The Gs/Gd virus was initially isolated from geese in Guangdong Province, China in 1996. The Gs/Gd lineages are capable of zoonotic infections of human hosts via direct contact with infected birds. In Hong Kong SAR in 1997, a derivative of this strain caused the first fatal human case of H5N1 infection [11,12]. Southern China has often been proposed to be an epicenter for the generation of influenza pandemics, and indeed this region has continuously demonstrated the greatest H5N1 genetic diversity, suggesting that in addition of being the country of origin of H5N1 HPAI strains, Southern China is also an ongoing source of emergent and re-emergent HPAI viruses [13].

The circulation of over 20 different HPAI H5 and H7 subtypes has been documented in the last 100 years, however, only the Gs/Gd lineage has become so geographically widespread. Since 2000, HPAI H5N1 viruses have been repeatedly detected in Hong Kong SAR in live poultry markets and multiple novel genotypes have arisen through reassortment of Gs/Gd lineage viruses [4,14,15]. The analysis of 318 viruses isolated from 1996-2006 revealed that Gs/Gd lineage generated a total of 44 different reassortants in China [16].

The Hong Kong outbreak of 1997 was caused by a Hong Kong/156/97–like virus that derived its HA gene from Gs/Gd-like viruses and other genes from H9N2 or H6N1 viruses found in quail and other game birds sold in live markets [17,18]. During this outbreak, 18 humans were infected with H5N1 [2]. Six of these cases were fatal, giving a fatality rate of only 33%, which was much lower than subsequent H5N1 fatality rates. In response to the Hong Kong outbreak in 1997, millions of poultry livestock were culled, and as a result, a virus with the same gene constellation has not been detected again, indicating that the control measures used during this outbreak resulted in the eradication of this virus lineage.

H5N1 outbreaks re-emerged in poultry in 2003, and continued to spread throughout China in 2004, prompting a mass culling campaign to eradicate the virus. A total of 98 outbreaks in poultry were reported to the World Organization for Animal Health (or Office International des Epizooties, OIE) between 2003 and 2009 [19]. H5N1 has also been detected in wild birds in 2002 and during each winter season from 2004 to 2009 [5,20,21]. Outbreaks in poultry have again occurred in late 2008/early 2009 [5].

The H5N1 virus was not detected in humans again until February 2003 when two cases were reported in Hong Kong in a family who had recently travelled to Fujian province. Additionally, in November 2003, a human case was identified in Beijing [2]. Since late 2005, human cases have been reported from time to time in several Chinese provinces, and to date a total of 38 human cases have been confirmed in China, including 25 fatalities [2].

In 2005, hundreds of wild migratory birds became infected with H5N1, causing mass die-offs in Qinghai Lake in central China [22-25]. Since 2005, only two other outbreaks have been reported in Qinghai province [5]. In April 2006, the H5N1 virus was detected in wild birds, but did not significantly spread. Interestingly, the latest outbreak was reported on the 17th May, 2009 in Qinghai province, at Genggahu Lake. In this outbreak, 121 wild birds were found dead, and 600 backyard poultry were subsequently culled.

H5N1 virus was introduced to Vietnam in late 2003 from Yunnan province, China, causing multiple poultry outbreaks and finally establishing itself endemically in this population [4,8,26] Outbreaks in poultry were first reported in 2004 and shortly followed by several human cases. Infections in humans were continuously reported throughout 2004-2005, with 65 cases reported during this period which at the time far outnumbered any other country [3]. Vaccination of poultry against H5 virus was initiated in Vietnam in August 2005, with inactivated H5N1 and H5N2 vaccines [27]. In efforts to control or eradicate H5N1 in poultry and to prevent human exposure, Vietnam has strengthened their vaccination campaign and associated surveillance programs in poultry, and improved poultry import controls and virological surveillance. Following the vaccination campaign, no influenza outbreaks were reported in poultry or humans from December 2005 until August 2006 [3]. The virus reappeared in poultry but no human cases were reported again until June 2007 and subsequently new human cases were regularly reported until the current time. At the time of writing, the total number of human infections detected in Vietnam was 111 with 56 mortalities, representing nearly one third of worldwide laboratory-confirmed human H5N1 infections [3,28]. Among 50 countries reporting H5N1 avian influenza in domestic poultry to the OIE, Vietnam has reported the greatest number, with 2539 outbreaks declared [5].

Avian influenza H5N1 virus was first reported in Thailand in January 2004 [29]. Several outbreaks were reported during the 2004-2005 period across many Thai provinces [29,30], prompting the government of Thailand to implement strict control and prevention measures against HPAI. The primary objectives were to identify geographic areas with confirmed H5N1 disease in poultry (e.g. 5-kilometer radius around an infected flock); to establish controls on the transport of poultry and poultry products out of affected areas, and to promote safe food-handling practices [29,31,32]. Small epidemics were reported in backyard flocks in 2006. However, following the tighter controls introduced during the same year, outbreak numbers and sizes were greatly reduced, leading to a much smaller natural reservoir which persists through the dry months [33]. Sporadic outbreaks continued to occur until the beginning of 2008 [34]. H5N1 transmission is seasonal in Thailand, as there have been several reports of backyard chicken deaths during the wet and cooler months of 2006/07 and 2007/08 [5,29,30,32-34].

Thailand has experienced 1141 outbreaks of H5N1 avian influenza in domestic poultry between 2003 and 2009 [19]. Approximately 63 million birds in three outbreaks were culled to prevent the further spread of the infection [30]. To date, 25 laboratory-confirmed human cases have been reported, including 17 fatalities [3]. The last human case of H5N1 was confirmed in September 2006 in a North-Eastern Thai province [3].

In Cambodia the first confirmed outbreak of HPAI subtype H5N1 in poultry was reported during January 2004, and was followed by 14 other H5N1 outbreaks during the same year [35]. Interestingly, a retrospective investigation also revealed that between December 2003 through January 2004, a HPAI H5N1 outbreak occurred in a wildlife center in Takeo province, infecting several captive wild birds species, and cats. During the first four months of 2005, four fatal human H5N1 cases were detected in Kampot province, South East Cambodia, coinciding with deaths among poultry in this province [3]. In 2006, four outbreaks were detected in domestic poultry, and two human cases were reported [36,37]. In April 2007, Cambodia confirmed the seventh fatal human case of H5N1 infection from Kampong Cham province. The most recent case of human H5N1 infection (8th human case) was also the first non-fatal case and occurred in December 2008 in Kandal province. From 2004 to date, Cambodia officially reported 21 outbreaks of H5N1 avian influenza in domestic poultry [19]. The control strategies in poultry are based exclusively on culling of infected flocks, and the prohibition of poultry imports [35].

H5N1 avian influenza was detected in poultry in Lao PDR in 2003. A mass culling campaign was rapidly conducted, resulting in the loss of approximately 155,000 poultry. Surveillance of live bird markets during 2005 and 2006 failed to detect the virus, or serological evidence of exposure to the virus, suggesting that the virus from the initial outbreak had been eradicated [38]. In February 2006, avian influenza H5N1 virus was isolated from healthy ducks at a farm in Vientiane. Two human cases (both fatal) were reported in 2007 from Vientiane Province [2] In 2008 and early 2009, Lao PDR reported new poultry outbreaks of H5N1, however, no human cases were declared [3]. Since 2003, a total of 18 H5N1 outbreaks in poultry have been reported by Lao PDR authorities [19].

Myanmar reported its first outbreak of H5N1 virus in poultry in March 2006, and subsequent outbreaks were reported in February, October, November, and December 2007 [3]. The sole human case of H5N1 infection (non-fatal) was reported in December 2007, in the Shan State province where an outbreak in domestic birds occurred at the same time. In March 2008, H5N1 sero-positive ducks were detected during routine surveillance conducted in the Shan State province [3]. To date, the country reported a total of 93 H5N1 avian influenza outbreaks in domestic poultry [19].

Indonesia has suffered the heaviest burden of avian influenza, with 141 human infections, of which 115 were fatal [2]. Indonesia first detected H5N1 outbreaks in poultry in December 2003, and then continuously through to August 2006. H5N1 virus is now endemic in many Indonesian islands, such as Java, Sumatra and Sulawesi. To date, 261 outbreaks in poultry were reported in the country [19]. The first human case was reported in July 2005, and then human infections continued to occur frequently until the current time. Three family clusters of H5N1 infection were identified in 2005 [39]. In May 2006, Indonesia reported a large family cluster involving 7 family members and human-to-human transmission was suspected [40].

Malaysia reported outbreaks its first outbreaks in poultry in August and September 2004. Additional outbreaks were reported in February and March 2006 in free-range poultry flocks and then again in chickens in June 2007. No human cases were detected in Malaysia [41]. Early detection and drastic culling measures could be attributed with such a successful control of H5N1 outbreaks in this country. A total of 16 H5N1 outbreaks in poultry were reported by Malaysian authorities [19].

The 1997 H5N1 outbreak in humans in Hong Kong resulted in a fatality rate of 33%. Since the re-emergence of H5N1 in 2003, this rate has been considerably increasing in most affected countries [2]. Of 429 confirmed human cases worldwide, 61% were fatal. South East Asia recorded 326 of these cases, including 222 fatalities, giving a fatality rate of 68% [2,42]. This rate is variable within each country. For example, China’s overall fatality rate is 66%, Indonesia’s 82%, Thailand’s 68%, Vietnam’s 50% (Table 1). The high fatality rate in Indonesia is striking, given the significant number of cases that have been reported there. Other South East Asian countries, such as Cambodia, Lao PDR, Myanmar, have small numbers of confirmed human cases (8, 2, 1, respectively), which means interpreting their fatality rates (88%, 100%, 0%, respectively) is difficult.

It is unclear whether H5N1 viruses in certain geographical regions differ in their pathogenicity. It would seem that clade 2.1 viruses (Indonesia) are more pathogenic than clade 1 viruses (Cambodia/Thailand/Vietnam) and 2.3 viruses (China) [43]. Drawing a link between H5N1 clade circulation and fatality rates is difficult, due to differences in health care practices and duration between onset of illness and treatment. Affordability of healthcare in developing countries and reduced access to anti-viral treatment at an early stage of the disease may partially explain increased fatality rates. Where surveillance is lacking, there is the possibility for under-representation of fatality rates, however, subclinical infections may also be under-estimated. Indeed, out of more than 600 blood-tested Cambodian villagers from areas where two children died in 2006, seven were seropositive for H5 antibodies indicative of asymptomatic infection. Thus, in this scenario the fatality rate was much lower than originally determined [44].

Diverse populations of endemic HPAI H5N1 viruses have continuously evolved in South East Asia, from 2003 until the present day, and several gene modifications have occurred within these viruses which may affect their transmissibility and pathogenicity. To become a pandemic strain, a H5N1 virus must be able to be efficiently transmitted between human hosts, a feature that existing H5N1 viruses have not yet acquired.

The HA is the main influenza antigen and determines cell binding, host range, and neutralizing antibody response. Thus, some modifications on the HA gene could provide the virus with advantageous properties. Several key amino acid residues on the HA have been identified.

Avian influenza viruses are defined as HPAI or LPAI viruses, based on the severity of the disease which they cause. HPAI viruses are highly contagious amongst poultry, and often result in a high fatality rate, especially in terrestrial poultry like chicken, quail or turkey (up to 100% in 2-3 days). LPAI viruses are responsible for mild diseases, with few or no symptoms in some bird species. The defining feature of HPAI is a multi-basic amino acid motif at the cleavage site present in the HA gene (PQRERRRKKR/G), which confers increased pathogenicity. LPAI viruses have a single arginine at the HA cleavage site, which can subsequently be cleaved only by trypsine-like proteases. Since these proteases are present in a restricted number of organs, the infection is usually limited to the respiratory or to the intestinal tract, without becoming systemic. In contrast, the presence of multiple basic amino acids allows the hemagglutinin to be cleaved by many different proteases, enabling a broader tissue tropism, and the ability to cause systemic infections. Some H5N1 strains have alterations on this site, such as Arg (R) or Lys (K) deletions or insertions. Alterations of this kind were observed, among others, on Cambodian, Indonesian, Thai and Vietnamese H5N1 isolates [8,49,50]. Although the real impact of such changes in the HA cleavage site is difficult to estimate, the high frequency of these occurrences highlights the importance of conscientious surveillance of this site.

The relevant residues involved in antigenic sites on the H5 HA have been described, and compared to H3 antigenic sites which have been comprehensively characterized [69,70]. The host immune pressure can induce mutations on these antigenic sites, which results in the emergence of immune-escape mutants. Therefore, a positive selective pressure on those sites can rapidly lead to ineffective host immune responses, which are a major constraint for the development of human or animal vaccines. In Vietnam and Indonesia, among virus isolated from August 2003 to June 2005, a positive selective pressure has been reported on 8 residues of the HA gene. Five of those residues were located on antigenic sites A and E (positions 83, 86, 138, 140 and 141), two of them were suggested to be involved in receptor binding (positions 129 and 175). The last one was a potential N-linked glycosylation site (position 156) [8].

The HA of human influenza viruses preferentially bind to sialic acid (SA) receptors linked to galactose (Gal) through an α-2,6 linkage (SA α-2,6Gal), whereas avian influenza viruses preferentially bind to SA receptors of the α-2,3 linkage (SA α-2,3Gal). The HA sites which bind these receptors are often under selective evolutionary pressure. Mutations of these binding sites from an avian to human receptor binding preference could facilitate human-to-human transmission and these sites have been proposed as markers for assessing the pandemic potential of H5N1 isolates [71]. Two key residues have been identified to determine receptor preference switching – position 226 and 228 in H2 and H3 isolates, which corresponds to 222 and 224 in the H5 gene [72]. However, the mutations Gln222Leu and Gly224Ser which are believed to enhance the affinity of the virus for the SA α-2,6Gal have never been observed in H5N1 field isolates. Nevertheless, several naturally occurring mutations were associated with an enhanced affinity of the avian virus to “human type” receptors SA α-2,6Gal. Mutations at position 182 (Asn182Lys) and 192 (Gln192Arg) can independently switch the receptor binding preference from avian to human [73]. Several other combinations of mutations were found to have a variable effect on receptor switching, certainly indicating that in the right circumstance a virus may adapt to bind human receptors more efficiently.

H5N1 HPAI viruses from clade 2.1 (found in Indonesia) may be under a lower positive selective pressure compared to the other clades [74]. Although such an observation is difficult to interpret, there may be a reduced necessity for the virus to evolve and adapt in Indonesia due to the high endemicity of the disease. The persistence in poultry reduces the need for the virus to jump from one species to another, especially to mammalian species, in order to maintain the chain of transmission [74].

The neuraminidase enzyme cleaves HA from sialic acids, facilitating the release of newly assembled viral particles from the host cell surface and thus enabling spread to other cells. Interestingly, all H5N1 viruses which have been isolated post-2003 contain a deletion in the stalk region of the protein, at positions 49-68 for clade 1 and 2 viruses, and at positions 54-72 for clade 3 viruses [75]. This deletion reduces the enzymatic activity of the NA, and may act to balance with the HA which has a reduced affinity interaction with receptors in poultry, compared to aquatic birds [76].

The NA is a target for the neuraminidase inhibitor (NAI) class of antiviral drugs, including Oseltamivir (Tamiflu^TM) and Zanamivir (Relenza^TM) which are commercially available. H5N1 viruses are typically sensitive to NAI’s , and these drugs are used as first line treatment for H5N1 infections in humans, and are a major component of pandemic planning as they are also recommended for chemoprophylaxis [77]. The His274Tyr mutation is associated with resistance to Oseltamivir. This mutation is now increasingly prevalent in seasonal H1N1 and H3N2 influenza strains, partly due to widespread use of Oseltamivir [77]. In addition to His274Tyr, amino acid substitutions Arg292Lys, Glu119Val and Asn294Ser have been associated with resistance or reduced sensitivity to Oseltamivir [78].

There have been several reports illustrating the development of Oseltamivir resistance during the treatment of H5N1-infected patients. Some patients developed the His274Tyr mutation, others the Asn294Ser mutation associated with a decreased sensitivity to NAI’s [79,80]. To assess the threat of a resistant H5N1 strain persisting with equal fitness and replicability, the H5N1 virus A/Vietnam/1203/2004 was engineered to possess several NA resistance mutations including His274Tyr, Glu119Val, Arg292Lys, and Asn294Ser. Viruses with His274Tyr and Asn294Ser were found to retain their infectivity and pathogenicity, highlighting the importance of ongoing surveillance of Oseltamivir resistance for early detection of resistant viruses [81].

In the absence of resistance mutations, genetic drift mutations also give rise to variations in Oseltamivir sensitivity [82,83]. Mutations which are remote from the active site may not confer absolute resistance, but may affect the dosage of drugs required to treat these infections. Suboptimal dosing may lead to the emergence of fully resistant viruses. Clade 2.3.4 viruses in Vietnam were recently found to be 8-fold less susceptible to Oseltamivir, but maintain their susceptibility to the adamantanes (M2 inhibitor drugs) [56]. Treatment combining NAI and adamantanes is recommended for patients infected with adamantane-sensitive H5N1 strains [56,84]. Ongoing evaluation of the evolution of drug sensitivity in both seasonal and H5N1 influenza viruses is vital in understanding the best treatment options in the wake of rapid virus evolution.

The M gene encodes two capsid proteins (M1 and M2). M1 is a RNA-binding protein, and M2 is a membrane ion channel protein involved in H⁺ proton transport, which facilitates the acidification of the endosome and the release of the viral particle inside the cytoplasm [85]. A comparison of synonymous and non-synonymous nucleotide substitutions in virus isolates from Indonesia and Vietnam (2003-2005) provided evidence of positive natural selection on the M2 gene [8]. Mutations on the M2 protein can result in different adaptative processes which affect the virus' capacity to uncoat during the early stages of the infection, and modulate its virulence. Also, mutation on the M2 gene can lead to adaptation to the different pH environments of aquatic or terrestrial hosts [8]. Moreover, the M2 protein is the target of the adamantane class of antiviral drugs (amantadine, rimantadine). Two mutations on this protein (Leu26Ile and Ser31Asn) lead to drug resistance, and these mutations are present in all genotype Z, clade 1 viruses [86]. Clade 2.3.4 viruses from Indonesia and Vietnam remain sensitive to amantadine [56], and clade 2.1 and 2.2 viruses have varying rates of resistance to this class of drug [86,87]. Thus, monitoring M2 mutations in H5N1 viruses remains an important aspect of avian influenza surveillance.

The polymerase complex of influenza viruses is composed of three subunits: PB1, PB2, and PA. These subunits, along with the NP protein and the RNA genome, form the ribo-nucleoproteic complex, necessary for the protection of the genome, as well as the replication and transcription processes. The PB1 gene also encodes the PB1-F2 protein by an alternative reading frame. Analyses conducted on viruses from Vietnam and Indonesia isolated between 2003 to 2005 showed that several sites on the PB1-F2 gene were under positive natural selection, although the significance of these particular sites is not known [8]. In mouse models and in cell lines, the PB1-F2 protein can influence viral pathogenicity, by sensitizing host cells to apoptotic stimuli (e.g. TNFα), thereby promoting apoptosis during infection [88,89]. However, the residues involved in this process have not yet been identified.

Mutations at residue 627 on the PB2 protein are associated with increased virulence and are thought to be relevant to the adaptation of H5N1 viruses to human hosts. The 627 residue on PB2 is a glutamic acid (Glu) in avian strains [76]. The Glu627Lys mutation on the PB2 improves replication efficiency, enhances adaptation of HPAI viruses to mammalian hosts, and increases transmission between mammalian hosts in experimental models [90-92]. The 627Lys mutation enhances growth at lower temperatures, consistent with those found in the human upper respiratory tract (around 33°C), compared to wild type viruses with 627Glu which grow optimally at 41°C [93,94]. Within clade 1 or clade 2.1 viruses, the Glu627Lys mutation is observed in some human strains but not in avian strains, suggesting a selective advantage for these isolates in mammalian hosts [8,95]. In clade 2.2 viruses, the Glu627Lys mutation is observed in both human and avian strains [47]. Interestingly, most of the strains isolated in birds from the Qinghai Lake outbreaks in 2005 (clade 2.2), which spread to Europe and Africa, contained the 627Lys mutation [60,96].

In addition to residue 627, several other changes on the PB2 protein and other proteins of the polymerase complex also contribute to adaptation to mammalian hosts and to regulation of virulence [90,97]. Positive selective pressure was also detected at several sites of unknown function, on PB2 and PA genes from Indonesian strains isolated from 2003 to 2007 [39]. Thus, surveillance of the substitutions occurring on the polymerase genes, and further elucidation of the importance of PB2 and PA genes in virus evolution is of ongoing interest.

The non-structural protein NS1 can regulate H5N1 virus virulence in humans by modulating the host immune response. The severity of human infections with H5N1 virus in 1997 was partly due to the resistance of these viruses to interferons and TNFα during the host immune response. The presence of glutamic acid at residue 92 on the NS1 was vital for these effects [98,99]. However 92Glu has not been identified in more recent human H5N1 strains. Ser42 and Ala149 can also inhibit immune response signaling pathways, including induction of interferons [100,101]. On the contrary, deletions in the NS1 gene can attenuate virulence [102]. The NS1 carboxy-terminal PDZ ligand binding motif is also a potential virulence factor. This motif binds cellular PDZ-containing proteins, disrupting a range of cellular signaling pathways [103]. Therefore, monitoring the evolution of the NS1 gene is important for the detection of novel viruses with increased pathogenicity in humans.