Permits and protocols
Protocols for animal care and experimental sampling procedures were approved by Michigan State University (MSU) Institutional Animal Care and Use Committee (AUF 12/16–211-00). All euthanasia procedures were in accordance with the Animal Welfare Act and Guidelines to the Use of Wild Birds in Research . Duck eggs were collected with permission from the U.S. Fish and Wildlife Permit (M Bl 94,270–2) and North Dakota Game and Fish Department License #GNF03639403.
Study species and locations
Mallards and teals used for this study were collected as eggs from the nests of wild birds in the southwest corner of Towner County, North Dakota, USA (48.4431853, − 99.3156225). In May–June 2015 we collected 90 mallard eggs (1–2 per nest) from a total of 50 nests, with each nest containing an average of eight eggs per clutch. The following summer, May – June 2016 we collected 80 blue-winged teal eggs (1–2 per nest) from a total of 40 nests. Nests were found and eggs collected by dragging a heavy metal-link chain behind two ATVs driving in parallel which initiated hens to fly off their nests . Eggs were candled in field to determine age, and any eggs that either had not started incubation or were between 15 and 22 days of incubation were shipped overnight to MSU in East Lansing, Michigan. Each year we made 2–4 shipments of 15 to 40 eggs each over a period of 6 weeks. Unless specified otherwise, all procedures were the same for each species/year.
Upon arrival at MSU, eggs were immediately placed into a climate-controlled egg incubator (Sportsman 1502 Egg Incubator, GQF Manufacturing Co., Savannah, GA) housed within a biosafety level two room within the MSU Research Confinement Facility. Eggs were incubated at 37.5 °C with 45–50% humidity and rotated electronically 10 times per day. Eggs were candled for viability and age once every three days. As soon as eggs pipped, they were moved into a hatching incubator (Sportsman 1502 Egg Incubator, GQF Manufacturing Co., Savannah, GA) at 37.2 °C with 70–80% humidity. Chicks remained in the hatcher until they were dry, approximately 12–24 h post hatching. Each bird was then weighed to the nearest 0.1 g, banded with a uniquely numbered plastic leg band, and placed in a brooder (30–35 °C). Birds were kept in brooders for two weeks, then moved to open-room housing where a maximum of 35 birds were housed per room (400sq feet). Each room maintained a temperature of 23 °C and 45–55% humidity, had two swimming pools (45″ diameter, 10″ depth), and two dry pools with aspen chip bedding. In both years, birds were maintained on a 13:11 h light:dark photoperiod.
Birds were fed ad libitum Purina® Flock Raiser® Crumbles (Purina, St. Louis, MO, USA) and supplemented with chopped dandelion greens twice per day. Rooms were fully cleaned twice per day. Birds were routinely checked for normal health and weighed every five days. One week prior to inoculation, mallards were separated into individual cages of 20 cages per room. Blue-winged teals were kept in the open room housing separated by experimental group.
LPAIV A/northern pintail/California/44221–761/2006 (H5N9), originally collected from a northern pintail cloacal swab and isolated in specific pathogen free embryonated chicken eggs (ECE), was acquired from the USGS National Wildlife Health Center in Madison, WI (USDA Veterinary Permit 44,372). We prepared stock virus propagating the virus in 9 to 11-day old ECE (Charles River, Norwich, CT, USA) . The infectious titer of the stock virus of 7.63 log EID50/ml was determined using the 50% egg infectious dose (EID50) and calculated using the Reed & Muench method . The viral inoculum was prepared by diluting the stock virus in Dulbecco’s Modified Eagle Medium (DMEM) (Gibco® by Life Technologies, Grand Island, NY, USA) to yield a final titer of 5.63 log EID50/ml.
Individual birds were assigned to one of two control groups for each species, one of five mallard LPAIV treatment groups, or one of four teal LPAIV treatment groups (Fig. 1). Experimental group assignment was done using pseudo-stratified randomization with birds being stratified by body mass, age, and sex. Additionally, individuals from the same nests were assigned to separate groups. Group names refer to their species (mallard = M, teal = B), whether they received LPAIV treatment (inoculated with virus = T, control = C), and the DPI they were sacrificed (# to follow T/C). The minimum sample size per group was based on individual viral load variation observed in populations as small as 10 individuals . Additional birds were placed in groups on DPI of most importance such as high viral shedding (DPI 1–3) and early detection of antibody titer (DPI 5) .
All LPAIV treatment group birds (also referred to 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 [48, 49]. All control birds 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.
We collected cloacal swabs on all live individuals. Cotton tipped swabs were collected from mallards on 1–5, 8, 11, 13, 15, 17, 19, 22, 24, 26, and 29 DPI, and from teals on 1–7, 9, 11, and 14 DPI (Fig. 1). Swabs were stored in 3.0 mL of brain-heart infusion broth (BHI), transported on ice, and stored in − 80 °C until sample processing.
Mallards, as described by their assigned groups, were sacrificed on 1, 2, 5, 15, and 29 DPI, and teals were sacrificed on 1, 3, 5, and 14 DPI (Fig. 1). Mallards sacrificed on one DPI were euthanized by intravenous lethal injection of pentobarbital sodium and phenytoin sodium solution (Beuthanasia-D Special, Merck Animal Health, Madison, NJ, USA). All other birds were euthanized by carbon dioxide inhalation. Bird carcasses were preserved on ice until necropsy was performed.
Necropsy and tissue collection
Mallard necropsy was performed in the same room where birds were kept under biosafety level two conditions mentioned above. Teal necropsies were performed under a biosafety cabinet. Necropsies were performed on mallards within one to six hours of being euthanized, with an average time of approximately four hours post euthanasia. Due to autolysis of tissue samples observed with mallards, we performed necropsies on teals within one hour of being euthanized, with the average time of 22 min post euthanasia. We examined birds for any abnormalities and the coelomic cavity for any gross pathology. We also assessed the birds’ body condition using a scale of one to five: one being emaciated and five being over-conditioned with presence of fat in intestinal mesentery. Sex was determined by examining the syrinx .
We collected 0.5 to 2 cm sections of intestine (duodenum, jejunum, ileum, cecum, colon) and bursa of Fabricius in 10% buffered formalin. The tissues were incubated at room temperature for 24–48 h to allow time for fixation, then transferred to a histological sectioning cassette in 70% ethanol and embedded in paraffin within 24 h. We also collected 2 mm sections of ileum and bursa in RNA stabilizing solution (RNAlater®, Sigma-Aldrich, St. Louis, MO, USA) for viral RNA analysis in these tissues.
Viral RNA isolation and RT-PCR
Virus in cloacal swabs, ileum tissue, and bursa tissue was quantified by isolating viral RNA using qPCR targeting the matrix protein gene . Unlike immunohistochemistry which stains for nucleoprotein antigen, qPCR is quantitative and can detect lower quantities of virus . Viral RNA was isolated from cloacal swab material using the MagMAX™-96 AI/ND Viral RNA Isolation Kit (Applied Biosystems® by Thermo Fisher Scientific, Vilnius, Lithuania) with modifications to the manufacturer protocol previously described . Viral RNA was extracted with host mRNA from 15 to 30 mg of ileum and bursa tissue from each bird using the Qiagen RNeasy Mini Kit (QIAGEN®, Hilden, Germany) according to the manufacturer’s protocol. For the RT-PCR working solution we used the TaqMan® RNA-to-Ct™ 1-Step Kit (Applied Biosystems® by Thermo Fisher Scientific, Foster City, CA, USA), primer 5′-AGATGAGTCTTCTAACCGTCTCTG (Sigma-Aldrich, St. Louis, MO, USA), probe 5′-[6FAM] TCAGGCCCCCTCAAAGCCGA [BHQ1] (Sigma-Aldrich, St. Louis, MO, USA), and 2 μL of sample RNA for a final well volume of 10 μL. Each sample was processed at least three times on a 384 well plate with a minimum of three negative control wells and three positive control wells. We used LPAIV H5N9 stock virus in a 10-fold dilution on each plate in three replicates to create a reference standard curve (Additional File 13). Ct values less than 40 were considered positive for virus. Using QuantStudio™ 6 and 7 Flex Real-Time PCR Software System v1.3, we calculated the standard curve, which was used to estimate virus quantity of each sample by correlating Ct values to 50% egg infectious dose per milliliter (EID50/mL). The reported limit of detection is 0.1 EID50 ; therefore, any samples with undetectable viral RNA were considered negative and assumed to be 0.00 EID50/mL. Virus quantity for each sample was averaged across sample replicates. Failed wells and suspected contaminated wells were removed from final calculations.
The quantification limit of the stock virus 10-fold dilution was approximately 400 EID50; however, 21% of our samples were detected to have positive virus between this threshold and 0.1 EID50. To validate the stability of our statistical analysis, multiple value random imputation  was used for any sample with positive virus between 0.1 and 400 EID50, and statistical analysis was repeated. Methods and results of this validation technique are outlined in supplemental material (Additional File 9).
We used lectin histochemistry to detect SAα2,3Gal in formalin fixed and paraffin embedded tissues of the intestines and bursa of Fabricius of each bird. Maackia amurensis I (MAL I) agglutinin is a plant lectin which binds specifically to Siaα2-3Galβ1-4Glc(NAc) [37, 56] and has been used in multiple receptor distribution studies in ducks and other influenza hosts [57, 58] to detect SAα2,3Gal. MAL II, which specifically binds Siaα2-3Galβ1–3 (Neu5Acα2–6) GalNAc , is another lectin commonly used in place of, or in conjunction with MAL I [13, 17, 18, 59]. Trial protocols were tested to determine the proper concentration of each lectin needed for proper binding and visual staining of SAα2,3Gal. The trial protocol resulted in a determined concentration for MAL I, but not MAL II; hence MAL I was the only lectin used given that H5 LPAIVs have similar affinity for the receptors targeted by each lectin [36, 37]; furthermore, any lack of specificity for sialic acid receptors is shared by both lectins .
Paraffin embedded tissue (duodenum, jejunum, ileum, cecum, colon, and bursa of Fabricius) from each bird was sectioned and stained with biotinylated lectin MAL I (Vector Laboratories, Burlingame, CA, USA), using previous described methods [17, 58] with minor modifications. Paraffin embedded tissue sections were deparaffinized and processed with the EnVision FLEX Target Retrieval Solution, Low pH kit wash buffers, blocking agents, and DAB plus chromogen working solution (Agilent, Dako Omnis, Santa Clara, CA, USA). Tissue sections were first treated with 100 μL of 3% Peroxide Block, then Avidin/Biotin blocking agent (Agilent, Dako Omnis, Santa Clara, CA, USA), and protein blocking. The tissue sections were incubated in 100 μL of MAL I for 32 min, and then treated for 20 min in 100 μL of streptavidin peroxidase (Agilent, Dako Omnis, Santa Clara, CA, USA). The working solution (200 μL) was applied and tissue sections were finally counter stained with 100 μL of hematoxylin (Gill’s III, 1:10 dilution) (Astral Diagnostics Incorporated, West Deptford, New Jersey, USA). All tissue sections stained in the same batch were also stained with a known positive control of duck (Anas platyrhynchos domesticus) tissue.
We assessed the abundance of SAα2,3Gal in the proximal intestine (combined duodenum and jejunum), ileum, cecum, colon, and bursa of Fabricius by estimating occurrence frequency of lectin stained cells. We estimated the percentage of lectin stained cells per 5 mm sections of tissue and cell type via an ordinal visual scoring method commonly used in histochemistry , which we called “lectin score.” Using brightfield microscopy (400x), we looked specifically at the bursa epithelial cells, and three cell types in each intestinal tissue: the brush border, villi enterocytes, and crypt enterocytes. We scored as many fields of view (FOV) as possible with a maximum of 10 FOVs per cell type in each tissue (Additional Files 14 and 15). Each FOV received a score based on the estimated percentage of cells stained in that FOV. A score of zero indicated that no cells were stained in that field of view. A score of 5 indicated that 1–10% of cells were stained. A score of 35 indicated that 11–60% of cells were stained. A score of 80 indicated that 61–100% of cells were stained. The scores for the FOVs were averaged to obtain a single score for each tissue and cell type, providing 13 separate lectin scores per bird. All samples were scored by the same individual (AD) to eliminate inter-observer error. In some cases, the tissue had become autolyzed and could not be scored, which was more common for the ileum and bursa tissues in mallards possibly due to longer processing times compared to teals.
Since the scoring method used to quantify the frequency of SAα2,3Gal was based off four categories of scores compared to a quantitative continuous scale, we validated our scoring method with the absolute counts of stained cells for 20 randomly selected birds from mallard groups MT1, MT2, and MT5. For each tissue, a single observer (AD) counted the number of stained cells out of 500 cells for each cell type of the ileum and colon, then calculated the percentage. With a total of 108 counts for 20 birds, we found high agreement between our scoring method and the absolute counts (R2 = 0.79, p < 0.001).
Statistical software R version 3.4.4  was used for all statistical analyses. P-values of less than 0.05 were considered statistically significant and assumptions of normality were met by Log10(value + 1) transforming all virus titer and lectin histochemistry data. These methods were performed for both mallards and teals unless otherwise indicated. All analyses only included virus titer data collected on one to five DPI, when the majority of virus was shed.
For birds sacrificed during the first five DPI, we used simple linear regression to analyze the relationship between virus titers in the cloacal swab, ileum tissue, and bursa tissue, since all three of these variables were collected at the time the bird was sacrificed. Only the cloacal swab collected on the day the bird was sacrificed was used in this analysis. Six total comparisons were evaluated, three for each species (swab vs. ileum, swab vs. bursa, ileum vs. bursa). In each comparison, the effect of DPI was also evaluated.
A repeated measures, linear mixed effects model  was used to test for differences in virus titer or lectin score between species, between sexes, and between control and infected birds (lectin score only). To account for repeated measures of individuals birds, each model was adjusted with a random intercept for each bird. Additionally, when variances of virus titer were different between the factors of the main effects, the model was adjusted to allow for unequal variances. Differences in variance were detected using the Fligner-Killeen test . ANOVA tables were visualized, and the post-hoc Tukey’s test was used to assess pairwise differences.
To analyze the effect of species on virus titer, species, DPI, and the species*DPI interaction were included in the model. To analyze the effect of sex on virus titer, we assembled two separate models: one for mallards and one for teals. Sex and DPI, plus their interaction, were included in each model.
To analyze the effect of lectin score on infection status (infected vs. control), mallards and teals were assessed in two separate models. For each species, infection status and tissue/cell-type, plus their interaction, were included in their respective model.
Using data from infected birds only, we also assessed species and sex-based differences in lectin score. To analyze the effect of species on lectin score, species, tissue/cell type, and their interaction, were included in the model. To analyze the effect of sex on lectin score, mallards and teals were analyzed in separate models. Sex and tissue/cell type, plus their interaction, were included in each model.
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 PCA to reduce the data into one component variable we called “[tissue type] PC.” Each PC variable accounted for greater than 80% of the variation between the cell types of that particular tissue. PC variables generated from the PCA were used in the MLR models to determine the relationship between virus titer and lectin score.
Virus titer and lectin score relationship was determined by assessing three different MLR models for each species using virus titer as the dependent variable. The virus titer variable in the first model consisted of virus titers from cloacal swabs collected on the DPI each bird was sacrificed. The second model used virus titers in ileum tissue, and the third model used virus titers in bursa tissue. Independent variables for the cloacal swab virus titer model consisted of the lectin score variables, the principal components described above (when appropriate), and five control variables: sex, BCS, LPAIV treatment group, body mass in grams at 55 days after hatch, and inoculation age in days. Independent variables for the ileum virus titer model included only the ileum lectin score variables and the five control variables. Only the bursa epithelium lectin score variable and the five control variables were included in the bursa virus titer model.
To determine the best fitting MLR model for each dependent variable, we followed a consistent procedure. Global linear models were tested for each dependent variable separately. To select parsimonious model fits to the data, we used stepwise variable selection based on the generalized Akaike’s Information Criterion (AIC). We then used variance inflation factor (VIF) scores 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 . When two VIFs were > 3.0 and < 1.0 in difference, we tested alternative models. Stepwise variable selection was used for each model to ensure the best fitting model. Residual plots were reviewed. For each of the three dependent variables, the model with both the lowest AIC, highest adjusted R2, and satisfactory residual patterns (e.g., no linear or nonlinear trend in residuals, little to no heterogeneous variance in residuals, and no suspected outlier observations) was chosen as the best fitting model to the data.