Risk Assessment of Fifth-Wave H7N9 Influenza A Viruses in Mammalian Models

The potential pandemic risk posed by avian influenza H7N9 viruses was heightened during the fifth epidemic wave in China due to the sudden increase in the number of human infections and the emergence of antigenically distinct LPAI and HPAI H7N9 viruses. In this study, a group of fifth-wave HPAI and LPAI viruses was evaluated for its ability to infect, cause disease, and transmit in small-animal models. The ability of HPAI H7N9 viruses to cause more severe disease and to replicate in brain tissues in animal models as well as their ability to fuse at a lower pH threshold than LPAI H7N9 viruses suggests that the fifth-wave H7N9 viruses have evolved to acquire novel traits with the potential to pose a higher risk to humans. Although the fifth-wave H7N9 viruses have not yet gained the ability to transmit efficiently by air, continuous surveillance and risk assessment remain essential parts of our pandemic preparedness efforts.

ABSTRACT

The fifth wave of the H7N9 influenza epidemic in China was distinguished by a sudden increase in human infections, an extended geographic distribution, and the emergence of highly pathogenic avian influenza (HPAI) viruses. Genetically, some H7N9 viruses from the fifth wave have acquired novel amino acid changes at positions involved in mammalian adaptation, antigenicity, and hemagglutinin cleavability. Here, several human low-pathogenic avian influenza (LPAI) and HPAI H7N9 virus isolates from the fifth epidemic wave were assessed for their pathogenicity and transmissibility in mammalian models, as well as their ability to replicate in human airway epithelial cells. We found that an LPAI virus exhibited a similar capacity to replicate and cause disease in two animal species as viruses from previous waves. In contrast, HPAI H7N9 viruses possessed enhanced virulence, causing greater lethargy and mortality, with an extended tropism for brain tissues in both ferret and mouse models. These HPAI viruses also showed signs of adaptation to mammalian hosts by acquiring the ability to fuse at a lower pH threshold than other H7N9 viruses. All of the fifth-wave H7N9 viruses were able to transmit among cohoused ferrets but exhibited a limited capacity to transmit by respiratory droplets, and deep sequencing analysis revealed that the H7N9 viruses sampled after transmission showed a reduced amount of minor variants. Taken together, we conclude that the fifth-wave HPAI H7N9 viruses have gained the ability to cause enhanced disease in mammalian models and with further adaptation may acquire the ability to cause an H7N9 pandemic.

IMPORTANCE The potential pandemic risk posed by avian influenza H7N9 viruses was heightened during the fifth epidemic wave in China due to the sudden increase in the number of human infections and the emergence of antigenically distinct LPAI and HPAI H7N9 viruses. In this study, a group of fifth-wave HPAI and LPAI viruses was evaluated for its ability to infect, cause disease, and transmit in small-animal models. The ability of HPAI H7N9 viruses to cause more severe disease and to replicate in brain tissues in animal models as well as their ability to fuse at a lower pH threshold than LPAI H7N9 viruses suggests that the fifth-wave H7N9 viruses have evolved to acquire novel traits with the potential to pose a higher risk to humans. Although the fifth-wave H7N9 viruses have not yet gained the ability to transmit efficiently by air, continuous surveillance and risk assessment remain essential parts of our pandemic preparedness efforts.

INTRODUCTION

Low-pathogenic avian influenza (LPAI) A(H7N9) viruses have caused annual epidemics of sporadic human infections in China since their first detection in humans in March 2013 (1). To date, China has experienced six waves of H7N9 outbreaks (the seventh wave is ongoing), with over 1,560 laboratory-confirmed human infections, including more than 600 deaths, reported to the World Health Organization (WHO) (2, 3). The fifth wave of the H7N9 epidemic (October 2016 to September 2017) in China caused 764 documented cases of human infection, including 281 deaths, among geographically extended regions, making it the largest H7N9 epidemic wave to date. In February 2017, human infections with highly pathogenic avian influenza (HPAI) H7N9 viruses possessing polybasic amino acid insertions at the hemagglutinin (HA) cleavage site were identified for the first time. These HPAI H7N9 viruses have resulted in 31 cases of human infection (4), among which 16 cases have been fatal (http://www.chinaivdc.cn/cnic/en/). Detection of H7N9 viruses in poultry from live bird markets and commercial farms in several provinces of China has resulted in the culling of thousands of birds (5). Although the epidemiological and clinical characteristics of human infection during the fifth epidemic wave remain similar to those of previous waves (6), it is unclear whether fifth-wave LPAI or HPAI H7N9 viruses pose a relatively higher threat to public health.

The H7N9 influenza viruses that have emerged in China are novel reassortants with surface genes from wild bird H7, N9 subtype viruses and internal genes from domestic poultry H9N2 viruses (7, 8). Since their first appearance in 2013, H7N9 viruses have continued to evolve and reassort, becoming enzootic in poultry in China (8). During the fifth epidemic wave, H7N9 viruses underwent considerable genetic changes, with the HA diverging into two phylogenetic lineages: the Yangtze River Delta (YRD) lineage and the Pearl River Delta (PRD) lineage. The majority of the fifth-wave influenza viruses fall into the YRD lineage, which is phylogenetically distinct from progenitor viruses represented by A/Anhui/1/2013 (Anhui/1) virus (6, 9). All HPAI H7N9 viruses detected to date fall in the YRD lineage and are clustered together, suggesting a single common ancestor (5). Serological studies have demonstrated that the fifth-wave H7N9 viruses from the YRD lineage have reduced cross-reactivity with ferret antiserum raised against the 2013 candidate influenza vaccine virus (CVV), which is derived from Anhui/1 and has prompted the WHO to update its recommendation for the H7N9 CVV composition (10).

Before the fifth epidemic wave, all H7N9 viruses were low pathogenic in chickens, causing no or mild clinical symptoms in infected poultry (11). However, unlike most other LPAI viruses, these H7N9 viruses could cause severe disease in humans and other mammalian species, indicating their intrinsic fitness in mammalian hosts (1, 12). Sequence analyses have revealed that most H7N9 viruses already possess certain mammalian adaptation markers. These include G186V and Q226L/I (H3 numbering) in the HA, which result in dual receptor binding to both avian-like (α2,3-linked) and human-like (α2,6-linked) sialic acids (12–14), and E627K in the PB2 protein (14), which contributes to efficient viral replication in the lower temperature of the mammalian upper respiratory tract (15). In addition to maintaining key mammalian adaptation residues shared with H7N9 viruses from previous waves, some fifth-wave H7N9 viruses have acquired additional virulence markers, such as PB2-A588V, which has been shown to enhance the pathogenicity of H10N8, H7N9, and H9N2 avian influenza viruses in mice (16). Furthermore, the majority (45/46) of fifth-wave human HPAI H7N9 virus isolates deposited in the Global Initiative on Sharing All Influenza Data (GISAID) database have acquired a 4-amino acid insertion (-KRTA-) and a substitution at G320R (H7 numbering) in the HA, resulting in a consensus sequence of -PEVPKRKRTA(R/G) (basic residues are underlined) in the HA cleavage site. The presence of four consecutive basic residues (-KRKR-) in the HA cleavage site of H7 viruses is a strong indicator of enhanced virulence in chickens; however, it remains unclear whether HPAI H7N9 viruses represent a higher risk to humans.

Despite the high number of human infections with H7N9 viruses during the fifth wave, sustained human-to-human transmission has not been documented (3). The inability of H7N9 viruses from previous waves to transmit efficiently in humans is in agreement with their limited airborne transmission between ferrets (12, 17, 18). However, considering the recent genetic and antigenic changes to these viruses during the fifth wave, it is necessary to conduct updated risk assessments for H7N9 viruses. In our study, representative H7N9 human isolates from the fifth wave were assessed for their virulence and transmissibility in both mouse and ferret models. We found that the HPAI H7N9 viruses are capable of causing a greater level of disease in both animal species with enhanced neurological involvement but transmit only among cohoused ferrets, whereas the LPAI H7N9 viruses exhibit pathogenicity and transmissibility traits similar to those of H7N9 influenza viruses from prior waves.

RESULTS

Genetic characteristics of representative fifth-wave H7N9 viruses.

H7N9 viruses have continued to evolve and acquire amino acid changes since their first emergence in China in 2013. Here, five representative fifth-wave H7N9 human isolates, inclusive of both LPAI (A/Hong Kong/4553/2016 [HK/4553], A/Hong Kong/125/2017 [HK/125], and A/Hong Kong/61/2016 [HK/61]) and HPAI (A/Guangdong/17SF003/2016 [GD/16] and A/Taiwan/1/2017 [TW/17]) viruses, were selected for sequence analysis with a focus on the residues that have been previously identified to be involved in receptor binding, mammalian adaptation, antiviral resistance, and HA cleavability (Table 1). Among the selected viruses, four H7N9 viruses (HK/4553, HK/125, GD/16, TW/17) from the YRD lineage and one virus (HK/61) from the PRD lineage share 96% to 100% similarities in the HA and neuraminidase (NA) proteins and more than 92% sequence similarity across the other major viral proteins (PB2, PB1, PA, NP, M1, NS1). Similar to the majority of H7N9 viruses (15), all H7N9 viruses analyzed in this study bear S138A, T160A, and G186V (H3 numbering) in the HA, which have been shown to enhance receptor binding for α2,6-linked sialic acids (19–22). However, the variability at position 226 between LPAI viruses (bearing human-like 226L) and HPAI viruses (bearing avian-like 226Q) indicates potential differences in receptor binding specificity among fifth-wave H7N9 viruses. In addition, TW/17 virus has an I126T (H3 numbering) mutation, which generates an N-linked glycosylation motif [N-X-(T/S)] at the asparagine (N) 124 residue. Compared to LPAI H7N9 viruses, HPAI H7N9 viruses have acquired a polybasic cleavage site (-KRKRT-) in the HA, indicative of enhanced pathogenicity in chickens. Examination of the PB2 protein identified the presence of a mammalian adaptation substitution (E627K) in two of the fifth-wave H7N9 viruses (HK/4553, TW/17) tested in this study; none of the viruses tested acquired the human adaptation marker, 701N. The HPAI H7N9 virus TW/17 possessed additional amino acid changes not present among other isolates, including the PB2-K526R substitution, which was previously shown to enhance the effects of E627K on influenza virus replication (23), and the NA-R294K (N9 numbering) mutation, which confers resistance to oseltamivir (24, 25). Collectively, while the genetic sequences of most fifth-wave H7N9 viruses remain largely similar to those of viruses from previous waves, select viruses have displayed variations in key mammalian adaptation residues, the cleavage site, and antiviral resistance markers.

TABLE 1.

Virus characterization and genetic analysis of fifth-wave A(H7N9) viruses

Virus full name	Virus name in this study	Epidemic wave/ lineagea	Patient, outcome of infection	Cell/egg passage history	GISAID identifier	HA cleavage siteb	Amino acid at the indicated position
							HA (H3 numbering)						PB2					NA
							126	138	160	186	226	228	526	588	627	701	702	294
A/Hong Kong/61/2016	HK/61	5th/PRD	70-yr-old male, fatal	C2/E1	240520	PEIPKG(R/G)	I	A	A	V	L	G	K	V	E	D	K	R
A/Hong Kong/4553/2016	HK/4553	5th/YRD	75-yr-old male, fatal	C2/E2	239994	PEIPKG(R/G)	I	A	A	V	L	G	K	V	K	D	K	R
A/Hong Kong/125/2017	HK/125	5th/YRD	62-yr-old male, recovered	C1/E2	256599	PEIPKG(R/G)	I	A	A	V	L	G	K	V	E	D	R	R
A/Guangdong/17SF003/2016	GD/16	5th/YRD	56-yr-old male, fatal	E3	266963	PEVPKRKRTA(R/G)	I	A	A	V	Q	G	K	V	E	D	R	R
A/Taiwan/1/2017	TW/17	5th/YRD	69-yr-old male, recovered	E2	248778	PEVPKRKRTA(R/G)	T	A	A	V	Q	G	R	A	K	D	K	K
A/Anhui/1/2013	Anhui/1	1st/NA	Adult female, fatal	E1	138739	PEIPKG(R/G)	I	A	A	V	L	G	K	A	K	D	K	R

^aYRD, Yangtze River Delta lineage; PRD, Pearl River Delta lineage; NA, not applicable.

^bUnderlined residues are the basic residues at the HA cleavage site of HPAI H7N9 viruses.

Replication efficiency of fifth-wave H7N9 viruses in Calu-3 cells.

To evaluate H7N9 virus replication in epithelial cells, the primary site for replication in mammals, we inoculated cells of the immortalized human airway epithelial cell line Calu-3 with a low dose for evaluation of multicycle replication kinetics. Compared to precursor avian H7 and H9 viruses, LPAI H7N9 viruses have demonstrated an enhanced ability to replicate in human airway epithelial Calu-3 cells, suggestive of mammalian host adaptation (26). The recent amino acid changes associated with select fifth-wave H7N9 viruses prompted us to compare their growth kinetics in this cell type. Calu-3 cells were inoculated at a multiplicity of infection (MOI) of 0.01 with fifth-wave H7N9 viruses, with the first-wave Anhui/1 virus serving as a control, and were cultured for 72 h at 37°C. All fifth-wave H7N9 viruses showed robust replication, with viral titers ranging 10^6.2 to 10^8.0 50% egg infectious doses (EID₅₀)/ml at 24 h postinoculation (p.i.), but these levels were significantly lower (P < 0.01) than the mean titer exhibited by Anhui/1 virus (10^9.5 EID₅₀/ml) at this time point (Fig. 1). However, all of the H7N9 viruses, including Anhui/1 virus, reached comparable titers of ≥10^9.3 EID₅₀/ml by 48 h, and the titers remained at similar levels at 72 h p.i. In summary, the fifth-wave H7N9 viruses have the ability to replicate efficiently in human airway epithelial cells, indicating that H7N9 viruses have retained their fitness for mammalian respiratory tract cells, despite recent genetic and antigenic divergence.

FIG 1.

Replication kinetics of fifth-wave A(H7N9) viruses in Calu-3 cells. Polarized Calu-3 cells grown on Transwell inserts were inoculated in triplicate with the H7N9 viruses shown at an MOI of 0.01. Cell supernatants were collected at 2, 24, 48, and 72 h postinoculation and were titrated in eggs. Viral titers, expressed as the log₁₀ EID₅₀ per milliliter, were plotted as the mean with standard deviation (SD). Statistical analysis was performed using two-way analysis of variance with GraphPad Prism software. **, P < 0.01 between Anhui/1 and all other H7N9 viruses. The detection limit was 1.5 log₁₀ EID₅₀/ml.

HPAI H7N9 viruses exhibit enhanced virulence and extended viral tropism in mice compared with LPAI H7N9 viruses.

LPAI H7N9 viruses from previous waves can replicate efficiently in mouse pulmonary tissues and cause lethal disease without prior adaptation (12). To compare the replicative ability and virulence of fifth-wave H7N9 viruses with those of viruses from previous waves, two LPAI viruses inclusive of both phylogenetically distinct YDR and PDR lineages (HK/4553 and HK/61, respectively) and two HPAI viruses (GD/16, TW/17) were intranasally (i.n.) inoculated into mice to monitor morbidity, mortality, and viral replication. Similar to LPAI H7N9 viruses from previous waves, HK/4553 and HK/61 viruses were able to cause severe disease in mice at high inoculation doses (50% mouse lethal dose [MLD₅₀], 4.8 and 5.5 log₁₀ EID₅₀, respectively; Table 2), comparable to that of Anhui/1 virus, which possessed an MLD₅₀ of 3.4 to 3.5 log₁₀ PFU (equivalent to 4.5 log₁₀ EID₅₀ based on stock virus titration) (12, 27). Both viruses exhibited robust viral replication in the murine lung at day 3 p.i., with mean viral titers of 10^6.5 and 10^4.0 EID₅₀/ml (10^3.0 EID₅₀ dose) and 10^8.5 and 10^7.8 EID₅₀/ml (10^6.0 EID₅₀ dose), respectively (Table 2). Viral replication persisted in the lungs at least until day 6 p.i. with mean viral titers of 10^6.8 EID₅₀/ml or higher at either inoculation dose. Additionally, both fifth-wave LPAI H7N9 viruses tested exhibited comparable titers (10^5.8 to 10^6.8 EID₅₀/ml) in the nose on days 3 and 6 p.i. at an inoculation dose of 10^6.0 EID₅₀. However, similar to Anhui/1 virus (12), both fifth-wave LPAI H7N9 viruses failed to consistently spread to the brains of infected mice, as only low levels of virus (≤10^2.5 EID₅₀/ml) were detected in a single mouse from each virus group.

TABLE 2.

Viral replication and pathogenicity of A(H7N9) viruses in mice

Virus	% wt lossa	MLD50b	Virus titer after inoculation withc
			103.0 EID50		106.0 EID50
			Lung (3)	Lung (6)	Lung (3)	Lung (6)	Nose (3)	Nose (6)	Brain (6)
HK/4553	25.0 (5)	4.8	6.5 ± 1.0	6.1 ± 0.5	8.5 ± 0	7.7 ± 0.7	5.9 ± 0.8	5.8 ± 0.4	2.5 (1/3)
HK/61	23.4 (6)	5.5	4.0 ± 1.1	6.9 ± 0.6	7.8 ± 0	6.8 ± 0.3	6.8 ± 0.4	6.3 ± 0.4	2.3 (1/3)
GD/16	23.3 (6)	3.2	5.4 ± 0.5	5.5 ± 0 (2/3)	7.7 ± 0.1	6.8 ± 0	5.8 ± 0.3	4.3 ± 0.4	4.1 ± 0.5
TW/17	25.0 (4)	ND	6.8 ± 0.4	6.8 ± 0.5	7.9 ± 0.2	7.5 ± 0d	4.7 ± 0.1	4.2 ± 0.3d	4.2 ± 0.3d

^aPercent mean maximum weight loss (five mice per group) following i.n. inoculation with 10^6.0 EID₅₀ of virus. The day that mean maximum weight loss was observed for each group is shown in parentheses.

^bThe 50% mouse lethal dose (MLD₅₀) is expressed as the log₁₀ EID₅₀ required to give a value of 1 for MLD₅₀. ND, not determined.

^cMean virus titers in mice (three mice per group) inoculated with 10^3.0 or 10^6.0 EID₅₀ of virus and the tissues shown were collected on day 3 or 6 p.i., as indicated in parentheses in the subheads. Virus titers are expressed as the mean log₁₀ EID₅₀ per milliliter ± standard deviation for three mice unless specified in parentheses in the table body, in which the number of mice with positive virus detection/total number of mice in the group is shown. The limit of detection was 1.5 log₁₀ EID₅₀/ml.

^dAll three animals were euthanized on day 5 p.i. due to the severity of disease.

HPAI H7N9 virus TW/17 replicated to higher titers than HPAI GD/16 virus in murine lung tissue on both day 3 and day 6 p.i. when mice were given a low-dose inoculum (10^3.0 EID₅₀) and at the later time point when a higher dose (10^6.0 EID₅₀) was administered (Table 2). GD/16 virus produced increased mortality in mice compared to the fifth-wave LPAI viruses (MLD₅₀, 10^3.2 EID₅₀). However, lower levels of HPAI TW/17 virus were detected in the nose of this animal model at day 3 and 5 p.i. compared with the levels of the LPAI viruses detected at days 3 and 6 p.i; all mice inoculated with 10^6.0 EID₅₀ of TW/17 virus required euthanasia before the scheduled endpoint at day 6 due to the severity of disease. Evaluation of brain tissue from infected mice revealed an increase in the detection of GD/16 and TW/17 viruses. All mice had virus present in brain tissues at day 5 (TW/17 virus) or 6 (GD/16 virus) p.i. with mean titers of 10^4.2 or 10^4.1 EID₅₀/ml, respectively, indicating an extended viral tropism to murine brain tissue for HPAI H7N9 viruses (Table 2). Taken together, the LPAI H7N9 viruses from the fifth wave do not show signs of increased pathogenicity in mice; however, recently emerged HPAI H7N9 viruses exhibit in this species enhanced virulence with higher lethality and a greater propensity to spread to extrapulmonary tissues.

HPAI H7N9 viruses cause severe disease with neurological involvement in ferrets.

Ferrets have been extensively employed to evaluate the pandemic potential of newly emerged influenza A viruses (28). The LPAI H7N9 viruses from previous waves were able to cause moderate weight loss and replicate efficiently in both the upper and lower respiratory tracts of ferrets (12, 26). To evaluate the virulence of fifth-wave H7N9 viruses, we i.n. inoculated 6 ferrets each with 10^6.0 EID₅₀ of the H7N9 viruses and monitored them for morbidity and disease progression. Similar to ferrets inoculated with the LPAI H7N9 viruses from previous waves (12, 26), HK/4553 virus-inoculated ferrets exhibited moderate clinical signs of disease, including transient fever (mean peak of 1.9°C above the baseline), intermittent sneezing and nasal discharge, and modest lethargy with a relative inactivity index (RII) score of 1.3 (Table 3). Inoculated ferrets lost approximately 10% of their preinoculation body weight, on average, but all survived the infection. In contrast, the HPAI H7N9 GD/16 and TW/17 viruses caused more severe disease; inoculated ferrets exhibited greater anorexia, dyspnea, and morbidity (mean maximal weight loss, 18.8% and 17.1%, respectively) with an RII score of 1.8 (Table 3). Additionally, 4/6 and 5/6 ferrets inoculated with GD/16 or TW/17 viruses, respectively, developed mild to moderate neurological symptoms, including hind-limb paresis and torticollis, and required euthanasia (Table 3). These pathogenic outcomes are characteristic of those observed in ferrets infected by HPAI H7N7 and H5N1 viruses (29–31).

TABLE 3.

Comparison of A(H7N9) virus pathogenesis and transmission in ferrets

Virus	Inoculated ferrets					Direct contact transmission model				Respiratory droplet transmission model
	Mean peak NW titera	% wt lossb	Feverc	RIId	Lethalitye	Virus detection in NW	Seroconversionf	Mean peak NW titer	Lethality	Virus detection in NW	Seroconversion	Mean peak NW titer	Lethality
HK/4553	7.7 (1–5)	10.2	1.9	1.3	0/6	3/3	3/3 (640–1,280)	7.3 (3–5)	0/3	1/3	1/3 (320)	8.3 (day 5)	0/3
GD/16	7.1 (1–5)	18.8	1.9	1.8	4/6 (8–13)	3/3	3/3 (80–160)	7.0 (5–7)	0/3	0/3	0/3	NDg	0/3
TW/17	6.5 (1–5)	17.1	2.0	1.8	5/6 (6–10)	2/3	2/2 (40–160)	4.5 (3)	1/3	0/3	0/3	ND	0/3

^aMean peak viral titers in nasal washes (NW) are expressed as log₁₀ EID₅₀ per milliliter. The range of days over which peak titers were exhibited is shown in parentheses. The limit of detection was 1.5 log₁₀ EID₅₀/ml.

^bPercent mean maximum weight loss for six inoculated ferrets.

^cMean maximum temperature rise over the baseline temperatures, which ranged from 38.1 to 39.9°C.

^dRII, relative inactivity index through day 1 to 14 p.i., which was calculated as previously described (30).

^eThe number of ferrets requiring euthanasia due to severe disease/total number of ferrets is shown. The range of days over which the ferrets were euthanized is shown in parentheses.

^fThe number of ferrets that seroconverted/total number of ferrets is shown. The range of HI titers is shown in parentheses.

^gND, not detected.

To better understand the neurological manifestations associated with fifth-wave H7N9 virus infection in ferrets, the brain tissues collected on day 3 p.i. from animals infected by LPAI HK/4553 virus or HPAI GD/16 virus were prepared for histopathological evaluation. We observed no prominent parenchymal, perivascular, or meningeal inflammation at this time point in ferrets infected by LPAI or HPAI virus (Fig. 2A and B). However, a mild, focal inflammatory infiltrate was found in the choroid plexus of ferrets infected by the HPAI virus (Fig. 2D), but such histopathological findings were not observed in ferrets infected by LPAI H7N9 viruses (Fig. 2C). Because the time of onset of neurological signs in ferrets ranged from days 6 to 9, we next evaluated neurological tissues collected at later time points from HPAI TW/17-inoculated ferrets, which were euthanized based on neurological symptoms. At days 6 and 9 p.i., the euthanized ferrets still maintained viral titers of 10^4.5 EID₅₀/g in brain tissues for both animals. We observed a spectrum of inflammation and tissue damage in the central nervous system of ferrets infected by HPAI TW/17 virus on days 3, 6, and 9 p.i. The degrees of inflammation and the severity of damage increased proportionally with the duration of infection. Day 3 p.i. showed scattered mild inflammatory infiltrate in the parenchyma (Fig. 3A) and choroid plexus (Fig. 3B). Day 6 p.i. showed multiple foci of a moderate inflammatory infiltrate in the meninges and parenchyma (Fig. 3C). Neuronophagia, neuronal necrosis, and glial nodules were also observed at this time point (Fig. 3D). Day 9 p.i. showed extensive hemorrhage and inflammatory infiltrate in the meninges, perivascular areas, and parenchyma (Fig. 3E and F). Taken together, our data indicate that the HPAI H7N9 viruses cause more severe disease and pathology in the ferret model than the LPAI H7N9 viruses. The HPAI H7N9 virus caused a spectrum of meningoencephalitis in accordance with the duration of infection, demonstrating a progression of neurological disease over the course of infection that ultimately resulted in a lethal outcome for the majority of infected animals. Further histopathological evaluation of H7N9 viruses in mammalian models may help us to better understand the potential for severe disease in humans.

FIG 2.

Hematoxylin and eosin staining of brain tissue collected from ferrets 3 days after LPAI HK4553 or HPAI GD/16 virus inoculation. (A, B) Representative parenchymal, perivascular, and meningeal tissues showing no prominent inflammation in ferrets infected with LPAI virus (A) or HPAI virus (B). (C, D) Choroid plexus of ferrets infected by LPAI, showing no overt inflammation (C), or HPAI virus, showing mild, focal inflammatory infiltrate (D). Bars, 200 µm (A and B) and 60 µm (C and D).

FIG 3.

Hematoxylin and eosin staining of brain tissue collected from ferrets 3, 6, and 9 days after HPAI TW/17 virus inoculation. (A, B) On day 3 p.i., a scattered mild inflammatory infiltrate was observed in the parenchyma (A) and choroid plexus (B). (C, D) Multiple foci of moderate inflammatory infiltrate in the meninges and parenchyma (C) and neuronophagia, neuronal necrosis, and glial nodules (D) were noted on day 6 p.i. By day 9 p.i., histopathology progressed to include extensive hemorrhage and inflammatory infiltrate in the meninges, perivascular areas, and parenchyma (E, F). Bars, 200 µm (A to D and F) and 300 μm (E).

Systemic spread of fifth-wave H7N9 viruses in ferrets.

To evaluate the tissue tropism of fifth-wave H7N9 influenza viruses in ferrets, we collected pulmonary and extrapulmonary tissues 3 days after inoculation and measured the level of infectious virus present in each tissue. Despite differences in disease severity, all three H7N9 viruses were found at comparable titers throughout the ferret respiratory tract; mean titers were approximately 10^7.0 EID₅₀/ml in nasal turbinates, 10^5.8 to 10^6.7 EID₅₀/g in trachea samples, and 10^5.6 to 10^7.1 EID₅₀/g in lung tissues (Fig. 4). However, in support of the histopathological findings, LPAI and HPAI H7N9 viruses exhibited substantial differences in their ability to spread to the brain. HPAI virus was detected at moderate levels in the olfactory bulb region of the brain of 3/3 and 2/3 ferrets infected by GD/16 and TW/17 viruses, respectively, and in the anterior and posterior brain tissue of all HPAI virus-infected ferrets (Fig. 4). In contrast, LPAI virus (HK/4553) was not detected in any region of the brain from any of the infected animals. This disparity between HPAI and LPAI viruses held true for intestinal tissues as well; all ferrets infected by GD/16 virus and 1/3 ferrets infected by TW/17 virus had titers of 10^2.9 to 10^3.8 EID₅₀/g in this tissue, but virus was not detected in the intestines from HK/4553 virus-infected ferrets. Ocular tissues were also evaluated for the presence of infectious virus following HPAI virus infection; infectious virus was detected in the eye and conjunctiva from GD/16 virus-infected ferrets (10^2.6 and 10^3.1 EID₅₀/ml, respectively) but not following inoculation with TW/17 virus (Fig. 4). Other extrapulmonary tissues and body fluids that were tested (spleen, kidney, liver, and blood) contained no detectable infectious H7N9 influenza virus. Therefore, despite the fact that fifth-wave LPAI and HPAI H7N9 viruses were found at high titers in both the upper and lower respiratory tracts of ferrets, only the HPAI viruses achieved consistent, moderate titers in the brain and olfactory bulb, which is indicative of an extended viral tropism for this emerging group of influenza viruses.

FIG 4.

Fifth-wave A(H7N9) virus replication in ferrets. Groups of 3 ferrets each were i.n. inoculated with 10^6.0 EID₅₀ of the H7N9 virus, as indicated, and were euthanized on day 3 p.i. to collect ferret tissue samples, including those of the nasal turbinates (NT), trachea, lung, olfactory bulb (OB), brain (pooled posterior and anterior brain), intestine (pooled duodenum, jejunum, and ileum), eye (pooled left and right), conjunctiva (pooled left and right), spleen, kidney, and liver, as well as blood samples. The presence of infectious viruses in the tissue and blood samples was determined by viral titration in eggs. Mean viral titers + SD are expressed as the log₁₀ EID₅₀ per gram or milliliter (for NT, eye, conjunctiva, and blood samples only). When virus was not detected in all three inoculated ferrets, the number of positive samples out of the total number is shown on top of the bar graph. The dashed line indicates the detection limit of 1.5 log₁₀ EID₅₀ per gram or milliliter. Statistical analysis was performed using two-way analysis of variance with the Tukey multiple-comparison test with GraphPad Prism software if all three ferret tissue samples from each group were virus positive. *, P < 0.05; **, P < 0.01. ND, not determined.

Fifth-wave H7N9 virus replication and transmissibility in ferrets.

We next examined the level of virus shedding in nasal washes and rectal swabs to determine the replication kinetics of fifth-wave H7N9 viruses in ferrets. Inoculated ferrets shed virus in nasal washes that peaked at levels ranging from 10^6.4 to 10^7.7 EID₅₀/ml (Table 3). By day 7 p.i., virus persisted in all surviving ferrets infected by HPAI H7N9 viruses (10^2.5 to 10^6.3 EID₅₀/ml) but was cleared in half of the LPAI virus-infected animals; titers of 10^2.3 to 10^3.8 EID₅₀/ml were detected in the animals still shedding HK/4553 virus at this time point (Fig. 5, left side of each panel). Like the LPAI H7N9 viruses from previous waves (12, 26), both LPAI and HPAI H7N9 viruses from the fifth wave were sporadically detected in rectal swab samples (1/6 ferrets from the LPAI H7N9 virus-infected group and 2/6 ferrets from each HPAI H7N9 virus-infected group) at low titers (≤10^3.5 EID₅₀/ml).

FIG 5.

Fifth-wave A(H7N9) virus transmission in ferrets. Groups of 6 ferrets each were i.n. inoculated with 10^6.0 EID₅₀ of the H7N9 virus shown. Each ferret was paired with an individual naive ferret at 24 h p.i. Animals used in the direct contact model (A) were cohoused in the same cage, and those used in the respiratory droplet model (B) were housed in separate, adjacent cages with perforated side walls that prevented contact but that allowed air exchange. Viral titers in nasal washes from individual inoculated ferrets (left) and contact ferrets (right), with bar color indicating animal pairs, on the days p.i. or p.c. shown are presented as the log₁₀ EID₅₀ per milliliter. Nasal wash specimens collected in advance of the scheduled time point due to required euthanasia are indicated d6 and d8 (days 6 and 8, respectively). The detection limit was 1.5 log₁₀ EID₅₀/ml.

Ferrets continue to represent one of the best small-animal models for evaluating influenza virus transmission; therefore, we used the previously established direct contact (DC) and respiratory droplet (RD) ferret transmission models to assess the transmissibility of fifth-wave H7N9 viruses. H7N9 viruses from prior waves were previously shown to transmit efficiently in the DC model but exhibit limited transmission by RD (12, 18, 26). In the same experimental setting, we found that the fifth-wave H7N9 viruses also demonstrated a capacity to transmit efficiently among cohoused ferrets, as 3/3 contact ferrets from the LPAI HK/4553- and HPAI GD/16-infected virus groups shed virus at titers similar to those for inoculated ferrets (Fig. 5A). TW/17 virus was detected in 2/3 contact ferrets at peak titers of 10^3.3 and 10^5.8 EID₅₀/ml (Fig. 5A); the ferret that shed 3 logs of virus required euthanasia on day 8 postcontact (p.c.) due to the development of neurological symptoms. However, by 21 days p.c., the two remaining ferrets from this group had seroconverted to homologous TW/17 virus (Table 3). In an RD transmission setting, one contact ferret from the LPAI HK/4553 virus-infected group shed virus in nasal washes that peaked on day 5 p.c. at 10^8.3 EID₅₀/ml. None of the contact ferrets from the HPAI GD/16 or TW/17 virus-infected groups shed detectable virus in nasal wash specimens, and all remained seronegative against homologous viruses at the end of the experiment (Fig. 5B; Table 3). Our findings show that fifth-wave LPAI and HPAI H7N9 viruses have not acquired the ability to transmit efficiently through the air among ferrets.

Fifth-wave H7N9 virus threshold pH for fusion.

The pH spectrum for HA activation has been recognized to contribute to influenza virus host adaptation and transmission (32). H7N9 viruses from previous waves have exhibited high pH thresholds (5.6 to 5.8) for fusion compared to most human seasonal influenza viruses (5.0 to 5.5) (26, 33, 34). To determine if high pH thresholds are also reflected in fifth-wave H7N9 viruses, we measured the highest pH at which syncytium formation was observed in 50% or more of HA-expressing Vero cells. Similar to the previously characterized Anhui/1 virus, the LPAI HK/61 virus, from the PRD lineage, triggered fusion at pH 5.8 or lower (Fig. 6). LPAI viruses from the YRD lineage (HK/125 and HK/4553) had a slightly reduced threshold for fusion, as syncytia were observed in at least 50% of cells at a pH of up to 5.7. Interestingly, the two HPAI H7N9 viruses examined (GD/16 and TW/17) could induce 50% or greater syncytium formations at pH 5.4 or lower, indicating that the pH threshold for the fusion of fifth-wave HPAI viruses was reduced to a level similar to that for seasonal influenza viruses. Our findings show that the fifth-wave HPAI H7N9 viruses have evolved and acquired the ability to fuse at a lower pH threshold than H7N9 viruses from previous waves. Further adaptation to the human host will increase the risk of the emergence of an H7N9 virus capable of causing a pandemic.

FIG 6.

Influenza virus fusion threshold, as measured by syncytium formation in Vero cells. Vero cell cultures were inoculated with each H7N9 virus at an MOI of 1 for 16 h. Infected cells were treated with 5 μl/ml of TPCK-trypsin for 15 min at 37°C and then incubated with fusion buffer with pH values ranging from 5.0 to 6.0. The fusion threshold was defined as the highest pH value at which 50% or more syncytia among NP-positive cells could be visualized (left). The syncytia at 0.1 pH unit higher than the fusion threshold are shown on the right as a control.

Fifth-wave H7N9 viruses isolated from ferrets after transmission display a reduced population of minor variants.

To monitor H7N9 virus subpopulations during virus replication and transmission in ferrets and to identify potential mammalian adaptive mutations, deep sequencing analysis was performed on virus inocula and a limited set of tissues and nasal wash specimens collected from selected paired sets of inoculated and contact ferrets infected by LPAI HK/4553 or HPAI TW/17 virus (NCBI SRA accession number PRJNA494222). Nasal wash specimens containing the largest amount of infectious virus were chosen for analysis. We found minor variants (above a 5% frequency threshold) representing nonsynonymous changes at a number of positions within nearly all of the HK/4553 and TW/17 viral proteins (Table 4; Fig. 7). When virus subpopulations found in ferret specimens were compared to the population in the inoculum, two nasal wash samples (both collected from HK/4553 virus-infected contact animals) were found to contain virus with a complete switch or a partial switch to a different major variant at position HA-493, NA-39, or PB2-508 (Fig. 7). The RD contact animal shed virus containing a complete switch at HA-493 (E to K) and PB2-508 (R to K), while the DC animal shed virus possessing an NA-39 major variant substitution (P to T). Three different positions were found to be either completely or partially switched in infected ferret samples compared to the sequence of the TW/17 virus inoculum. PA-192 and PA-X-191 were switched (R to S and F to L, respectively) in the contact nasal wash and trachea and olfactory bulb tissue samples, and NP-89 was switched (H to P) in trachea and lung tissue samples. When the virus subpopulations isolated during transmission experiments from the nasal washes of paired inoculated and contact ferrets were compared with each other, influenza viruses isolated from the contact ferrets were found to be more homogeneous than the viruses isolated from the paired inoculated ferrets (Table 4; Fig. 7). During the HK/4553 RD transmission experiment, variants at positions HA-225, HA-493, PB2-508, PB2-701, PB1-577, and NP-50 were found to be completely homogeneous in the contact animal, unlike in the inoculated animal with which it was paired (Fig. 7). In the HK/4553 DC transmission experiment, the shift in homogeneity was less striking after transmission; however, minor variants at positions PB2-508 and NP-50 were no longer detectable in virus isolated from the contact animal of the pair. Because TW/17 virus did not transmit to any ferrets by RD, we only evaluated variants isolated during the DC transmission experiment. Variants at positions PB2-444, PA-42, PA-192, PA-X-42, PA-X-191, NP-89, and NS1-75 that were isolated from the contact animal were found to be homogeneous compared to those that were isolated from the paired TW/17-inoculated animal. Collectively, these results demonstrate a reduction in minor variants of both LPAI and HPAI H7N9 viruses after the virus is passaged through a ferret via inoculation, and the minor variants are then further reduced at certain positions after transmission to a contact ferret. The relevance and potential effect of these selections are not clear. Notably, however, the LPAI HK/4553 virus variant with PB2-K627, a molecular marker for mammalian adaptation, increased in the frequency of detection in all of the tissue and nasal wash samples from infected animals that we analyzed compared to that in the inoculum. The PB2-E627 variant (commonly found in avian isolates) was present at levels below the level of detection in lung tissue and in the samples collected during the RD transmission experiment for this virus. TW/17 virus did not reveal minor variants at this position in the inoculum or ferret samples; K627 was fixed for this virus. Furthermore, despite previous reports of reduced fitness associated with viruses bearing the oseltamivir resistance mutation (NA-R294K) (35), we did not detect reversion of this substitution in any TW/17-infected ferret samples, and NA-K294 was maintained. To fully understand the selective pressures on the genetic diversity of influenza viruses that are passaged through mammalian hosts, additional study is needed. However, our findings demonstrate that the inclusion of deep sequencing analysis in the risk assessments of emerging influenza viruses offers the opportunity to identify key mammalian adaptive markers that could give rise to the next pandemic strain of influenza virus.

TABLE 4.

Genetic diversity of fifth-wave A(H7N9) viruses in ferrets

Virus	Inoculated or contacta	No. of sites with minor variation detected	No. of sites with minor or fixed variants from reference sequence that resulted in nonsynonymous changesb
HK/4553	Inoculum	12	7
	DC inoculated	10	6
	DC	11	7
	RD inoculated	9	6
	RD contact	0	2
TW/17	Inoculum	9	7
	DC inoculated	9	6
	DC contact	1	2

^aDC, direct contact transmission model; RD, respiratory droplet transmission model.

^bSequences were compared to the reference sequences deposited in GISAID.

FIG 7.

Frequency of minor variants found in inoculum and ferret specimens. Minor variants in the HK/4553 (top) or TW/17 (bottom) viral RNA samples extracted from the virus inoculum, day 3 tissue samples, and peak titer nasal wash specimens that were isolated from inoculated or contact ferrets in direct contact (DC) or respiratory droplet (RD) transmission experiments are shown. The amino acid (H3 numbering) representing the minor variant at the indicated position is shown on the left, with the length of the colored bar corresponding to the frequency of detection, while the dominant amino acid at that position is shown on the right within that column. Minor variants shown on this graph were present at levels above the 5% cutoff threshold. dpi, day postinoculation.

DISCUSSION

Despite restrictions posed by receptors, avian influenza viruses can occasionally cross species barriers and cause sporadic zoonotic infections in humans. H7N9 viruses have been ranked at the top of the list of novel influenza viruses with pandemic potential, according to the CDC Influenza Risk Assessment Tool (IRAT) (36). The fifth H7N9 epidemic wave was characterized by the largest number of human cases and, for the first time, by the emergence of HPAI H7N9 viruses. In this study, we performed a comparative risk assessment study with both LPAI and HPAI fifth-wave H7N9 viruses in an experimental setting previously used to characterize the H7N9 viruses from the first three waves. By employing both mouse and ferret models, we demonstrate that the fifth-wave LPAI H7N9 viruses share similar pathogenic features as viruses from previous waves but identify that HPAI H7N9 viruses have gained the capacity to cause enhanced disease in mammals, an extended tropism in the brain, and a reduced pH threshold for fusion. However, like H7N9 viruses from previous waves, fifth-wave H7N9 viruses have not yet developed efficient airborne transmissibility.

The emergence of HPAI H7N9 viruses with a characteristic polybasic insertion at the HA cleavage site during the fifth epidemic wave has caused considerable concerns for both poultry and human health. Compared to LPAI viruses with a typical monobasic arginine (R) residue (with cleavage occurring after the R residue, named the P1 position) in the HA cleavage site, which rely on the presence of exogenous proteases, such as trypsin-like proteases located in the mammalian respiratory tract or chicken intestinal tract, HPAI viruses with a polybasic cleavage site can be cleaved by ubiquitous intracellular subtilisin-like proteases, such as furin, conferring the ability for systemic spread (37). Previous studies have demonstrated that the presence of R or lysine (K) at the P4 position (the fourth residue upstream of the HA cleavage site) is critical for furin cleavage, with the number of basic residues and the sequence at P2 to P6 also contributing to the cleavage efficiency (38, 39). Although the presence of a polybasic cleavage site in the HA of avian influenza viruses does not necessarily result in severe disease in mammals (29), the insertion of three basic residues (-RKR-) in the HA of H7N9 viruses has considerably heightened their virulence in mice and ferrets. The enhanced mammalian virulence associated with HPAI H7N9 viruses observed in our study is disconcerting for public health. While current epidemiological data on the limited number of HPAI H7N9-infected patients have not revealed any significant changes in disease severity or mortality from that exhibited by LPAI H7N9 virus-infected individuals, a study focusing on a subset of human HPAI H7N9-infected cases noticed disease with a more rapid progression and a greater severity than that in LPAI H7N9 virus-infected patients (40, 41). Interestingly, sequence analysis of HPAI H7N9 viruses identified from poultry and environmental specimens revealed that the majority (35/58) of the viruses deposited in GISAID have a glycine (G) instead of an R at the P6 position [-GKRTA(R/G)-] upstream of the HA cleavage site (underlined). Considering the potential role of a basic residue at the P6 position in enhancing HA cleavability (38, 39), examination of whether HPAI H7N9 viruses with a different residue (either G or R) at this position would exhibit virulence differences in chickens or mammals warrants further study.

A number of molecular markers across all eight gene segments have been identified to contribute to mammalian pathogenicity and the transmissibility of avian influenza viruses (15). Like the majority of H7N9 viruses from previous waves, the fifth-wave LPAI HK/4553 virus possesses key substitutions, including S138A, T160A, G186V, and Q226L in the HA, for dual receptor binding (α2,3- and α2,6-linked sialic acids). The dual receptor binding capacity has been demonstrated by a glycan array binding assay with a closely related fifth-wave LPAI virus, HK/125, which has the identical residues at the receptor binding site as HK/4553. Additionally, HK/4553 virus possesses E627K in the PB2 for efficient replication in mammalian hosts; this virus exhibits phenotypic features comparable to those of previously characterized H7N9 viruses, emphasizing the significant role played by the aforementioned molecular markers in virus pathogenicity. However, although HK/4553 virus possesses an additional virulence marker (PB2-A588V), this substitution did not appear to increase the virulence of HK/4553 virus in mice or ferrets. Both HPAI H7N9 viruses examined possess HA-226Q (avian influenza virus-like), in agreement with recent studies showing that GD/16 virus has a binding specificity for α2,3-linked sialic acids in glycan arrays, whereas a certain level of binding for 2,6-linked sialic acid was detected by biolayer interferometry and enzyme-linked immunosorbent assays (40, 42, 43). Interestingly, TW/17 virus has acquired an additional glycosylation motif in the HA due to an I126T substitution. The HPAI A/Netherlands/219/2003 (NL/219, H7N7) virus, isolated from a fatal case from The Netherlands in 2003, has a glycosylation site at N129, which has been identified to enhance the receptor binding affinity for α2,6-linked sialic acids (21). Whether these two closely located glycans share similar functions requires further study. Despite the presence of both PB2-627K and -526R in TW/17 virus, which have been shown to act synergistically to enhance the virulence of LPAI H7N9 viruses (23), this virus exhibited virulence in animal models comparable to that of GD/16 virus bearing PB2-627E and -526K, indicating that PB2-A588V or other unidentified mutations in the polymerase genes of GD/16 virus may compensate for the lack of these virulence markers. Interestingly, the limited transmissibility by respiratory droplets shared by most LPAI H7N9 viruses from previous waves was ablated among both HPAI H7N9 viruses. The presence of 226Q in the HA for both HPAI H7N9 viruses may not fully account for their reduced transmissibility, as LPAI H7N9 viruses, such as A/Shanghai/1/2013 with HA-226Q, exhibited similar levels of transmissibility as Anhui/1 virus (HA-226L), emphasizing our incomplete knowledge of molecular correlates of H7N9 virus transmission.

Like most avian influenza viruses, H7N9 viruses from previous waves have exhibited high pH (pH 5.6 to 5.8) thresholds for fusion (26, 33). An acid-stable HA (often correlated with higher thermostability) has been linked to efficient airborne transmission, as human seasonal influenza viruses, which transmit efficiently in mammals, often have a lower fusion pH than less transmissible avian influenza viruses of the same subtype (44). Furthermore, airborne transmissible H5N1 mutant viruses have exhibited a reduced pH threshold for fusion and improved thermostability compared to wild-type (WT) H5N1 viruses that cannot transmit through the air (45, 46). Some LPAI H7N9 viruses from the second and third waves exhibited a slightly reduced pH threshold for fusion (pH 5.6 versus pH 5.8); as shown here, the fusion pH has been further reduced to pH 5.4 for two fifth-wave HPAI H7N9 viruses. Sequence analysis showed that there are only five unique mutations (I48T, A122P, K173E, and L226Q in the HA1 and E64K in the HA2, H3 numbering) shared by those two HPAI H7N9 viruses, in addition to the 4-amino-acid HA cleavage site insertion, compared to the sequences of the LPAI viruses included in our study. Although the HPAI H7N9 viruses did not show improved transmissibility, despite the lowered fusion pH, further study to identify the exact mutation(s) responsible for the reduced fusion threshold and to investigate whether such a mutation(s) confers an enhanced airborne transmissibility for LPAI H7N9 viruses would be important for surveillance and risk assessment of H7N9 viruses in the future.

The HPAI virus GD/16 was isolated from a fatal case treated with a neuraminidase inhibitor (oseltamivir), and the inoculum employed in our study contains a minor population (8.8%) with NA-294K, which was most likely derived from the original clinical specimen. A recent study by Imai et al. demonstrated that GD/16 virus bearing quasispecies of NA-294R (94%) and NA-294K (6%) exhibited a similar virulence in both mice and ferrets, as well as generally comparable airborne transmissibility results, as presented here (42). Moreover, in the same study, the authors assessed the virulence of recombinant viruses (GD/16 NA-294R and NA-294K) in both in vitro and in vivo models and found that GD/16 NA-294K displayed delayed growth in differentiated human bronchial epithelial (NHBE) cells, was attenuated in mice, and demonstrated somewhat enhanced transmissibility compared to the GD/16 WT virus group, although the difference in viral transmission was not statistically significant (42). In our study, we observed that the TW/17 isolate with the NA-294K mutation (containing no detectable minor population of NA-294R) exhibited reduced viral replication only at an early time point (24 h p.i.) during viral infection in Calu-3 cells (Fig. 1) but shared similar virulence and transmissibility in animal models as WT GD/16 virus with a predominant NA-294R population, emphasizing that H7N9 virulence and transmissibility are multifactorial traits and that the combination of different molecular markers in H7N9 viruses ultimately determines their phenotypic features in mammals.

While our study focused on the human fifth-wave H7N9 virus isolates, a recent study from Shi et al. reported that the HPAI H7N9 A/chicken/Guangdong/SD008/2017 virus isolated from healthy chickens in a live poultry market was avirulent in mice and caused moderate disease in ferrets, despite being highly pathogenic in specific-pathogen-free chickens (47). However, the mutant A/chicken/Guangdong/SD008/2017 virus bearing the PB2-627K or -701N substitution isolated from inoculated ferrets exhibited substantially increased virulence in mice and ferrets and was able to transmit efficiently (6/6 and 3/3 ferrets, respectively) by respiratory droplets between ferrets (47). Although the efficient transmission of the HPAI H7N9 viruses reported in their study may partially reflect strain-specific properties, the differences in experimental setups to assess respiratory droplet transmission, such as the direction and rate of airflow in isolator units and caging arrangements, may also contribute to the difference in transmissibility results. The same group also reported efficient respiratory droplet transmission for A/Anhui/1/13 virus in the ferret model (48), whereas inefficient respiratory droplet transmission was reported by several other groups (12, 17, 18), necessitating responsible interpretation and comparison of the transmission data generated from different research groups.

Deep sequencing has recently become a powerful tool to analyze influenza virus genetic diversity and monitor adaptive mutations in hosts (49). In our study, we did not observe the emergence of variants with any other known mammalian adaptation markers, such as mutations in the HA to enhance binding for human-like receptors, except the increased dominance of the LPAI HK/4553 variant with PB2-K627. The four mutations (HA-E493K and PB2-R508K of HK/4553 virus and PA-R192S and PA-X-F191L of TW/17 virus) that became fixed in contact ferrets have not been previously shown to be markers for mammalian adaptation. Further investigation will be needed to determine whether these mutations confer a fitness advantage or simply benefit from the stochastic bottleneck associated with transmission between ferrets (50, 51). Our current study is limited by the small number of ferret specimens used for deep sequencing analysis, and in the future, more extensive study to compare viral genetic diversity and adaptation for HPAI and LPAI H7N9 viruses may help us to better understand the potential of H7N9 viruses to become more transmissible in mammals.

The public health threat posed by H7N9 viruses will most likely continue to persist as these viruses remain enzootic in poultry in China. As demonstrated in our study, fifth-wave viruses, including both LPAI and HPAI H7N9 viruses, have not yet gained enhanced transmissibility compared to viruses from previous waves, indicating that the increased number of human infections during the fifth wave was most likely due to the expanded virus distribution and increased virus prevalence in poultry. Nevertheless, the emergence of HPAI H7N9 viruses with the capacity to cause more severe disease in mammals and with a more stable HA and distinct antigenicity has further challenged our pandemic preparedness efforts. Continued surveillance and effective measures to reduce virus transmission from poultry to humans, coupled with comprehensive risk assessments, represent essential components in future influenza virus infection and disease control strategies.

MATERIALS AND METHODS

Viruses and cells.

Patient background information, virus passage history, and virus sequence identification numbers in the GISAID database for select fifth-wave H7N9 viruses are listed in Table 1. Virus stocks were prepared in the allantoic cavity of 10-day-old embryonated hens’ eggs at 37°C for 42 h for LPAI A(H7N9) viruses (HK/4553, referred to as A/Hong Kong/VB16184091/2016 in GISAID; HK/125; HK/61, referred to as A/Hong Kong/VB16189623/2016 in GISAID; Anhui/1) and for 26 h for HPAI A(H7N9) viruses (GD/16, TW/17) as previously described (26). Stock viruses were exclusivity and sterility tested to ensure no contamination of other subtypes of influenza A viruses. The titers of the stock viruses were determined in eggs and expressed as the 50% egg infectious dose (EID₅₀) per milliliter, calculated by the method of Reed and Muench (52). Madin-Darby canine kidney (MDCK) cells, Vero cells, and cells of a human bronchial epithelial cell line (Calu-3) were cultured and passaged as described previously (53). All experiments were performed under biosafety level 3 containment with enhancements, as required by the U.S. Department of Agriculture and the National Select Agent Program (54). The titer of the virus stock of Anhui/1 described previously (12) was determined in both MDCK cells and eggs, which was 1.9 × 10^9.0 PFU/ml and 3.2 × 10¹⁰ EID₅₀/ml, respectively.

H7N9 virus growth kinetics in Calu-3 cells.

Polarized Calu-3 cells grown on 12-well plates with semipermeable inserts (Corning) were inoculated in triplicate with fifth-wave LPAI or HPAI H7N9 viruses at an MOI of 0.01 according to previously described methods (53). Cell supernatants collected at 2, 24, 48, and 72 h p.i. from the apical side of infected cells cultured at 37°C were stored at −80°C until titration in 10-day-old embryonated hens’ eggs.

H7N9 virus infectivity and replication in mice.

All animal experiments were performed under the guidance of the Centers for Disease Control and Prevention’s Institutional Animal Care and Use Committee in an Association for Assessment and Accreditation of Laboratory Animal Care International-accredited facility. The virulence of fifth-wave H7N9 viruses in mice was evaluated based on morbidity (measured by weight loss), mortality, and the 50% mouse lethal dose (MLD₅₀). To reduce the variations caused by strain-specific differences, two representative isolates of both HPAI and LPAI H7N9 viruses were included in the mouse experiment. Briefly, groups of five female BALB/c mice (The Jackson Laboratory, ME), 6 to 8 weeks old, were anesthetized by isoflurane inhalation and were each inoculated i.n. with 50 µl of serial 10-fold dilutions of H7N9 virus ranging from 10^2.0 to 10^6.0 EID₅₀ diluted in phosphate-buffered saline (PBS). Inoculated mice were monitored daily; any mouse with the loss of 25% or more of its preinoculation weight was euthanized. To evaluate virus replication in pulmonary tissues (lung, nose) and extrapulmonary tissues (brain), groups of six mice each were i.n. inoculated with 10^3.0 or 10^6.0 EID₅₀ of H7N9 virus. Three mice from each group were euthanized on days 3 and 6 p.i., unless otherwise indicated, for collection of tissues, which were then homogenized in 1 ml of PBS and titrated in 10-day-old embryonated hens’ eggs.

H7N9 virus pathogenicity and transmission in ferrets.

The virulence and transmissibility of fifth-wave H7N9 viruses in both direct contact (DC) and respiratory droplet (RD) transmission models were assessed as described previously (26). Ferrets were housed in Duo-Flow bioisolator units (BioClean Lab Product Inc.) with a vertical airflow rate of 20 cubic feet per minute (CFM). In brief, each of nine ferrets (Triple F Farms, Sayre, PA, or Marshall Farms, North Rose, NY), which were 8 to 10 months old and serologically negative for currently circulating influenza viruses, were i.n. inoculated with 10^6.0 EID₅₀ of a fifth-wave H7N9 virus prepared in 1 ml of PBS. Six of these nine inoculated ferrets served as donors in transmission studies and were monitored daily for clinical signs of disease, including coughing, sneezing, fever, weight loss, and lethargy (scored and used to calculate RII as described previously [30]). Nasal wash samples were collected from these animals on alternating days from days 1 to 11 p.i., and rectal swab samples were collected on days 1, 3, and 5 p.i. The remaining three inoculated ferrets were humanely euthanized on day 3 p.i. to collect tissues for virus titration. The viral titers in the ferret samples were determined in 10-day-old embryonated hens’ eggs with a limit of detection of 1.5 log₁₀ EID₅₀/ml. The transmission studies were set up as detailed previously to include 3 pairs each of cohoused ferrets for the DC transmission model and 3 pairs each for the RD transmission model (26). Virus transmissibility was decided by the presence of infectious viruses in nasal washes from contact ferrets and seroconversion against homologous virus on day 21 p.c.

Histopathological evaluation.

The central nervous system tissues from euthanized animals were fixed in 10% neutral buffered formalin and embedded in paraffin. Four-micrometer sections from formalin-fixed, paraffin-embedded specimens were stained with hematoxylin and eosin (H&E) for histopathological evaluation.

pH threshold for fusion of fifth-wave H7N9 viruses.

The fusion pH of H7N9 viruses was evaluated by a syncytium formation assay in Vero cells as described previously (55). In brief, Vero cell cultures inoculated with H7N9 viruses at an MOI of 1 were incubated with fusion buffer (20 mM HEPES, 2 mM CaCl₂, 150 mM NaCl, 20 mM citric acid monohydrate/sodium citrate tribasic dehydrate) with pH values ranging from 5.2 to 6.0 (at 0.1-unit increments) for 5 min following tosylsulfonyl phenylalanyl chloromethyl ketone (TPCK) trypsin treatment for 15 min. Cells were fixed at 3 h after fusion induction and stained for anti-NP antibody (clone A1, A3 blend; EMD Millipore) for immunofluorescence microscopy. The fusion pH threshold is defined as the highest pH value at which 50% or more syncytia can be achieved.

Next-generation sequencing and genetic analysis.

Viral RNA from ferret nasal washes was extracted using a QIAamp viral RNA minikit (Qiagen), and the full influenza virus genome was amplified and deep sequenced on an Illumina MiSeq instrument as described previously (56). Genome assembly and minor variant analyses were performed using IRMA (57). Minor variants were considered above a 5% cutoff threshold and visualized across sample conditions using Tableau software.

Accession number(s).

The sequences of the H7N9 virus subpopulations detected by deep sequencing analysis of the virus inocula and a limited set of tissues and nasal wash specimens collected from selected paired sets of inoculated and contact ferrets infected by LPAI HK/4553 or HPAI TW/17 virus were deposited in the NCBI SRA database under accession number PRJNA494222.