While the ferret has proved essential for the study of influenza virus pathogenesis and transmission, the use of this species to examine alternate inoculation methods has been limited 
. Characterizing the progression of disease following alternate routes of inoculation with influenza virus will assist in the better understanding of unique features and the relative severity and risk associated with different exposure routes. In this study, we established an in vivo
model using the ferret to assess the ability of influenza viruses of multiple subtypes to use the eye as a gateway to establish a productive infection. Both human and avian influenza viruses were capable of mounting a respiratory virus infection in ferrets following i.o. inoculation. The detection of virus in ocular samples collected from ferrets inoculated by either ocular or intranasal routes demonstrates the importance of studying ocular involvement in respiratory virus infection. Divergent patterns of virus transmissibility by respiratory droplets following use of different inoculum volumes and routes of inoculation highlights the complexity of properties which govern virus transmission.
The high similarity of respiratory tract tissue between humans and ferrets makes the ferret model an attractive one for modeling human respiratory disease and investigating the role of receptor specificity in influenza virus pathogenesis, providing an advantage over murine models 
. We found that the sialic acid composition of ferret corneal epithelial sheets more closely mimics that of humans compared with a mouse model, demonstrating another physiological parallel between ferrets and humans 
. Bridging the α2–3 rich corneal and conjunctival epithelial surfaces with α2–6 rich upper respiratory tract tissue is the lacrimal duct, which expresses both α2–3 and α2–6 linked sialic acids 
. Characterization of the distribution of sialic acids in the ferret conjunctiva and lacrimal duct, in addition to the composition of ferret ocular mucins, will allow for a better understanding of virus attachment and replication in these locations. However, in a previous study, we demonstrated that the ability of influenza viruses to bind to or replicate in ocular tissue cannot be explained by sialic acid binding specificity alone 
. Our detection in ferret ocular tissue of both human and avian influenza viruses with distinct binding specificities further underscores this point (, ).
Macroscopic signs of ocular disease in ferrets were not observed during the course of infection with any virus tested, similar to prior observations in mouse and rabbit models following deposition of influenza virus on the corneal surface 
. A previous study in ferrets reported mild conjunctival inflammation following i.o. inoculation with an H7N3 virus, however this may be attributable to strain-specific differences or the use of younger (3–5 month old) ferrets 
. Despite the absence of visible ocular complications, virus was consistently detected in CW samples from ferrets inoculated by the intranasal or ocular route (, ). Levels of viral M1 RNA generally correlated with the magnitude of virus isolation, and were a more sensitive detection method compared with virus isolation in CW samples, similar to that observed in human eye swabs (, ) 
. The presence of virus in RS samples following both i.n. and i.o. inoculation with influenza virus has been previously reported and likely originates from virus swallowed during inoculation 
, as indicated by deposition of virus in the esophagus following initial virus inoculation by both inoculation routes (, ).
Unlike in a murine model, the ferret model supported virus replication of both human and avian influenza viruses following i.o. inoculation 
. In this ferret model, H7 viruses were detected at higher titer in NW samples and with a higher frequency in ocular CW samples compared with H5N1 viruses, suggesting a recapitulation of the tropism of the H7 virus subtype observed in humans (). However, the permissiveness of multiple virus subtypes to cause a productive infection following i.o. inoculation in the ferret points to a greater capacity of influenza viruses to use the eye as a portal of entry in an experimental in vivo
setting, just as previous in vitro
studies have demonstrated that numerous human ocular cell types distributed throughout the ocular area support infection and replication with both avian and human influenza viruses 
. Cumulatively, these previous in vitro
studies suggest that the apparent ocular tropism associated with viruses of the H7 subtype is not due to a superior ability to replicate in ocular cells compared with other virus subtypes. Future studies evaluating potential fine receptor specificity differences on the ocular surface and the composition of ocular mucins which may restrict exposure to the ocular epithelial surface to selected virus subtypes may provide a greater understanding of this property.
Consistently high titers of human influenza viruses in ocular samples following both i.n. and i.o. inoculation indicates that these H1N1 and H3N2 viruses are not exhibiting a preferential tropism for ocular infection but more likely are a reflection of the high titers observed in the upper respiratory tract in these tissues independent of the initial inoculation route. Specifically, the nasolacrimal duct which links the ocular lacrimal sac to the nasal meatus could serve as a conduit for virus-containing fluid exchange between ocular and respiratory tract tissue 
. Numerous reports have documented drainage of vaccine or immunizing agents to nasal tissue following topical ocular administration as well as the spread of intranasally administered solutions to the conjunctival mucosal surface 
. However, spread of virus from respiratory tract tissue to ocular tissue following i.n. inoculation with human or avian influenza viruses has not been observed previously in the ferret and only sporadically reported in the mouse, possibly due to relatively low titers of virus in nasal tissue or other anatomical differences 
. Detection of both human and avian influenza viruses in ferret eyes and conjunctival tissue following i.n. inoculation indicates that virus can circulate proximal to the nasal cavity and nasolacrimal canal more readily than previously considered and that more routine collection of ocular tissue during standard virus pathotyping in mammalian models is warranted to better understand the extent of viral ocular dissemination (). While it is unlikely that ferret grooming practices are solely responsible for virus spread between these locations, human studies with numerous respiratory viruses have demonstrated the ability to self-inoculate between ocular and respiratory sites, and it is possible that self-inoculation could be further contributing to virus spread in this model 
. Inoculation of ferrets by the ocular route with 106
of selected human and avian influenza viruses in a 20 µl volume resulted in comparable results to those obtained employing a 100 µl inoculation volume, indicating that replication-independent drainage of virus to respiratory tract tissue and subsequent virus detection in NW and CW samples reported in this study was not contingent on the inoculum volume (data not shown).
Visualization of virus deposition using fluorescently-tagged virus has allowed for a new understanding of dissemination patterns following in vivo
inoculation. Using this technique, we confirmed previously hypothesized reports of replication-independent drainage from ocular to respiratory tract tissue 
. Viral load measured in respiratory tract tissues day 3 p.i. following i.n. or i.o. inoculation largely mirrored initial virus deposition patterns, with tissues possessing the greatest quantity of virus reflecting those sites of greatest initial virus deposition during virus inoculation (–, ). Both i.o. and i.n. inoculation using a 100 µl volume resulted in detection of virus at high titers in upper and not lower respiratory tract tissues day 3 p.i., indicating that inoculation with a reduced volume leads to limited initial virus deposition in respiratory tract tissues regardless of inoculation route. The delay in high virus titer recovery from lower respiratory tract tissues during HPAI virus infection after i.o. inoculation and the delay in onset of severe disease and lethal outcome with Thai/16 virus is likely due to replication-dependent spread (and not deposition) of virus to the lower respiratory tract, suggesting that during inoculation, the majority of virus was retained in conjunctival tissues, drained to the nasal turbinates, or swallowed and diverted away from the lower respiratory tract; future study of ferret lacrimal tissue in this role is warranted. Comparable delays in onset of severe disease compared with traditional i.n. inoculation were observed in a murine model of i.o. inoculation and a ferret model of aerosol inoculation, despite ultimately similar lethal outcomes 
. The finding here of H5N1 subtype viruses using the eye as a portal of entry to initiate a lethal infection, shown previously in a mouse model, underscores the risk of ocular exposure to influenza viruses, even those subtypes not typically considered to have a tropism for this tissue 
Despite efficient replication of seasonal H1N1 and H3N2 viruses in the upper respiratory tract of ferrets following i.o. inoculation, these viruses did not result in frequent detection of sneezing and nasal discharge, and did not transmit efficiently to naïve contacts by respiratory droplets (, ). Infrequent sneezing is commonly observed among influenza viruses which do not transmit efficiently by respiratory droplets and could be contributing to the reduced transmissibility seen here 
. Further research is needed to better understand the virologic and immunologic properties which confer the incidence of sneezing and nasal discharge and the role of these properties in virus transmissibility 
. Additionally, the efficiency of virus transmission by respiratory droplets following i.o. inoculation was likely influenced by the reduced initial virus deposition and subsequent limited replication in the ferret trachea, as reduced virus transmissibility by respiratory droplets was observed following i.n. inoculation when using a 100 µl but not 1 mL volume (, , and data not shown). Despite similarly high virus titers and duration of virus shedding in NW samples, the presence of expelled virus particles originating from tracheal replication which was present during traditional i.n., but not i.o. or i.n. inoculation using a reduced volume, may have contributed to differing transmission efficiencies between inoculation routes and may represent a previously unrecognized role for virus replication in tracheal tissue in virus transmissibility by respiratory droplets. In contrast, virus transmission in the presence of direct contact did not differ between inoculation routes. The reduced transmissibility of virus following i.o. inoculation is in agreement with epidemiological studies which demonstrate that the majority of human cases of conjunctivitis following H7 influenza virus exposure are self-limiting 
. Further study evaluating the shedding of virus into the environment among persons infected with influenza viruses which cause respiratory or ocular disease will shed light on potential differences in transmission dynamics independent of virus subtype.
The diversity of potential exposures to influenza virus underscores the importance of studying the development of respiratory disease resulting from alternate exposure routes. This knowledge is critical for both a greater understanding of the establishment of influenza virus respiratory disease as well as differences in virus transmission dynamics following differing exposure routes. The facile dissemination of virus inoculum from ocular to nasal tissue, and the detection of virus in both NW and CW samples throughout the acute phase of ferret infection, highlights the ability for concurrent ocular and respiratory disease following influenza virus infection; not surprisingly, reports of conjunctivitis and influenza-like illness in the same individual have been documented during H7 outbreaks resulting in human infection 
. While much regarding the properties which regulate the ocular tropism of influenza viruses remains to be determined, our results highlight the potential for a range of influenza A subtypes to initiate infection through the eye and support the use of eye protection during occupational exposure to aerosols containing influenza viruses