The association between SAα2,3Gal and avian influenza viral load in mallards (Anas Platyrhynchos) and blue-winged teals (Anas discors)

Background Individual heterogeneity in pathogen load can impact disease transmission dynamics; therefore, identifying intrinsic factors responsible for variation in pathogen load is necessary for determining which individuals are prone to be most infectious. Since low pathogenic avian influenza viruses (LPAIV) preferentially bind to alpha-2,3 sialic acid receptors (SAα2,3Gal) in the intestines and bursa of Fabricius in wild ducks ( Anas spp.), we investigated juvenile mallards ( Anas platyrhyncos ) and blue-winged teals ( Anas discors ) orally inoculated with A/northern pintail/California/44221-761/2006 (H5N9) and the virus titer relationship to percentage of SAα2,3Gal in the intestines and bursa. To evaluate the natural variation of free-ranging duck populations, birds were hatched and raised in captivity from eggs collected from nests of free-ranging birds in North Dakota, USA. Data generated from real time RT-PCR used to quantify virus titers in cloacal swabs, ileum tissue, and bursa of Fabricius tissue and lectin histochemistry used to quantify percentage of SAα2,3Gal was analyzed through a series of multiple linear regression and mixed models. Results In mallards, we found high individual variation in virus titers significantly related to high variation of SAα2,3Gal in the ileum, with sex being a significant factor. In contrast to mallards, individual variation in teals was minimal and significant relationships between virus titers and SAα2,3Gal were not determined. Collectively, teals had both higher virus titers and a higher percentage of SAα2,3Gal compared to mallards, which may indicate a positive association between viral load and SAα2,3Gal. Statistically significant differences were observed between infected and control birds indicating that LPAIV infection may influence the percentage of SAα2,3Gal, or vice versa, but only in specific tissues. Conclusions The results of this study provide quantitative evidence that SAα2,3Gal is related to LPAIV titers; thus, SAα2,3Gal should be considered a potential intrinsic factor of variation in LPAIV load.

SAα2,3Gal expression in the duodenum and jejunum compared to other Anseriformes (18). Variation was also found within species, such as mallards, based on the type of lectin used, Maackia amurensis I (MAL I) vs. Maackia amurensis II (MAL II) (18). While the previous literature suggests there is variation in SAα2,3Gal intensity and distribution within and across species, the percentage of SAα2,3Gal in the intestines and bursa has yet to be statistically quantified and related to LPAIV load, a first step in understanding this potential source of AIV variability across individuals and species.
In this study, we address this knowledge gap by investigating the relationship between SAα2,3Gal and LPAIV load in mallards and blue-winged teals (Anas discors, hereafter referred to as "teal"). Both species are important hosts for LPAIV. The mallard is important because of their worldwide distribution, their periodomesticity, and the large diversity of AIV strains isolated from them, including highly pathogenic strains causing high mortality in poultry and people (3,22,23). Teals have high infection prevalence (24) and an important role in over-wintering the virus in the southern United States (25,26).
We hypothesized that higher percentages of SAα2,3Gal in mallard and teals corresponds with higher LPAIV titers. Additionally, we hypothesized that the relationship between virus titers in cloacal swab, ileum tissue, and bursa tissue would all be positively related to each other. Sex-based differences, species-based differences, and comparisons in the percentages of SAα2,3Gal between control and infected birds was also analyzed, where we did not expect to see differences. This research provides a first look into this putative intrinsic factor responsible for LPAIV individual variation in mallards and blue-winged teals.

Viral Infection of Mallards and Teals
All birds inoculated with LPAIV H5N9 (mallard = 60, teal = 44) were infected as demonstrated by positive titers on at least one day between one to five days post infection (DPI) in cloacal swabs, ileum tissue, and/or bursa tissue (Additional File 1). No birds shed virus past 15 DPI, and of the birds that survived to 15 DPI, 99.9 (mallard) and 98.5 (teal) percent of the total virus shed by those birds occurred in the first five DPI. As expected with LPAIV, we observed no clinical signs of disease such as ruffled feathers, lethargy, respiratory distress, or any pathology.

Species and Sex-based differences in Viral Shedding
Statistically significant differences in viral shedding were found between mallards and teals, but not between males and females within species. Mallards had statistically higher variation than teals in cloacal swab viral titers on one, two, three, and five DPI (Fligner-Killeen p < 0.05; Table 1  . Post-hoc analysis shows the lectin score in the cecum brush border (p < 0.001) and cecum villi (p < 0.001) was higher in infected birds than control birds (Fig. 3). Higher intertissue and inter-individual variation was observed in mallards compared to teals for all tissue/cell types (Fligner-Killeen p < 0.001; Table 2; Fig. 4). LPAIV-infected teals had statistically higher lectin staining than LPAIV-infected mallards (F 1,102 = 309.92, p < 0.001) with a statistically significant interaction between species and tissue/cell type (F 11,1067 = 9.95, p < 0.001). In mallards, the ileum, cecum, and colon had statistically similar lectin scores for most cell types; however, lectin scores for most cell types in the proximal intestine were significantly lower (p < 0.05) than the lectin scores in ileum, cecum, and colon (Fig. 5). In teals, most tissues/cell types had similar lectin scores, except for the cecum brush border and cecum villi, which were statistically significantly lower than all other tissue/cell types (Fig. 5).
In both species, lectin staining was not significantly different between males and females (mallard:   Table 3) was selected as the best fitting model. Our results indicated that the lectin score of the ileum villi was positively associated with a higher virus titer. LHS in the ileum brush border was negatively associated with a higher virus titer.
Sex was not a significant factor in this model.

Discussion
Mallards and blue-winged teals are important reservoir hosts for avian influenza viruses (3,24,25); they are both widely distributed waterfowl species and commonly infected with both LPAIV and HPAIV.
Our study documents both within and between-species variation in viral shedding as well as for SAα2,3Gal, the viral receptor for many LPAIVs. In mallards, but not teals, we found viral shedding was related to percentage of SAα2,3Gal. While we expected to see positive linear relationships between virus titers and SAα2,3Gal in all tissues and cell types, the mallard ileum was the most predictive of virus titers, with a positive relationship between virus titers and SAα2,3Gal in ileum villi enterocytes, and a negative relationship between virus titers and SAα2,3Gal in the ileum brush border. Despite the lack of relationship between viral shedding and SAα2,3Gal in teals, we observed significantly higher viral shedding by teals, and a higher percentage of SAα2,3Gal compared to mallards.
As the direction (positive or negative) of the correlation between SAα2,3Gal and virus titer varied across mallard tissue locations, our data highlight the importance of understanding tissue-specific tropism as it relates to cell surface SAα2,3Gal distribution. Within mallards, the positive relationship between virus titer and SAα2,3Gal in the ileum villi enterocytes was expected given that LPAIV replicates in intestinal enterocytes by binding SAα2,3Gal on the surface of the cell for cell entry (27).
A reason why ileum villi enterocytes were most correlated with viral titer compared to ileum crypt enterocytes may be that the villi have closer direct contact with digesta and as a result, closer direct contact with virus passing through the gut. For example, previous studies have found LPAIV antigen via immunohistochemistry more consistently in mallard villi enterocytes compared to the crypts (12,13). Surprisingly, however, the other three intestinal tissue types: proximal, cecum, and colon, were not associated with virus titers in the MLR models.
The lack of a statistically significant relationship between SAα2,3Gal and virus titer in the mallard colon was unexpected, given numerous studies have indicated the colon as a site for high LPAIV replication (11)(12)(13)28). The lack of a statistically significant relationship between colon SAα2,3Gal and viral shedding may be due to the statistical approach we used. The MLR method was designed to identify the tissue or tissues which most contributed to the variation observed in virus titers while eliminating issues from the model that exhibited multicollinearity. Since SAα2,3Gal in the ileum and colon were 63% correlated with each other (Additional File 11), the colon could also have a contributing effect to viral load, but not as strongly as the ileum. positive staining goblet cells since they produce mucins which also express SAα2,3Gal and may inhibit cell entry and viral replication (29,30). Although the consensus is not clear concerning the presence of SAα2,3Gal in the mallard proximal intestine; previous findings indicate that positive viral antigen via immunohistochemistry is more commonly found in the ileum, cecum, and colon when cloacal swab virus titers are high (12,13), which would indicate that the proximal intestine is not a main site of LPAIV replication. Therefore, we suggest that the lower percentage of SAα2,3Gal in the proximal enterocytes could explain the lack of a relationship to shedding of LPAIV.
Two hypotheses could explain the negative relationship between SAα2,3Gal in the ileum brush border and virus titer. Initially, we expected to see a positive relationship between SAα2,3Gal in the brush border of all intestinal tissues and virus titers since the receptors are located on the surface of the cell and more likely to be exposed to virus (31). However, as a virion attaches to a receptor, the virion along with the receptor becomes engulfed by the cell for replication, therefore removing the receptor from the surface of the cell (32). This idea is also consistent with the differences observed in percentage of SAα2,3Gal between infected and control mallards, where control mallards had higher SAα2,3Gal in the ileum and colon brush border compared to infected birds. Secondly, mucus is also found along the brush border and LPAIV has been found to bind SAα2,3Gal in mucus, which would prohibit the virus from reaching the enterocyte for virus replication (15,30,33); thereby reducing the quantity of virus shed. Up-regulation of mucins have also been observed in response to other viruses which bind sialic acid receptors (29), such as human rotavirus infections (34). To investigate the true explanation for the negative relationship between percentage of SAα2,3Gal in the ileum brush border and virus titers, further experimental research is warranted.
The bursa epithelial cells are also considered to be an important site of replication for LPAIV in waterfowl, including mallards (12,13). However, given autolysis of tissue samples we were unable to analyze the relationship between SAα2,3Gal in the bursa and viral shedding in mallards. In teals, lectin staining was very high in the bursa; however, it was not significantly related to viral shedding.
Lack of a significant relationship to viral titer in teals could be attributed to the lack of individual variation in SAα2,3Gal expression in the bursa or to a sporadic correlation between bursa and cloacal swab virus quantity.
The premise of our study was to determine if the percentage of SAα2,3Gal in the intestines and bursa may be associated with cloacal shedding; hence, we predicted the variation of SAα2,3Gal in control and infected birds would not differ. Our data suggest this is not the case. In the cecum, the percentage of SAα2,3Gal was higher in the crypts of infected mallards compared to their conspecific controls. Similarly, in teals the percentage of SAα2,3Gal was higher in the cecum villi and brush border of infected birds. The ceca have a unique role in the functioning of the vertebrate immune system. The cecal tonsils, a major lymphoid tissue in the cecum, enlarges during gut infections due to infiltration of immune cells, which includes macrophages (35). Macrophages express Gal-specific receptors (36), which could explain the higher abundance of SAα2,3Gal in the cecum of infected birds relative to controls. White leghorn chickens have a greater abundance of sialic acid receptors than silky fowl and this corresponded with a higher number of immune cells in the cecum in the leghorns (37). The cecum has a unique response to LPAIV infection compared to other intestinal tissues, which warrants further analysis of SAα2,3Gal in this tissue.
Contrary to differences in SAα2,3Gal expression between LPAIV-infected and control birds in the cecum, control mallards expressed more SAα2,3Gal in the ileum and colon brush border than infected mallards. Franca et. al (13) found that SAα2,3Gal was lower in the cecum, colon, and bursa of infected birds compared to control birds. Their hypothesis indicated that the SAα2,3Gal expression level may decrease after infection because the neuraminidase function of the virus allows cleaving of the receptor releasing virions from the cell (38). When the receptor is cleaved, it is no longer present on the cell surface which would reduce lectin binding. While Franca et. al (13) did not specify whether the decrease in lectin staining was on the surface of the enterocyte, we found mallards to have a higher percentage of SAα2,3Gal only in the brush border. Our results indicate the importance of assessing the specific location of SAα2,3Gal in determining their function in influenza studies.
An interesting finding was the species-specific differences between mallards and teals in variation and viral shedding. The variation observed in mallards in our study is consistent with what has been observed in other experimental infection studies (10,39 (45)). Pepin et al. (45) observed significantly higher cloacal shedding of LPAIV in males than females but did not draw a conclusion about the mechanism underlying this difference.
There is more evidence for sex-based differences in infectiousness for other host-pathogen systems (46); however, much of the difference is related to a bird's reproductive state and/or hormone levels (47,48). The birds in our study were only 6-12 weeks old and not reproductively mature; hence, any sex-related differences are unlikely due to variation in hormone production.
The identified positive relationships between viral RNA in cloacal swabs, ileum tissue, and bursa tissue further supports the importance of the ileum and bursa for cloacal shedding of LPAIV. Prior to this study, it was well known that LPAIV replicates in duck intestines and the bursa of Fabricius (11)(12)(13). While testing for virus in cloacal swabs is the standard method for determining AIV fecal shedding (49,50), the direct relationship between tissue replication and virus shed by the cloaca was unknown. Through quantifying viral RNA via RT-qPCR in ileum and bursa tissue, significant positive relationships were found between virus titers in cloacal swabs, ileum tissue, and bursa tissue, indicating the contribution of these tissues to the cloacal virus shed. The positive relationship between virus titers in the ileum and cloacal swabs provides additional evidence to support our conclusion that ileum SAα2,3Gal was associated with virus titer. Lastly, these positive relationships add validity to collecting cloacal swabs as an indicator of virus titer in the ileum and bursa and perhaps the infection status of the bird as a whole.

Conclusion
Understanding the mechanism underlying variation in infection severity and viral shedding can provide insight into why a few individuals in a population are more infected than others, and perhaps, why some are more infectious than others. LPAIV is a gut-associated pathogen in wild waterfowl; hence, the physiology of the host's gut is an important determinant of within-host-pathogen interaction. The results of this study provide evidence that sialic acid receptors in the gut are associated with viral load. Since sialic acid expression varies both between species (18, 21) and within species (13)

Study Species and Locations
Mallards and teals used for this study were collected as eggs from the nests of wild birds in the
All T-group birds (hereafter referred to collectively as "infected") were inoculated with 1.0 mL of 5.63 log EID50/ml viral inoculum on 0 DPI, diluted in DMEM by placing one drop on each eye and each nare, then dispensing the rest in the esophagus (56,57). All C-group birds (hereafter referred to collectively as "control") were sham-inoculated with 1.0 mL of sterile DMEM in a similar fashion.
During the inoculation and after inoculation, birds were kept in biosafety level two conditions and personal protective equipment consisted of non-vented, full coverage eye goggles, hair cap, N95 respirator, double gloves, tyvek suit, and plastic booties.
For each model, Cloaca Titer was the dependent variable. To test species differences, species and DPI, plus their interaction, were set as fixed effects and individual birds were a random effect. To test sex-based differences, mallards and teals were analyzed in two different models with sex and DPI, plus their interaction, set as fixed effects and individual birds were a random effect. Differences in variance were detected using the Fligner-Killeen test (69). When differences in variance were detected, the "weights" argument was used as an adjustment for the model. ANOVA tables were visualized using the "anova.lme" function. We used the post-hoc Tukey's test to assess inter-variable differences.
Lectin Histochemistry scores were analyzed using a similar approach to virus titer, but with lectin score as the dependent variable. Using the linear mixed model methods described above, we tested for lectin score differences between infected and control birds, setting infection status and tissue/celltype plus their interaction as fixed effects and individual birds as a random effect. Using data from infected birds only, we also assessed species and sex-based differences in lectin score. Species differences were assessed by setting species and tissue/cell type and their interaction as fixed effects and individual birds as a random effect. Sex-based differences were analyzed in separate models for each species with sex and tissue/cell type, plus their interaction, as fixed effects and individual birds as a random effect.
We also looked at lectin score correlations between cell types within intestinal tissue type using Pearson's r coefficient. We considered cell types within a tissue type (proximal, ileum, cecum, colon) with a coefficient of 0.8 or higher to indicate a strong correlation. If all three cell types within a tissue were highly correlated, we used the "pccomp" function, a principle component analysis Also, some additional tissue samples were too autolyzed to assess, which reduced the sample size for mallard MLR models from 40 to 25 birds (T1 = 6; T2 = 8; T5 = 11), and teal MLR models from 36 to 32 birds (T1 = 9, T3 = 11, T = 12).
To determine the best fitting MLR model for each dependent variable, we followed a consistent procedure. Using the "lm" function, global models were tested for dependent variables Last Cloaca Titer, Ileum Titer, and Bursa Titer separately. To select parsimonious model fits to the data, we used the "step" function for stepwise variable selection based on the generalized Akaike's Information Criterion (AIC) (70). We then used variance inflation factor (VIF) scores, the "vif" function in the "car" package (71), to identify problematic co-linear predictors from the stepwise-chosen models.
Independent variables with VIFs > 3.0 were determined problematic and were removed from the model one at a time until all VIFs < 3.0 (72       Teals had higher lectin scores than mallards, but with less variation. Mean lectin scores + 95% confidence intervals for intestinal tissues proximal (duodenum and jejunum), ileum, cecum, and colon for LPAIV H5N9 infected mallards and blue-winged teals. Across all panels, points with different letters are considered significantly different (p < 0.05).

Supplementary Files
This is a list of supplementary files associated with this preprint. Click to download.