris and doublets. The count versus PE histogram was used to observe PE fluorescence in the different conditions (Supplemental Fig. 3). Data was analysed with FlowJo™ 10.8.2.
Bead-based cross-reactivity depletion
To determine the cross-reactivity of any H5 IAV neutralising and binding antibodies, a serum depletion assay was developed using the previously HA-conjugated beads. The aim was to isolate antibodies which bound to seasonal IAV-coated beads and determine if any H5 binding or neutralising ability remained.
For pseudotyped neutralisation assays, the bead mixture was diluted to a concentration of 100 beads/µL/bead (total concentration of 300 beads/µL). An equal concentration of A/Wisconsin/67/2005(H3N2), A/California/07/2009(H1N1), and A/New York/18/2009(H1N1) coated beads were used. Human serum was added to this bead mixture at a 1:20 dilution. This was incubated for 30mins, and the mixture was incubated on the magnet. The flow-through (media and serum not bound to the magnet) was transferred to an Eppendorf™ tube with fresh beads. This process was repeated four more times, for a total of five passages through fresh beads. Five passages generated the optimal depletion (highest level of depletion whilst conserving resources) (Supplemental Fig. 7a).
After the final passage, the mixture was incubated on the magnet for 1min. While on the magnet, 100uL of flow-through was added in duplicate to Row A of a white Nunc™ MicroWell™ 96-well plate and serially diluted to Row H. Each plate also contained untreated serum, a virus-only control row, and a cell-only control. The neutralisation assays were then completed and analysed as per the previous neutralisation results. NT50s of the treated and untreated conditions were compared in Prism using the Extra sum-of-squares F-test, with p < 0.05 as the cut-off for statistically significant differences in NT50 values.
The effect of the beads themselves was also tested at the final assay concentration (300beads/µL) using non-coated beads. No effect was observed with the uncoated beads (Supplemental Fig. 7b).
Environmental risk factor analysis
We aimed to assess the spatial distribution of H5 binding, wild bird populations, and domesticated poultry. Binomial generalized linear mixed effects models (GLMMs) were used with the outcome as the proportion of individuals per household with high H5 binding. For poultry ownership the number of poultry owners per village was used as an outcome. For wild bird contact, the outcome was whether there was a bird sighting at a particular location.
Households included in the study were geolocated and integrated with remote sensing-derived environmental data on land cover and climatic factors. All environmental data was curated as in Klim et al.25 and Fornace et al.38,49. The possible predictor variables were mean-centred and scaled. The model with lowest AIC is was selected using MASS::stepAIC using a stepwise parsimonious approach (both forward and backward selection)50 (Supplemental Tables 2–4). Odds ratios were calculated and plotted using the sjPlot51 package in RStudio41,44.
Spatial autocorrelation of the residuals from the GLMM results was determined with Moran’s I, where p < 0.05 was considered statistically significant. Models exhibiting residual spatial autocorrelation were integrated into a Bayesian framework with integrated nested Laplace approximations (INLA) in R-INLA52,53. Spatial effects were modelled as a Matérn covariance function, using the stochastic partial differential equation (SPDE) method54. For the model intercepts and fixed effects coefficients, weakly informative priors of Normal (0, 100) were used55.
The final models were evaluated using the deviance information criteria (DIC). Posterior probabilities were estimated using 1000 posterior samples. These posterior probabilities were then used to predict the probability of the outcome variable (H5 binding or species distribution) across the whole study region. Uncertainty for these predictions was visualized through standard deviation. Posterior probabilities were visualized with ggplot242 and the wesanderson colour palette (copyright Karthik Ram, 2022).
Acknowledgements
We would like to thank the Director General of Health, Malaysia for permission to publish this study. We acknowledge the UK Medical Research Council, Natural Environment Research Council, Economic and Social Research Council and Biotechnology and Bioscience Research Council for funding received through the Environmental and Social Ecology of Human Infectious Diseases Initiative (Grant No. G1100796). KMF is supported by a Sir Henry Dale fellowship jointly funded by the Wellcome Trust and Royal Society (Grant No. 221963/Z/20/Z). This research was supported in part by US Food and Drug Administration Medical Countermeasures Initiative (contract 75F40120C00085), and the UK Public Health Rapid Support Team (Grant No. PSR00130). We acknowledge the reagents obtained through BEI Resources, NIAID, NIH and the specific providers listed in Supplemental Table 6. HK is supported by the Future of Humanity Institute at the University of Oxford DPhil Scholarship programme.
Author Contributions
HK undertook serological assays, geostatistical analysis, and study design. CB assisted with experimental design and flow cytometry experiments. THC and TW collected data and contributed to study design. HBM assisted with serological assays. GSR contributed to conceptualising the study. JM and TT assisted with experimental design and serological assays. JLJ, MSJ, and KA provided Kota Kinabalu control samples and facility use at UMS BMHRC, and KA contributed to the study design. CPT assisted with the experimental design and contributed reagents. KMF, CJD, and MWC conceptualised the study and analysed the data. KMF performed geostatistical analyses.
Conflict of Interest
The authors declare that the research was conducted in the absence of any commercial or financialrelationships that could be construed as a potential conflict of interest.
Data Availability Statement
All environmental data are available on reasonable request. As survey and serological data includes identifiable information on household coordinates, data is only available following approval by the relevant ethics committees in Malaysia and the UK.
- World Health Organization: Global Influenza Programme. Assessment of risk associated with recent influenza A(H5N1) clade 2.3.4.4b viruses. 3 (2022).
- Food and Agriculture Organization of the United Nations, The World Health Organization & The World Organisation for Animal Health. in Situation analysis and advice to countries from FAO, WHO, WOAH Geneva/Paris/Rome (2023).
- Gilbert, M., Slingenbergh, J. & Xiao, X. Climate change and avian influenza. Rev Sci Tech 27, 459-466 (2008).
- Prosser, D. J., Teitelbaum, C. S., Yin, S., Hill, N. J. & Xiao, X. Climate change impacts on bird migration and highly pathogenic avian influenza. Nature Microbiology 8, 2223-2225, doi:10.1038/s41564-023-01538-0 (2023).
- Fuller, T. et al. The Ecology of Emerging Infectious Diseases in Migratory Birds: An Assessment of the Role of Climate Change and Priorities for Future Research. EcoHealth 9, 80-88, doi:10.1007/s10393-012-0750-1 (2012).
- Carlson, C. J. et al. Climate change increases cross-species viral transmission risk. Nature, doi:10.1038/s41586-022-04788-w (2022).
- Venkatesan, P. Avian influenza spillover into mammals. The Lancet Microbe 4, e492, doi:10.1016/S2666-5247(23)00173-8 (2023).
- Mirolo, M. et al. Highly pathogenic avian influenza A virus (HPAIV) H5N1 infection in two European grey seals (Halichoerus grypus) with encephalitis. Emerg Microbes Infect 12, e2257810, doi:10.1080/22221751.2023.2257810 (2023).
- Maemura, T. et al. Characterization of highly pathogenic clade 2.3.4.4b H5N1 mink influenza viruses. eBioMedicine 97, doi:10.1016/j.ebiom.2023.104827 (2023).
- Ulloa, M. et al. Mass mortality event in South American sea lions (Otaria flavescens) correlated to highly pathogenic avian influenza (HPAI) H5N1 outbreak in Chile. Vet Q 43, 1-10, doi:10.1080/01652176.2023.2265173 (2023).
- European Centre for Disease Prevention and Control. (2023).
- Gomaa, M. et al. We are underestimating, again, the true burden of H5N1 in humans. BMJ Glob Health 8, doi:10.1136/bmjgh-2023-013146 (2023).
- M. Bamford, D. W., W. Bancroft, G. Tischler, and J. Wahl. Migratory Shorebirds of the East Asian – Australasian Flyway; Population Estimates and Internationally Important Sites. (Wetlands International, 2008).
- Ocampo-Peñuela, N., Garcia-Ulloa, J., Kornecki, I., Philipson, C. D. & Ghazoul, J. Impacts of Four Decades of Forest Loss on Vertebrate Functional Habitat on Borneo. Frontiers in Forests and Global Change 3, doi:10.3389/ffgc.2020.00053 (2020).
- Gaveau, D. L. A. et al. Rapid conversions and avoided deforestation: examining four decades of industrial plantation expansion in Borneo. Scientific Reports 6, 32017, doi:10.1038/srep32017 (2016).
- Wulffraat, S. et al. Supplementary Report: The Environmental Status of Borneo. (World Wildlife Fund, 2017).
- Richards, D. R. & Friess, D. A. Rates and drivers of mangrove deforestation in Southeast Asia, 2000–2012. Proceedings of the National Academy of Sciences 113, 344-349, doi:doi:10.1073/pnas.1510272113 (2016).
- Posa, M. R. C., Wijedasa, L. S. & Corlett, R. T. Biodiversity and Conservation of Tropical Peat Swamp Forests. BioScience 61, 49-57, doi:10.1525/bio.2011.61.1.10 (2011).
- Mellouli, F. E. et al. Molecular Detection of Avian Influenza Virus in Wild Birds in Morocco, 2016–2019. Avian Diseases 66, 29-38, 10 (2021).
- El-Shesheny, R. et al. Replication and pathogenic potential of influenza A virus subtypes H3, H7, and H15 from free-range ducks in Bangladesh in mammals. Emerg Microbes Infect 7, 70, doi:10.1038/s41426-018-0072-7 (2018).
- Wille, M. et al. Serologic Evidence of Exposure to Highly Pathogenic Avian Influenza H5 Viruses in Migratory Shorebirds, Australia. Emerging Infectious Disease journal 25, 1903, doi:10.3201/eid2510.190699 (2019).
- Leow, B. L. et al. Molecular characterisation of hemagglutinin and neuraminidase genes of the first highly pathogenic Avian Influenza H5N1 2.3.2.1c virus isolated from Sabah, Malaysia. Trop Biomed 39, 160-169, doi:10.47665/tb.39.2.001 (2022).
- Quaza Nizamuddin Bin Hassan Nizam. Self-declaration on the recovery of freedom from highly pathogenic avian influenza by Malaysia. (World Organisation for Animal Health, 2018).
- Mahlapuu, M. H., J.; Ringstad, L.; Bjorn, C. Antimicrobial Peptides: An Emerging Category of Therapeutic Agents. Frontiers in Cellular and Infection Microbiology 6 (2016).
- Klim, H. et al. Quantifying human-animal contact rates in Malaysian Borneo: Influence of agricultural landscapes on contact with potential zoonotic disease reservoirs. Frontiers in Epidemiology 2, doi:10.3389/fepid.2022.1057047 (2023).
- Dhingra, M. S. et al. Global mapping of highly pathogenic avian influenza H5N1 and H5Nx clade 2.3.4.4 viruses with spatial cross-validation. Elife 5, doi:10.7554/eLife.19571 (2016).
- Centers for Disease Control and Prevention. Highlights in the History of Avian Influenza (Bird Flu) Timeline – 2010-2019, <https://www.cdc.gov/flu/avianflu/timeline/avian-timeline-2010s.htm> (2022).
- Temperton, N. J. et al. A sensitive retroviral pseudotype assay for influenza H5N1-neutralizing antibodies. Influenza and Other Respiratory Viruses 1, 105-112, doi:10.1111/j.1750-2659.2007.00016.x (2007).
- Thompson, C. P. et al. Detection of neutralising antibodies to SARS-CoV-2 to determine population exposure in Scottish blood donors between March and May 2020. Eurosurveillance 25, doi:10.2807/1560-7917.es.2020.25.42.2000685 (2020).
- Sullivan, B. L. et al. eBird: A citizen-based bird observation network in the biological sciences. Biological Conservation 142, 2282-2292, doi:https://doi.org/10.1016/j.biocon.2009.05.006 (2009).
- Ekiert, D. C. et al. Antibody recognition of a highly conserved influenza virus epitope. Science 324, 246-251, doi:10.1126/science.1171491 (2009).
- Sui, J. et al. Structural and functional bases for broad-spectrum neutralization of avian and human influenza A viruses. Nat Struct Mol Biol 16, 265-273, doi:10.1038/nsmb.1566 (2009).
- Nachbagauer, R. et al. Defining the antibody cross-reactome directed against the influenza virus surface glycoproteins. Nat Immunol 18, 464-473, doi:10.1038/ni.3684 (2017).
- Nobusawa, E. et al. Comparison of complete amino acid sequences and receptor-binding properties among 13 serotypes of hemagglutinins of influenza A viruses. Virology 182, 475-485, doi:https://doi.org/10.1016/0042-6822(91)90588-3 (1991).
- Cantoni, D. et al. Correlation between pseudotyped virus and authentic virus neutralisation assays, a systematic review and meta-analysis of the literature. Front Immunol 14, 1184362, doi:10.3389/fimmu.2023.1184362 (2023).
- Centers for Disease Control and Prevention. Bird Flu in Birds, <https://www.cdc.gov/flu/avianflu/avian-in-birds.htm> (2022).
- QGIS Geographic Information System v. 3.30.2 (Open Source Geospatial Foundation, 2009).
- Fornace, K. M. et al. Environmental risk factors and exposure to the zoonotic malaria parasite Plasmodium knowlesi across northern Sabah, Malaysia: a population-based cross-sectional survey. The Lancet Planetary Health 3, e179-e186, doi:10.1016/s2542-5196(19)30045-2 (2019).
- Jeffree, M. S. et al. High incidence of asymptomatic leptospirosis among urban sanitation workers from Kota Kinabalu, Sabah, Malaysian Borneo. Sci Rep 10, 19442, doi:10.1038/s41598-020-76595-0 (2020).
- Thom, R. et al. Longitudinal antibody and T cell responses in Ebola virus disease survivors and contacts: an observational cohort study. The Lancet Infectious Diseases 21, 507-516, doi:https://doi.org/10.1016/S1473-3099(20)30736-2 (2021).
- RStudio: Integrated Development for R. (RStudio, PBC, Boston, MA, 2020).
- H, W. ggplot2: Elegant Graphics for Data Analysis. (Springer-Verlag, 2016).
- Visualizations with statistical details: The 'ggstatsplot' approach (Journal of Open Source Software, 2021).
- R Core Team. R: A language and environment for statistical computing. , <https://www.R-project.org/.> (2020).
- Thompson, C. P. et al. A naturally protective epitope of limited variability as an influenza vaccine target. Nature Communications 9, doi:10.1038/s41467-018-06228-8 (2018).
- Brown, E. P. et al. High-throughput, multiplexed IgG subclassing of antigen-specific antibodies from clinical samples. J Immunol Methods 386, 117-123, doi:10.1016/j.jim.2012.09.007 (2012).
- Barrett, J. R. et al. Phase 1/2 trial of SARS-CoV-2 vaccine ChAdOx1 nCoV-19 with a booster dose induces multifunctional antibody responses. Nature Medicine 27, 279-288, doi:10.1038/s41591-020-01179-4 (2021).
- Tomic, A. et al. Divergent trajectories of antiviral memory after SARS-CoV-2 infection. Nature Communications 13, 1251, doi:10.1038/s41467-022-28898-1 (2022).
- Fornace, K. M. et al. Association between Landscape Factors and Spatial Patterns of Plasmodium knowlesi Infections in Sabah, Malaysia. Emerg Infect Dis 22, 201-208, doi:10.3201/eid2202.150656 (2016).
- Venables, W. & Ripley, B. Modern Applied Statistics with S. 4th edn, (Springer, 2002).
- sjPlot: Data Visualization for Statistics in Social Science v. 2.8.15 (2023).
- A breakthrough in the development of a universal flu vaccine, <https://www.ox.ac.uk/research/research-impact/breakthrough-development-universal-flu-vaccine> (
- Martins, T. G., Simpson, D., Lindgren, F. & Rue, H. Bayesian computing with INLA: New features. Computational Statistics & Data Analysis 67, 68-83, doi:https://doi.org/10.1016/j.csda.2013.04.014 (2013).
- Lindgren, F., Rue, H. & Lindström, J. An explicit link between Gaussian fields and Gaussian Markov random fields: the stochastic partial differential equation approach. Journal of the Royal Statistical Society: Series B (Statistical Methodology) 73, 423-498, doi:https://doi.org/10.1111/j.1467-9868.2011.00777.x (2011).
- Simpson, D., Rue, H., Riebler, A., Martins, T. G. & Sørbye, S. H. Penalising Model Component Complexity: A Principled, Practical Approach to Constructing Priors. Statistical Science 32, 1-28, 28 (2017).
Table 1: Abundance of species observed in the Sabah region of Malaysian Borneo from July 1979 to August 2023 and reported to the eBird database.
There is NO Competing Interest.