Does exposure to poultry and wild fowl confer immunity to H5N1?

Human outbreaks of highly pathogenic avian influenza (HPAI) such as H5N1 and novel avian strains such as H7N9 have provoked significant public health concern. An outbreak of H5N1 in humans was first reported in Hong Kong (HK), China in 1997. This event was curtailed by a variety of public health measures including the culling of over 1.5 million chickens in the city.¹ H5N1 has since re-emerged in multiple countries with over 600 reported infections and, since 2003, a case fatality of about 59% (World Health Organization (WHO) report, as of January 24, 2014). A clear picture of the prevalence and transmissibility of avian influenza strains in humans is lacking, in part due to sparse epidemiological data and limitations in current detection methods. A recent meta-analysis estimated a 1.2% (95% CI: 0.6%–2.1%) seropositive rate for H5N1 in humans.² Here we argue that H5N1 seroprevalence could be region specific, due to differing levels of exposure to wild and domestic fowl. We hypothesize that persistent environmental exposure to avian influenza strains may enhance cross-immunity against HPAI strains such as H5N1, and thus lower H5N1 seropositive rates. In contrast, people with limited exposure to poultry and wild birds may have less cross-immunity against HPAI (e.g., H5N1) and thus could be more susceptible to the virus.

Our hypothesis is supported by two independent lines of evidence: one epidemiological, and the other immunological. The former evidence shows that populations with frequent exposure to poultry have lower H5N1 seropositive rates; the latter indicates that cross-immunity between human and avian strains exists and its strength may depend on the frequency of exposure. Each of these lines of evidence is detailed herewith.

Epidemiological evidence: lower H5N1 seropositive rates in rural Southeast Asia

To assess human susceptibility to H5N1, we reviewed 33 H5N1 seroprevalence studies on populations at high risk from 11 regions (Table 1). We define populations at high risk as those with suspected exposure to H5N1 from either poultry or H5N1 patients, or both; exposed individuals include poultry workers (PWs), healthcare workers, and close contacts of H5N1 patients. To minimize potential bias due to differing detection methods, we included only studies that adopted the WHO criteria for serological diagnosis. Among these studies the seropositive rates were reported to be: 1.4% (25/1 805) in Cambodia, 1.4% (9/641) in mainland of China, 0.9% (10/1 064) in Vietnam, 0 (0/2 290) in Thailand, and 0 (0/1 422) in Indonesia (Table 2). In comparison, investigations on the 1997 outbreak in HK indicated that about 8.4% (179/2 135) of participants were positive for H5N1, over five times higher than in other regions. As this comparison is based on studies conducted in similar circumstances (i.e., recent potential exposure to H5N1), the higher seropositive rates observed in the HK studies are likely not an artifact of serosurvey timing. Additionally, we included the 1997 HK outbreak in this analysis as it is confirmed to have been caused by an H5N1 strain and led to severe illnesses and a case-fatality rate of 33%.³

Table 1.

Compiled data describing seroprevelance of H5N1 virus infection in populations at high risk^*

Region	Study population	When	Positive (% (pos/total))	Criteria	Note	Reference
Hong Kong,  China	217 HCWs Exp to H5N1 cases	1997; ca. 2 weeks after exposure	3.69 (8/217)	WHO		17
	293 GWs in poultry culling; 1525 PWs	December, 1997–January, 1998; paired sera (0–7 d after the culling, and 2 weeks later) for GWs; single smp from PWs	3 (9/293) for GWs; ca. 10 (ca. 155/1 525) for PWs	WHO		18
	101 CCs (51 HHCs, 26 tour group members, 24 exp coworkers)	Nov 1997; Most paired sera collected ≤11 d and ≥21 daysafter exposure	6.93 (6/51, 1/26, 0/24)	WHO		19
Cambodia	351 villagers resided ≤ 1 km of 1 case	June, 2005; ca. 3 months after suspected exposure	0 (0/351)	WHO		47
	80 CCs (10 relatives, 28 HCWs, 42 villagers)	January, 2005–April, 2006	0 (0/80)	WHO		48
	674 villagers resided ≤ 1 km of 2 cases	May, 2006; ca. 7 weeks after suspected exposure	1.04 (7/674)	WHO		49
	700 villagers in same village of 1 case	April, 2007; after identifing 1 case	2.57 (18/700)	WHO equivalent	2 Pos had contact w/ case; 11 had contact w/ and ate sick poultry	6
Mainland of China	231 PWs in H5N1 outbreak areas	April–June 2004	3.03 (7/231)	HI ≥20	22 (9.52%) H9N2 Pos, 1 (0.43%) H7N7 Pos in the PWs; 2.34% H5N1 Pos and 3.76% H9N2 Pos in 983 non-PWs	14
	191 CCs of 2 cases	October, 2005	0 (0/191)	Obs	Observed for 10 d after last exposure, none sick; cases occurred in rural area	50
	640 CCs of 6 cases	2005–2006	0 (0/640)	Obs	Observed for 2 wk, 5 had febrile RI but neg for H5N1 by MN; cases occurred in cities	51
	110 poultry purveyors at LBMs visited by a H5N1 case	ca. February–March, 2006	0.91 (1/110)	WHO	1 Pos had HI=320 but no recent RI and denied contact w/ sick birds	52
	91 CCs of 2 cases (son and father)	December, 2007; paired sera (≤1 week, and ≥3–4 weeks after last exposure)	0 (0/91)	WHO	Poultry sold in the LBM were H5- vaccinated	53
	134 matched Ctrls of 28 cases (matched by gender, age, and location)	2008; 11–486 d after onset of case- patient (median=360 days)	0 (0/134)	WHO	Exposure statues of Ctrls not clear	54
	252 poultry retailers; 244 poultry wholesalers; 125 PWs in large-scale poultry breeding enterprise; 869 farmers in small-scale rural poultry farms	March, 2007–July, 2008	0.79 (2/252) for poultry retailers; 0.82 (2/244) for poultry wholesalers; none from others	Not clear (could be HI≥ 20,55*** pos sera further confirmed)	5.77% (86/1490) H9N2 Pos; 0 H5N1 Pos, 13 (1.85%) H9N2 Pos in 701 non-PWs	15
	306 backyard PWs	July–Auguest, 2010	2.61 (8/306)	WHO (HI ≥160)	None reported serious RI	7
Vietnam	200 PWs; 200 Ctrls with similarly high EE to poultry	October, 2001	4 (8/200) in PWs; 1 (2/200) in Ctrls	WHO equivalent	ca. 50% Exp to backyard poultry in both groups	56
	106 ctrl-respondents of 28 cases, matched by gender, age, and place of residence	2004	0 (0/106)	MN (>40)	ca. 74% Ctrls had EE (e.g., backyard poultry)	8
	60 HWs	January, 2004; paired serum smp (ca. 1 week and 3 weeks after last exposure to patient)	0 (0/60)	WHO		57
	183 PWs; 317 cullers	2005; ca. 1–6 months after suspected exposure	0 (0/500)	WHO		58
	747 residents from a community with previous outbreaks in human and poultry	December, 2007; human infections in community in 2004, additional cases in nearby villages in 2007 and 2008	4.95 (37/747)	HI ≥40		59
Thailand	25 Exp HCWs	2004; paired sera (one w/in 1 week of exposure, the other 2 weeks later)	0 (0/25)	WHO		60
	242 ICU patients from area with confirmed cases	February, 2005–December, 2008	0 (0/242)	WHO	6 smp (2.4%) with low titers (10-40)	61,62
	322 PWs from 5 districts with outbreaks	May, 2004; 2–4 months after exposure	0 (0/322)	WHO	7 (2%) had low titers (not specified)	63
	901 villagers from 4 villages had ≥1 case	October, 2005; ca. 12–14 months after suspected exposure	0 (0/901)	WHO equivalent	68.1% Exp to backyard poultry, 25.7% Exp to sick/dead chicken; 13 had low titers (5-20)	64
	800 villagers from a province with outbreak in 04–05	April–October, 2008	0 (0/800)	HI ≥40	65% Exp to poultry; detected elevated titers (HI ≥10) against H9N2 and H5N1	9
Indonesia	841 residents from villages with H5N1 outbreaks in birds	December, 2005; ca. 18 months after the first chicken outbreaks	0 (0/841)	WHO	Vaccination to birds in 2004 and 2005	65
	495 farmers from Sec III farms (small, family-run)	January–February, 2007; 1 bird outbreak in June 2006	0 (0/495)	WHO	11% reported fever and cough during previous 6 mo; 3% reported ILI; 1 w/ HI=20	66
Nigeria	295 PWs, 25 lab workers, all Exp to suspected infected poultry	March–April, 2006; ca. 1–2 months after H5N1 outbreaks in poultry	0 (0/320)	WHO	97% H3N2 pos; 1-2 participants had MN =40, 4–8 had MN=20 (Fig. 2 in reference)	33
South Korea	176 PWs, 1327 involved in culling poultry w/ suspected infection (GWs, soldiers, animal husbandry men, and civilians), 70 others (epidemiologists, public health officials, new reporters)	December, 2003–March, 2004; paired sample, on the day of culling completion in each region (exposure) and <4 weeks later.	0.36% (0/176, 9/1327, 0/70)	WHO	19 outbreaks in poultry; killed ca. 5 million poultry in infected farms and those w/in 3 km	67
Turkey	125 CCs (28 family contacts, 97 HCWs); 95 poultry cullers; 75 residents in outbreak regions	February, 2006; ca. 3–4.5 weeks after exposure	0.34% (1/295)	HI ≥20; MN ≥10	1 family contact had HI=20 and MN=10	68
Germany	97 participants (firemen, soldiers, etc.) collected wild birds w/ suspected infection	March, 2007; ca. 1 year after suspected exposure to H5N1 infected birds	0 (0/97)	PN or MN (>20)	5 pos by PN but neg by WHO-CC confirmatory testing	69
Israel	201 PWs and cullers worked during H5N1 outbreaks in poultry	March, 2006	0 (0/8)	Not specified	176 (87.6%) had full OP, the rest had partly OP; tested 8 paired sera	70

^*Populations at high risk are defined as those with recent suspected exposure to H5N1 from either poultry or H5N1 patients, or both; HI and MN titers are recorded as the number of dilution in this table. HCWs: health care workers; PWs: poultry workers; GWs: government workers; HHCs: household contacts; CCs: case contacts; Exp: exposed; Ctrls: controls; Pos: positive; Neg: negative; Obs: observation. Smp: sample; RI: respiratory illness; PN: plaque neutralization; OP: oseltamivir prophylaxis; LBM: live bird market; EE: environmental exposure.

Table 2.

Summary of seroprevalence of H5N1 virus infection in populations at high risk

Summary Region	Positive (% (pos/total))	Note/Risk factors	Referen- ces
Hong Kong, China	8.38 (179/2136)	Poultry are almost exclusively supplied by mainland of China; live bird markets are very common	13
Cambodia	1.39 (25/1805)	Backyard poultry raising is very common; live bird markets exist; seropositive participants were more likely to report bathing or swimming in community ponds	6
Mainland of China	1.40 (9/641) per WHO; 1.09 (43/3 942) per report	Backyard poultry raising is common in rural areas; live bird markets are very common in Southern China (especially in cities); national H5-vaccination in poultry started in 2005	7,50, 52, 54
Vietnam	0.94 (10/1 064) per WHO; 2.60 (47/1 811) per report	Included the controls, because they had high environmental exposure. Backyard poultry raising is very common; live bird markets exist; some villagers have no indoor water source	8
Thailand	0 (0/2 290)	Free-grazing poultry are common; live bird market is not practical; poultry are processed in slaughter houses	4,9, 10
Indonesia	0 (0/1 422)	More poultry are vaccinated; backyard poultry are common; live bird markets are common
South Korea	0.36 (9/2 512)		67
Nigeria	0 (0/320)		33
Turkey	0 (0/295) per WHO; 0.34 (1/295) per report		68
Germany	0 (0/97)		69
Israel	0 (0/8)		70

We hypothesize that the higher seropositive rate in HK versus other regions in Southeast Asia (SE Asia) is due to differing rates of exposure to poultry. Many of the regions in SE Asia with H5N1 outbreaks are rural. Exposure and interaction with both domestic and wild fowl is frequent in these regions due to the common rearing of backyard, free-grazing poultry, as well as sharing of community ponds with poultry.^4-6 These factors have been repeatedly identified as risk factors for H5N1 infection.^4,6-10 Additionally, environmental contamination with avian strains including H5N1 has been documented.^11,12 In contrast, in HK there is no backyard poultry rearing; rather, all poultry are imported, and most locals are only exposed to live birds at poultry markets.¹ Indeed, visiting a poultry market has been identified as the most significant risk factor for HK cases.¹³ We postulate that in rural SE Asia more frequent exposure to birds and the multiple avian influenza strains these birds may carry confer cross-protection against H5N1 and potentially other novel avian strains (e.g., H7N9). In HK, residents have less cross-protection and are more likely to experience severe H5N1-related illness, adaptive immune response, and the generation of H5N1-specific antibodies. Consequently, H5N1-specific seropositive rates are higher in HK than in rural SE Asia.

Our hypothesis is contingent on the idea that populations in agricultural areas are subject to more frequent, low-level exposure to avian influenza strains. Such environmental exposure may enhance the immune system of the local population against invasion by alternate avian strains. Specifically, cross-immunity against HPAI strains (e.g., H5N1) could be conferred by prior infections by more commonly existing low pathogenic avian influenza (LPAI) strains, such as H9N2.^9,14-16 If cross-immunity is sufficient to resolve mild infection, specific H5N1 antibodies might not be produced and thus not detected. Consequently, in regions with frequent environmental exposure, the population would be less likely to generate specific H5N1 antibodies. Seroprevalence surveys in these regions may thus yield low positive results for specific avian strains, in particular less common strains such as H5N1 (e.g., 0 in Thailand and Indonesia). Further, due to partial protection conferred by cross-immunity, a higher dose of H5N1 would be needed to overwhelm the immune system to cause severe illness in these populations. Populations in regions with frequent contact to poultry may thus appear less exposed to H5N1 and indeed may be less susceptible to H5N1 infection.

Conversely, most citizens in HK only have limited exposure to live birds at poultry markets; moreover, the live birds in HK markets were less likely to carry avian influenza strains due to stringent screening prior to importation. As a consequence, the immune systems of HK residents would be less prepared to deal with a novel avian strain than those of people in rural SE Asia. Specifically, with less exposure to avian influenza strains, LPAI antibodies in this population might not exist; and seasonal flu antibodies would remain at lower titers and be less likely to confer cross-immunity against HPAI. Further, due to the practice of live poultry trading and slaughter in the markets, the risk of exposure to HPAI is great when infected poultry are imported. This relatively low level of conferred cross-immunity combined with the sudden exposure to H5N1 might have contributed to the aggregate 17 cases in the 1997 outbreak.^1,13 For the same reason, specific antibodies would more likely be produced in response to an H5N1 invasion in the HK population. This may account for the higher seropositive rates among the HK participants as has been reported.^17-19

Immunologic evidence: cross-immune response against H5N1

Past studies have shown that antibodies induced by infections early in life can be preferentially reinforced later in life in response to new infection by antigenically related strains, a phenomenon termed original antigenic sin (OAS).^20,21 In contrast to OAS, which exists between variants of the same subtype, here we refer to cross-immunity as partial protection conferred across subtypes (e.g., H9N2, H5N1, H3N2). Numerous investigations, including serosurveys, vaccine trials in humans, and animal model experiments, provide evidence for this type of broader cross-protection (Table 3). These studies have shown that immunity conferred by prior infection or vaccination with human strains can cross-react with avian strains and be protective.^22,23 Such cross-immunity could be effected by different mechanisms, including neutralizing antibodies against either hemagglutinin (HA)^24-28 or neurominidase,^22,29 or memory T cells.^30-32 As described next, the development of cross-immunity appears to be affected by two aspects: (1) the frequency of encountering viral invasion and (2) the diversity of influenza strains encountered.

Table 3.

Studies indicating cross-reactivity/protection between different influenza subtypes

Subtypes	Experiment Summary	Mechanisms Proposed in the Study	Note	References
HA (stem) antibodies
H3 HA-specific against H1,  H2, H3, H5, H9, and H13	Generated a monoclonal antibody (MAb) with A/Aichi/2/68 (H3N2); this MAb was able to broadly react to H1, H2, H3, H5, H9, and H13 subtypes by ELISA, and showed neutralization and hemagglutination inhibition (HI) activities against particular strains of H1, H2, H3, and H13 subtypes; passive immunization of mice with this MAb provided heterosubtypic protection	Their molecular modeling study showed that the binding site was located on the globular head of the HA	While the HI test is considered a “gold standard” for subtyping, this study shows that, though rare, heterosubtypic HI activity does exist	24
H3 against H5, or other avian  strains	Tested 320 participants with exposure to suspected H5N1-infected poultry by microneutralization (MN); none with an H5 titer ≥1:80 (i.e., positive), but the majority had a titer of 1:10; 97% positive for H3N2, mostly with titers ≥1:320	-	This is the Nigeria study; the low titers against H5 might be due to anti-H3 cross- reactivity. H3 and H5 belong to different groups, so the cross-reactivity is more likely due to anti-H3 against H5 with low affinity. Alternatively, low titer of broad stem antibodies binding to both groups may be present	33
H2 against H9	H9 vaccine trial; By MN, 24 of 29 (83%) samples from individuals born before 1969 had reactive samples, whereas none of 31 individuals born after 1968 were reactive. By HI, 8 of 29 (28%) and none of 31 were reactive, respectively	People born before 1969 were exposed to H2N2 (which discontinued to circulate in human after 1968); H2 cross-reacts with H9	Group 1 includes H1, H2, H5, and H9; group 2 includes H3 and H7. Those shown positive by MN could be due to stem antibodies, while those positive by HI may indicate cross-activity between different HA heads	23
H3N2/H2N7 against H9	The prevalence of neutralizing antibodies to H9 (including H9N3, H9N2, and H9N7) decreased after the sera were pre-treated by H3N2 or H2N7 adsorption	Cross-reactivity between different influenza strains	H3 and H9 belong to different groups. The cross-activity between H3 and H9 could be due to broad stem antibodies binding to both groups, or H3/H2 antibodies binding to H9 with low affinity	47
Neurominidase (NA) antibodies  Human (H1)N1 aga ins t  avian N1 (7:1 reassortant;  PR8+NA [A/Vietnam/DT-  0361/2005 H5N1])	Tested NA activity of purified intravenous immunoglobulin (IVIg) containing IgG from >1000 blood donors, using neuraminidase inhibition (NI) assay; NI titers were shown to be 258–986 against H1N1, 1 309–3 274 against H3N2, and 143–231 against avian N1 reassortant. The IVIg preparations had low HI titers against H3N2 and H1N1 (about 20–40), but negative for H5N1	Cross-reactive antibodies against avian influenza virus A (H5N1), due to similar NA		29
Human (H1)N1 against avian  (H5)N1	Mice were immunized against the NA of a human H1N1 strain by DNA vaccination, then challenged with recombinant PR8-avN1 or H5N1 virus. Immunized mice were partially protected	Cross-reactive neuraminidase antibodies afford partial protection against H5N1 in mice		22
T cells
H3N2 against H5N1	Tested with mice and ferrets. Prior infection with H3N2/vaccination reduced clinical signs, body weight loss, mortality, and virus replication in the lungs as compared to naive mice/ferrets infected with H5N1	Cytotoxic T lymphocytes against H3N2 cross-reacted with H5N1, and they were selectively expanded upon challenge of H5N1, whicha likely accelerated clearance of H5N1		30, 31
Human strains against H5N1	Tested recognition of H5N1 internal protein by T cells isolated from healthy individuals from United Kingdom and Vietnam	Memory T cells established by seasonal human influenza A infection cross- react with H5N1		32

More frequent exposure may lead to stronger cross-immunity

Cross-immunity may be enhanced as people encounter additional strains, either by natural infection or by vaccination. An H9N2 vaccine trial²³ found that people with previous H2N2 infection possessed H2N2 antibodies that could neutralize H9N2, and that their titers for H2N2 increased after administration of H9N2 vaccine.

In a manner consistent with laboratory findings, natural exposure to avian strains might boost pre-existing immunity induced by human strain(s), which in turn could feed back and provide further protection against avian strains. A study investigated H5N1 seroprevalence in PWs and lab workers (LWs) with suspected exposure during multiple H5N1 outbreaks in poultry in Nigeria.³³ None of the participants were positive for H5N1 by WHO diagnosis criteria, although over 90% of them had H5N1 neutralizing antibody titers ≥1:10. However, 97% of participants were positive for H3N2, and the majority had high neutralizing titers (≥1:320). Clearly, these titers might have resulted from an H3N2 outbreak, but that outbreak would have to have been unusually comprehensive (97% positive for H3N2). Given the frequent contact with poultry among the PWs, and their likely repeated exposure to a variety of LPAI strains, it is possible the high H3N2 titers evident in this study merely reflect the presence of cross-protective antibodies that had been boosted during the culling event. That is, the MN titer of 1:10 in the majority of the workers may have stemmed from the high levels of H3N2 neutralizing antibodies due to the potential exposure to avian influenza. This line of reasoning begs the question: could this low MN titer against H5N1 be sufficiently protective, if the PWs were indeed infected by H5N1 or other avian strains?

It is not yet clear how strong immunity, be it specific or cross-reactive, needs to be to protect humans from H5N1 infection. Rockman et al³⁴ showed that ferrets with pre-exposure anti-H5 HI titers of 1:4–1:8, obtained from hyperimmune serum administration, only experienced slight weight loss after challenged with a lethal dose of H5N1. If a low level of antibody (e.g., those shown in the Rockman et al study) was protective, then the worker sera, with an anti-H5N1 MN titer of 1:10, might have been sufficient to clear the H5N1 infection, or at least reduce the number of cells infected such that the infection simply manifested as milder influenza-like illness (ILI). Indeed, 5% of PWs and 28% of LWs reported ILI, and about 16% of them reported cough and fever within the about 2-month period. Thus, the high rates of H3N2 seroprevalence and rates of ILI might be consistent with the proposal that the H3N2 antibodies were boosted upon encountering avian influenza, including LPAI, and that they conferred cross-immunity against these infections. Clearly, this idea is highly speculative; however, if cross-protective influenza antibodies are more prevalent than currently thought, the boosting of antibodies produced early in life, such as due to childhood H3N2 infection, upon infection with avian influenza strains (both HPAI and LPAI), might be commonplace.

As elaborated above and in the additional findings summarized in Table 3, it appears that exposure to avian influenza strain(s) may stimulate cross-subtype immunity and help maintain the titers of these cross-subtype antibodies at high levels. As a consequence, populations with daily contact with poultry would more likely have stronger cross-subtype immunity, thanks to frequent exposure to various avian strains. In addition, individual cross-subtype protection would likely be enhanced as a person ages and is more likely to suffer additional influenza infections (either human or avian strains or both). Immunity conferred after each recovery could provide additional cross-protection to avian strains, particularly in areas where poultry, wild fowl, and LPAI strains are commonly encountered. This development of increased cross-subtype protection with time could in part explain the relatively young age of H5N1 patients.

Exposure to strains of greater diversity may lead to broader cross-immunity

Recent studies reveal a new class of antibodies that can bind to the conserved HA stalk domain.^25-28 Based on antigenic properties and major structural features, influenza The viruses can be classified into two groups; group 1 includes human subtypes H1 and H2, avian subtypes H5 and H9, and another six subtypes; group 2 includes human subtype H3, avian subtype H7, and another four subtypes.³⁵ Stem antibodies that can broadly bind to viruses of multiple subtypes exclusively from the same group^25,26 or from both groups²⁷ have been isolated.

The discovery of stem antibodies has led to recent theories that cumulative cross-immunity from these antibodies may be responsible for the extinction of influenza virus subtypes^36,37 and may account for the seemingly stronger immunity against H5N1 in the older segment of population thanks to previous exposure to H2N2.³⁸ We further postulate that frequent contact with poultry may stimulate production of stem antibodies against H9, a more common LPAI strain in the same group with H5, which in turn confers cross-immunity against H5. Indeed, seroprevalence surveys have found higher seropositive rates of H9N2 than H5N1 in populations of multiple Asian countries.^9,14-16

Studies also indicate that only exposure to an antigenically dissimilar HA head can efficiently enhance the production of stem antibodies.^37,39 This finding suggests that people with contact to multiple distinct strains are likely to have higher titers of stem antibodies and strengthened broad protection. Consequently, people in rural SE Asia may be more likely to develop stem antibodies and further enhance their production due to frequent potential exposures to multiple avian strains.

Thus, if strain diversity is key to the development of stem antibodies and the HK population lacks frequent exposure to non-human influenza strains, then stem antibody production would not be enhanced. Additionally, phylogenetic analysis on sequences of H3N2 isolates indicates relatively low levels of genetic diversity in tropical SE Asia and HK.⁴⁰ The isolation of the HK population from poultry and wild birds along with the relative low diversity of human influenza strains could have increased the risk of severe infection by H5N1.

Taken together, environmental exposure to influenza strains, with both higher frequency and greater strain diversity, may lead to stronger and broader cross-immunity and less specific immunity against specific strains (e.g., H5N1). This is consistent with the serosurveys that H5N1 seropositive rates are lower in rural SE Asia and higher in industrialized HK.

Implications of our hypothesis

The interpretation of serosurvey data is challenging due to differences among study populations (e.g., differing anthropomorphic characteristics) and their living environments, the lack of standardized study design, the complexity of the serological assays, and the broad diversity of influenza viruses. These factors can all contribute to the varying H5N1 seroprevalence rates reported in different regions. To control for these variations, we limited our comparison to study cohorts with recent suspected exposure to H5N1 among populations in SE Asia where outbreaks of avian influenza in both poultry and human are more frequent. We also restricted our comparison to studies adhering to WHO diagnosis standards. Under these premises, we found that the seropositive rates reported for the HK cohorts were substantially higher than in other regions of SE Asia.

There have been a number of recent debates over whether studies based on the 1997 HK H5N1 outbreak should be included when estimating H5N1 exposure and case-fatality rates.^41-45 Indeed, some have argued that the 1997 HK outbreak is irrelevant due to the genetic dissimilarity of the 1997 strain from later H5N1 strains. It is possible, as was suggested in a recent review,⁴⁶ that strains of different genotypes may differ in infectivity and transmissibility characteristics. One might argue that this genetic dissimilarity is responsible for the higher seroprevalence associated with the HK outbreak. Here we have offered an alternate explanation for the high H5N1 seroprevalence in HK that is consistent with emerging epidemiological and immunological evidence. Because of the unique exposure status of residents in HK (i.e., lack of daily environmental exposure plus potential high-level exposure at the live poultry market), the studies associated with the 1997 outbreak instead appear to provide valuable insight into the interplay between avian strains and human immunity.

Our analysis suggests that the higher seroprevalence reported in the HK studies may be due to the greater susceptibility of the HK population, as opposed to those in SE Asia who typically have more frequent environmental exposure to avian influenza strains (This is true for both PWs and non-PWs). Further, these data indicate that susceptibility to H5N1, as well as other emerging avian strains, could be higher for populations without daily environmental exposure to poultry and wild birds. Many populations in other regions of the world, for example, the urban United States, have even more limited exposure to poultry and wild fowl (i.e., there are no live poultry markets). It is thus conceivable that infection rates might be higher than observed in SE Asia, were an HPAI (e.g., H5N1) virus to be introduced in these areas. However, this assumes no other cross-immunity gleaned from alternate virus exposure routes. Thus, despite a case-fatality rate for H5N1 that could be much lower than current estimates (due to underreporting of subclinical infections), the highly pathogenic nature and uncertainty of population susceptibility to the virus urges caution and better preparedness for the potential pandemic threat posed by H5N1 and other non-human strains.

On the other hand, our analysis suggests that populations with frequent contact with poultry could be partially protected due to cross-immunity conferred from exposure to other avian influenza strains. These individuals when exposed to H5N1 may manifest as asymptomatic or mild unreported infections. Current detection methods might miss these cases, as they may not account for immune mechanisms other than specific neutralizing antibodies. A more precise and complete picture of avian and swine influenza seroprevalence in the broader human population is warranted. Further investigations of individuals with subclinical infections could help better distill how the immune system combats viral invasion and the extent to which cross-immunity and stem antibodies confer protection against new infections.

Venues to test our hypothesis

A comprehensive understanding of the mechanisms underlying human immunity against avian influenza is still lacking; given this circumstance, evidence provided in all serosurveys, including those with relaxed diagnosis standards, should be carefully examined. We draw our hypothesis from both epidemiology and immunology studies, including information that might have previously been overlooked, for instance, potential cross-protection stimulated by frequent contact with poultry, and low antibody titers commonly found yet disregarded in serosurveys. The idea presented here is potentially controversial, and as of yet, epidemiological studies to conclusively support our hypothesis have not been conducted; however, existing data paint a partial picture that is consistent with our proposed idea.

To investigate our hypothesis, more studies of how cross-immunity works are needed that test a range of questions, including the following: (1) At what titer levels do circulating cross-reactive antibodies effectively confer protection against novel influenza infections? (2) What is the adaptive immune response in the presence of cross-reactive antibodies? For instance, does cross-protection down-regulate or preclude adaptive immune system generation of new HA-specific antibodies? (3) How can cross-protection best be identified in serosurvey studies? And how does this differ from current HI and MN assays? (4) How important are different arms of the immune system, both the innate and adaptive immune systems (both humoral and cell-mediated immunity components), against novel influenza infections? How do they correlate and regulate one another during different stages of infection and under various exposure/infection histories?

In addition, more in-depth epidemiological surveys on the prevalence of cross-immunity among populations in different regions are warranted to complement immunological studies. Specifically, our hypothesis can be tested through serosurveys of populations with differing exposure to poultry and wild fowl. As urbanization continues in SE Asia, an increasing new generation of people will reside in cities segregated from rural areas. These populations (e.g., citizens in big cities such as Shenzhen and Guangzhou in Southern China), similar to those in HK, would have limited exposure to poultry, such as when visiting live bird markets. Serosurveys on these populations if they were exposed to H5N1 (e.g., during an outbreak) would provide new data to test our hypothesis. Additionally, investigations could explore differences in the regional population prevalence of different immune effectors such as stem antibodies. For instance, we would expect a higher prevalence of stem antibodies in rural SE Asia, where exposure to backyard poultry and bird-invested waters is greater than in more industrial regions. Further, such investigation may shed light on how these broad antibodies function in humans, how potent each antibody isolate would be against exposures of varying dose, and how long they could persist.

Alternatively, cross-subtype immunity could be examined following outbreaks of seasonal influenza, in which neutralizing antibodies for both the outbreak strain as well as other subtypes, including LPAI, could be assayed more comprehensively. At the same time a re-evaluation of the detection methods, and even WHO standards, may provide more information on cross-immunity. For instance, a titer lower than the WHO diagnosis threshold may be due to mild infection, as it might result from cross-immunity stimulated by exposure to avian strains. Such possible inferences would have to be pursued in the context of known circulating strains, but could provide valuable further insight into rates of cross-immunity. Altogether, combining mechanisms revealed by immunological studies with epidemiological studies may unveil more efficient ways to combat a diversity of influenza viruses (either human or non-human strains).