Genome sequences of haemagglutinin cleavage site predict the pathogenicity phenotype of avian influenza virus: statistically validated data for facilitating rapid declarations and reducing reliance on in vivo testing

ABSTRACT

Based on the pathogenicity in chickens, most H1–H16 avian influenza viruses (AIV) cause mild diseases, whereas some of the H5 and H7 AI viruses cause severe, systemic disease. The number of basic amino acids in the haemagglutinin (HA) cleavage site of AIV plays a critical role in pathogenicity. As we gain a greater understanding of the molecular mechanisms of pathogenicity, genome sequencing of the HA0 cleavage site has assumed a greater role in assessment of the potential pathogenicity of H5 and H7 viruses. We validated the use of HA cleavage site motif analysis by comparing molecular pathotyping data against experimental in vivo (intravenous pathogenicity index [IVPI] and lethality) data for determination of both low pathogenicity and high pathogenicity AI virus declaration with the goal of expediting pathotype confirmation and further reducing the reliance on in vivo testing. Our data provide statistical support to the continued use of molecular determination of pathotype for AI viruses based on the HA cleavage site sequence in the absence of an in vivo study determination. This approach not only expedites the declaration process of highly pathogenic AIV (HPAIV) but also reduces the need for experimental in vivo testing of H5 and H7 viruses.

KEYWORDS:

• Avian influenza virus
• highly pathogenic avian influenza virus
• cleavage site
• haemagglutinin
• pathogenicity
• low pathogenic avian influenza virus

Introduction

Avian influenza viruses (AIV) attach to host cell receptors using the viral haemagglutinin (HA) protein to initiate the infection life cycle (reviewed in Swayne & Sims, 2021). After the virus enters the cell, the HA protein must be cleaved at the HA0 cleavage site by cellular proteases of the host to form functional subunits HA1 and HA2 (Bosch et al., 1979; Senne et al., 1996). The HA0 cleavage site of AIV can be classified as mono-basic (e.g. PQRETR/GLF) or multi-basic (e.g. PQRRRKKR/GLF) (Lee et al., 2021). The number of basic amino acids in the HA0 cleavage site plays a critical role in pathogenicity, determining which proteases can cleave HA and in which tissues the AIV can replicate. The mono-basic cleavage site usually contains less than two basic amino acids in the critical position and is cleaved by trypsin or trypsin-like proteases that limits virus replication of low pathogenic AIV (LPAIV) to principally the epithelial cells of the intestinal and respiratory tracts of birds. By contrast, the multi-basic cleavage site contains several basic amino acids in the critical position and is cleaved by several common cellular proteases that permit cleavage in most cells throughout the body, causing systemic disease and lethal infection of high pathogenicity AIV (HPAIV) in gallinaceous poultry species. All naturally occurring HPAIV to date have been H5 and H7 subtypes which have arisen from LPAIV via changes in the HA proteolytic cleavage site, including mutation of multiple non-basic to basic amino acids, loss of a shielding glycosylation site, duplication of basic amino acids, or recombination with insertion of cellular or viral amino acids (Swayne & Sims, 2021).

The classification of AIV into LPAIV (H1–16) and HPAIV (subset of H5 and H7) is currently officially based on experimental in vivo testing in young chickens through intravenous inoculation. At the First International Symposium on Avian Influenza held in 1981, it was resolved to define HPAIV based on the ability to produce > 75% mortality within 8 days in at least eight susceptible 4- to 8-week-old chickens inoculated by the intramuscular, intravenous, or caudal air sac route (Bankowski, 1981). The World Organisation for Animal Health (WOAH) definition of HPAIV is that HPAIV should have an intravenous pathogenicity index (IVPI) in 4-to 8-week-old chickens of > 1.2 or, as an alternative, cause at least 75% mortality in 4- to 8-week-old chickens inoculated experimentally intravenously (OIE, 2022a, 2022b). Using either approach, H5 and H7 viruses of low pathogenicity in chickens which do not have an IVPI of greater than 1.2 or cause less than 75% mortality in an intravenous lethality test, or have an IVPI of less than 1.2, should be sequenced to determine whether multiple basic amino acids are present at the cleavage site of the HA molecule. It is noted that field presentation should also be considered.

As we gain a greater understanding of the molecular mechanisms of pathogenicity, genome sequencing of the HA0 cleavage site has assumed a greater role in assessment of the potential pathogenicity of H5 and H7 viruses. In 2004, the 13th edition of the WOAH Manual of Diagnostic Tests and Vaccines for Terrestrial Animals stated that H5 and H7 AIV with IVPI of ≤ 1.2 or lethality of < 75% should additionally be sequenced to determine whether the cleavage site motif matched any previously reported for HPAIV, and, if so, the virus should be classified as HPAIV. In the EU, a similar definition of HPAIV was adopted in Commission Delegated Regulation (EU) 2020/689 of 17 December 2019 supplementing Regulation (EU) 2016/429 of the European Parliament and of the Council as follows: an influenza A virus of H5 and H7 subtypes or any influenza A virus with an IVPI > 1.2 or an influenza A virus of H5 and H7 subtypes with a sequence of multiple basic amino acids present at the cleavage site of the HA molecule that is similar to that observed for other HPAI isolates (Commission Delegated Regulation, 2019). OFFLU, the WOAH and Food and Agriculture Organization Animal Influenza Expert Network, maintains a haemagglutinin cleavage site document for H5 and H7 LPAIV and HPAIV to inform national veterinary authorities of accurate molecular and phenotypic data for timely decision-making processes (https://www.offlu.org/wp-content/uploads/2022/01/Influenza-A-Cleavage-Sites-Final-04-01-2022.pdf). The WOAH Terrestrial Manual specifies that determination of the cleavage site by sequencing, or a molecular method such as lineage-specific pathotyping assay, are preferred for initial pathogenicity assessment, and indicates this should be confirmed by either inoculation of specific pathogen-free (SPF) chickens or deep sequencing using high throughput sequencing to exclude the presence of HPAI virus (OIE, 2022a).

The purpose of this study was to validate the use of HA cleavage site motif analysis by comparing molecular pathotyping data against experimental in vivo (IVPI and lethality) data for determination of both LPAIV and HPAIV declaration with the goal of expediting pathotype confirmation and further reducing the reliance on in vivo testing.

Materials and methods

Pathotyping test data

The experimental in vivo pathotyping (IVPI and lethality) test data of 587 AIV, including 516 from the U.S. Department of Agriculture (USDA) and 71 from the Animal and Plant Health Agency (APHA), UK, were collected in this study, including 266 H5, 230 H7 and 91 other HA subtype viruses (online supplemental data 3 and 4). The in vivo experiments were conducted in SPF chickens according to the WOAH Terrestrial Manual (OIE, 2022a). The challenge studies and all experiments with live viruses were conducted in biosafety level 3 facilities at the National Veterinary Services Laboratories, Animal and Plant Health Inspection Service, USDA (Ames, IA, USA), and APHA (Addlestone, Surrey, UK) in accordance with approved institutional Animal Care and Use Protocols.

Literature search

We initially searched PubMed (database URL: http://www.ncbi.nlm.nih.gov/PubMed) from inception until December 2023 using the following terms: “avian influenza”, “pathogenicity”, “cleavage site”, “intravenous pathogenicity index”, and “chicken”. The search strategy was based on all fields. Only articles in the English language literature were evaluated. The searches were supplemented by checking the bibliographies of any eligible articles for additional references.

Dataset

The experimental in vivo pathotyping (IVPI and lethality) test data with no HA cleavage site sequences available were excluded from the analysis, including 20 H5 and 61 H7 viruses.

A total of 446 H5 and 198 H7 viruses were used in the analysis.

Statistical analysis

The confidence interval for each category was calculated to estimate the proportion with the binomial results, high or low pathogenicity using the Sample Size Calculators (available at https://www.sample-size.net/). Fisher's exact test was used for comparisons between cleavage site sequence motifs and phenotypes of LPAIVs and HPAIVs.

Results and discussion

None of the H5 (n = 253) and H7 (n = 148) viruses that were characterized as LPAIV based upon molecular pathotyping met the criteria for HPAIV during experimental in vivo testing in SPF chickens (Tables 1 and 2). For both subtypes, the Fisher exact test statistic values between molecular definition and experimental in vivo pathogenicity data of HPAIV and LPAIV were < 0.00001.

Table 1. Summary of molecular and in vivo pathotyping study data for H5 HPAIVs and LPAIVs.

Molecular determination		Cleavage site sequences	In vivo pathogenicity
			High	Low
HPAI	Gs/GD lineage H5Nx	PQRRRRKKR/G, PLREKRRKR/G, PLRERRRKR/G, PLRERRRRR/G, PLRGKRRKR/G, PPREKRRKR/G, PPRERRRKR/G, PQREKRKKR/G, PQRERRKKR/G, PQRERRRKR/G, LRRERRRKR/G, PLREKRRRKR/G, PQREKRRKKR/G, PQREGRRKKR/G, PQIEGRRRKR/G, PQGERRRKKR/G, PQIERRRRKR/G, PQRERRRKKR/G, PQSRRKKR/G	152 (98.06%, 95% CL: 94.45–99.60%)	3 (1.94%, 95% CL: 0.40–5.55%)
	Sporadic LPAI-to-HPAI mutation	PQKKKR/G, PQRKKR/G, PQKRKR/G, PQRRKKR/G, PQRKRKKR/G, PQRRRKKR/G, PQRKRKRKTR/G, PQRKRKTR/G, PQRETRRQKR/G, PERKRKTR/G	32 (84.21%, 95% CL: 68.75–93.98%)	6 (15.79%, 95% CL: 6.02–31.25%)
LPAI		PQRETR/G, PQKKTR/G, PQREIR/G, PQKETR/G, PQRKTR/G, PQTRTR/G	0 (0%, ^A97.5% CL: 0.0–1.45%)	253 (100%, ^A97.5% CL: 98.55–100.0%)

CL: confidence level.

^A One-sided 97.5% CL.

Table 2. Summary of molecular and in vivo pathotyping study data for H7 HPAIVs and LPAIVs.

Molecular determination	Cleavage site sequences	In vivo pathogenicity
		High	Low
HPAI	PETPKRRKR/G, PEIPKKRKR/G, PEIPKRKKR/G, PEIPKRRRR/G, PEIPRKRKR/G, PEIPKKKKR/G, PETPKRKRR/G, PETPKRRRR/G, PEIPKRRR/G, PEIPKKRKKR/G, PEIPKKREKR/G, PETPKRRERR/G, PETPKRKRKR/G, PEIPKKKKKKR/G, PENPKKRKTR/G	50 (93.10%, CL: 92.89–100%)	0 (0%, CL: 0.00–7.11%)
	PEIPKGRHRRPKR/G, PEIPKGSRVRR/G, PEKPKTCSPLSRCRETR/G, PEKPKTCSPLSRCRKTR/G, PELPKGTKPRPRR/G, PENPKDRKSRHRRTRTR/G, PENPKQACQKRMTR/G, PENPKQAHQKRMTR/G, PENPKQAYHKRMTR/G, PENPKQAYKKRMTR/G, PENPKQAYQKRMTR/G, PENPKQAYRKRMTR/G, PENPKTTKPRPRR/G, PENPRQAYRKRMTR/G
LPAI	PEIPKGR/G, PEKPKKR/G, PEIPKKR/G, PEKPKPR/G, PEKPKTR/G, PELPKRR/G, PENPKIR/G, PENPKNR/G, PENPKPR/G, PENPKTR/G, PENPNTR/G, PESPRPR/G, PQRETR/G, REPKPR/G	0 (0%, CL: 0.00–2.46%)	148 (100%, CL: 97.54–100.0%)

CL: one-sided 97.5% confidence level.

All H7 viruses that met the molecular criteria for classification as HPAIV (100%, n = 50) also met the criteria for HPAIV based upon experimental in vivo testing in SPF chickens (Table 2).

Of the H5 HPAIVs that met the molecular criteria for classification as HPAIV, the majority (95.8%, 184 of 192) also met the criteria for HPAIV based upon experimental in vivo testing in SPF chickens (Table 1). However, a total of eight H5 HPAIVs that met the molecular criteria for classification as HPAIV were of low pathogenicity for chickens upon experimental in vivo testing. The Gs/GD lineage of H5 HPAIV, first identified in a domestic goose in southern Guangdong province of China in 1996, has caused deaths in wild birds, poultry, and mammals. All of the Gs/GD lineage viruses in this study met the molecular criteria for classification as HPAIV and the majority (152 of 155) were of high pathogenicity for chickens upon experimental in vivo testing (Table 1) by Avian Influenza Reference Laboratories for the WOAH, Food and Agriculture Organization (FAO), or World Health Organization (WHO). The three Gs/GD lineage viruses that did not meet the criteria for high pathogenicity based upon experimental in vivo testing [A/duck/Hebei/0908/2009(H5N2), A/goose/Guangdong/2/96(H5N1), and A/chicken/Tibet/LZ01/2010(H5N2)] were reported by researchers. It is unclear why the molecular and in vivo pathotyping results did not align for these three viruses; however, the results were not confirmed by an official influenza reference laboratory. The corresponding authors of these publications were contacted to obtain more information without any reply.

For other H5 HPAIVs (excluding the Gs/GD lineage) the results of molecular and in vivo pathotyping were concordant among 84.2% (32 of 38; Table 1). Among the six discrepant viruses, (four from gallinaceous poultry and two from ostriches) all met the molecular criteria for HPAIV based upon cleavage site motif but did not meet the criteria for high pathogenicity based upon experimental in vivo testing (Table 3). The A/chicken/Pennsylvania/1/83(H5N2) virus had four basic amino acids at the HA cleavage site (PQKKKR/G) but the experimental in vivo testing did not meet HPAI criteria (Deshpande et al., 1987). In this case, the presence of N-linked glycosylation in the HA stalk influenced the pathogenicity by shielding the cleavage site, preventing cleavage by ubiquitous furin-like proteases. Other related H5N2 viruses of the same lineage that subsequently caused HPAI outbreaks possessed a single point mutation at position 13 of the HA1 that eliminated the glycosylation site conformationally close to the HA cleavage site. The loss of this glycosylation site at HA1 position 13 was exclusive to the isolates with high lethality. For A/turkey/England/87-92BFC/91(H5N1) (Alexander et al., 1993; Wood et al., 1994) and A/chicken/Texas/298313/04(H5N2) (Lee et al., 2005) there were closely related isolates or passages of the viruses with the same multi-basic amino acid motif that did meet the criteria for HPAIV based upon experimental in vivo testing in SPF chickens. The A/turkey/England/87-92BFC/91(H5N1) virus isolated from diseased turkeys caused no disease in experimentally inoculated chickens (IVPI = 0.0) despite having the same multi-basic amino acid motif (PQRKRKTR/G) as the other non Gs/GD H5N1 isolates that were classified as HPAIV based upon in vivo testing (IVPI = 3.0) (Wood et al., 1994). The two viruses from ostrich [A/ostrich/South Africa/AI1091/2006(H5N2) and A/ostrich/South Africa/AI2214/2011(H5N2) (Abolnik et al., 2012)] approached but did not meet the experimental in vivo criteria for HPAIV: the IVPI for A/ostrich/South Africa/AI2214/2011(H5N2) was 0.8, and was comparable to the IVPI of A/ostrich/South Africa/AI1091/2006(H5N2) which was 0.56 (Abolnik et al., 2009; Howerth et al., 2012). Pathogenicity testing of viruses recovered from ratites such as ostrich often may not meet the experimental in vivo criteria for HPAIV even when the molecular pathotyping criteria are met; in these studies, none of the chickens died and all recovered, which is a typical feature of ostrich derived HPAIVs which may indicate inadequate adaptation to chickens to infect, replicate and express the maximum pathogenicity. Recently, it has been reported that Mexican-lineage H5N2 with four basic amino acids at the HA cleavage site (PQKRKR/G) showed low pathogenicity in experimental in vivo testing, but the molecular mechanism remains uncertain (Xu et al., 2022).

Table 3. H5 viruses characterized as HPAIV based upon molecular pathotyping that did not meet HPAIV criteria based upon experimental in vivo testing.

Virus	Cleavage site sequences	Pathogenicity in chickens	Molecular mechanism
A/chicken/Pennsylvania/1/83(H5N2)	PQKKKR/G	Low	Presence of glycosylation site at HA1 position 13 masks the cleavage site
A/turkey/England/87-92BFC/91(H5N1)	PQRKRKTR/G	Low	Unknown
A/chicken/Texas/298313/04(H5N2)	PQRKKR/G	Low	Unknown
A/ostrich/South Africa/AI1091/2006(H5N2)	PQRRKKR/G	Low	Unknown^a
A/ostrich/South Africa/AI2214/2011(H5N2)	PQRRRKKR/G	Low	Unknown^a

^a Presumed host attenuation for Galliformes.

Discrepancy between genotype and phenotype may occur with co-infection of more than one strain (Quer et al., 2022) and where quasispecies are present during replication and mutation from LPAI to HPAI. Traditional Sanger sequencing does not allow the detection of a minority in a mixed virus population; therefore, these can be challenging to detect and analyse using Sanger sequencing methods which may underestimate the diversity and complexity of the virus population. All discrepant viruses in this study were H5 HPAIVs (4.7%, nine of 193) and all met the criteria for HPAIV based upon molecular pathotyping. The use of deep sequencing methods (e.g. next-generation sequencing) allows investigation for quasispecies populations with greater accuracy and sensitivity, particularly when conducted directly from the sample, versus on a virus isolate which could bias the population. Either LPAIV or HPAIV virus in a mixed virus sample can outcompete during virus isolation using embryonating eggs or in vivo pathotyping using chickens and become a dominant population. Therefore, deep sequencing of clinical samples can be used to identify and quantify the variations present and help resolve discrepancies by providing a high-resolution and unbiased view of the virus populations.

To date, HPAI has been characterized only from a selection of H5 and H7 subtypes. For other subtypes, HA cleavage site motif and experimental in vivo pathotyping test data of 217 LPAIV, including H1 (n = 22), H2 (n = 24), H3 (n = 18), H4 (n = 20), H6 (n = 49), H8 (n = 1), H9 (n = 50), H10 (n = 21), H11 (n = 5), H12 (n = 2), H13 (n = 4), and H15 (n = 1) were analysed in this study (online supplemental data 3). The majority (97.7%, 212 of 217) of other HA subtype viruses that have been characterized by both molecular and in vivo pathotyping demonstrate a mono-basic cleavage site with low pathogenicity (online supplemental data 3). It should be noted that five nephropathogenic strains of H10 AIVs that do not harbour multiple basic amino acids in the cleavage site of the HA protein (Swayne & Alexander, 1994; Si et al., 2022) have been shown to be highly lethal when chickens were inoculated intravenously (IVPI > 1.2) including two old H10 viruses (A/mandarin duck/Singapore/805/F-72/7/1993 and A/turkey/England/1979) that have PEIMQGR/G motif at the HA cleavage site and three wild bird origin H10 viruses identified from South Korea from 2012–2018 that have PELMQGR/G motif. The pathogenesis of these H10 viruses has been demonstrated to differ from the typical H5 and H7 HPAIVs as the lesions responsible for lethality of the former viruses were due to virus replication and necrosis in the kidneys of infected chickens, while the latter H5 and H7 HPAIVs produced systemic infection of diverse cell types (Swayne & Alexander, 1994).

Overall, these data provide statistical support to the continued use of molecular determination of pathotype for HPAIV and LPAIV based on the HA cleavage site sequence in the absence of an in vivo study determination. This approach not only expedites the declaration process of HPAIV but also reduces the need for experimental in vivo testing of H5 and H7 viruses. Deep sequencing methods should be used to further support findings and in vivo testing reserved for specific case scenarios, such as cleavage site changes of H5 and H7 subtypes, or where the field presentation includes high mortality or other signalments consistent with HPAIV, regardless of the HA subtype or molecular pathotype. Deep sequencing of all H5 and H7 viruses would be a valuable addition to routine diagnostics.

Disclosure statement

No potential conflict of interest was reported by the authors.

Additional information

Funding

Dong-Hun Lee is supported by the Korea Institute of Planning and Evaluation for Technology in Food, Agriculture, Forestry (IPET) through Animal Disease Management Technology Advancement Support Program, funded by Ministry of Agriculture, Food and Rural Affairs [grant number 122057-2]. Mia K. Torchetti, Mary Lea Killian, Ian Brown and David E. Swayne were funded through their respective institutions.