All animal work was carried out in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. The protocol was approved by the CDC Institutional Animal Care and Use Committee, Protocol # 07–011. All efforts were made to minimize suffering.

Five to seven-day-old and two- to three week-old chicks (Gallus gallus) (Northern Colorado Feeder Company, Fort Collins, CO) and eight- to twelve-week-old female Syrian golden hamsters (Mesocricetus auratus) (Harlan Sprague Dawley Inc., Indianapolis, IN) were used. Chicks were housed individually and hamsters in groups of two in large mouse cages. House sparrows (Passer domesticus) were captured by mist nets (Avinet Inc., Dryden, NY) in Weld County, Colorado and transported in rubber-coated wire-mesh finch flight cages (PetCo Inc., Fort Collins, CO). Sparrows were aged using standard plumage characteristics [8], sexed, and banded with uniquely-numbered aluminum leg bands. A 0.3-mL blood sample was taken by jugular venipuncture from each bird. Blood was collected directly into Microtainer serum separators (Beckton-Dickinson Co., Franklin Lakes, NJ). Sparrows were given a two-week acclimation period in captivity prior to experimentation. During this period they were physically examined by the attending veterinarian, treated with anti-helminth medication (Avermectin), and transferred in groups of 12 to screened non-human primate cages equipped with perches. Animals were provided food and water ad libidum: standard rodent feeder diet (hamsters), hen grower pellets (chicks), or mixed bird seed (sparrows). Wild-caught sparrows were confirmed negative for neutralizing antibody to WNV and HJV by plaque reduction neutralization test (PRNT) [9], using challenge doses of approximately 100 plaque forming units (pfu) of WNV (strain NY99-4132) or HJV (strain B230). Birds with >70% neutralization of WNV were excluded from experimental infection with that virus. Conservatively, birds with >50% neutralization of HJV were excluded from infection with that virus due to the probability that birds exhibiting alphavirus neutralizing antibody most likely had a past infection with a member of the Western equine encephalitis virus (WEEV) complex.

Prior to blood sampling, hamsters were anesthetized with 0.025 mL/10 g body weight using a mixture of ketamine (dose 200 mg/kg) and xylazine (10 mg/kg), injected intraperitoneally (ip). Chicks and sparrows were administered 0.01 mL/10 g body weight ketamine (dose 50 mg/kg) and xylazine (10 mg/kg) mixture intramuscularly (im). Following experimentation, animals were euthanized by cervical dislocation performed under anesthesia.

Aedes albopictus (Lake Charles, LA strain) eggs were hatched by placing egg papers submerged in de-ionized water in a vacuum for approximately 10 minutes. Flooded eggs were left in the vacuum for one hour before transferring to larval rearing pans containing 0.4% liver powder solution. Culex quinquefasciatus (Sebring strain) larvae were reared on liver powder solution and rabbit chow. Adult mosquitoes were maintained in environmental chambers held at 27°C with approximately 90% humidity, on a 12L:12D cycle. Mosquitoes were provided 5% sucrose solution which was removed 24–48 h prior to experimental blood feeds. Adult mosquitoes were at least four days old at the time of experimentation.

Animals, viruses, and sampling days are presented in Table 1. Viruses were selected for their viremogenic potential in chicken, hamster, and house sparrow [1], [10]–[13]. All viruses were obtained from the reference collection at the Division of Vector-Borne Diseases, Centers for Disease Control and Prevention (Fort Collins, CO). Animals were needle-inoculated subcutaneously with approximately 10⁴ pfu of WNV (strain NY99-4132) or HJV (strain B230) virus suspension in 0.1 mL BA-1 media (Hanks M-199 salts, 0.05M Tris pH 7.6, 1% bovine serum albumin, 0.35 g/L sodium bicarbonate, 100 U/mL streptomycin, 1 µg/mL Fungizone). All birds were inoculated in the breast and hamsters in the lower abdomen with a 26 g ½-in needle attached to a tuberculin syringe. Inoculations were accompanied by the bite of 1–3 uninfected Ae. albopictus mosquitoes at the site of virus inoculation to more closely simulate natural infection [4].

Between days one and four post-inoculation when peak viremia was expected [1], [13]–[14], blood was collected by mosquitoes and syringe. Ae. albopictus and Cx. quinquefasciatus mosquitoes were starved 24–48 h prior to exposure to infected animals, and fed either by placing anesthetized animals onto screened pint cups containing 10–15 female mosquitoes or by holding a 2-dram screened vial containing 5–10 female mosquitoes to a manually-restrained animal (two- to three-week-old chicks only). After feeding, mosquitoes were killed by freezing. Fully engorged mosquitoes were homogenized in TenBroeck glass tissue grinders containing either one or two milliliters of BA-1+20% fetal bovine serum (FBS), individually or in pools of two to five mosquitoes. Mosquito homogenates were transferred to 2.0 mL microcentrifuge tubes and centrifuged for two minutes at approximately 6,000×g. Supernatants were transferred to cryovials and stored at −80°C until testing (chicken and hamster). House sparrow samples were immediately inoculated onto Vero (African green monkey kidney) cell monolayers.

Syringe samples were taken within 20 minutes of mosquitoes feeding on each animal. 0.1 mL blood was collected using a syringe attached to a 26 g ½-in SubQ needle. Blood was collected by jugular venipuncture of chicks on two occasions (one day apart) for each chick, and of sparrows on one or two occasions (three days apart), and by cardiac puncture of hamsters (post-mortem). Syringe samples were either expelled directly into Microtainer serum separators and centrifuged to isolate the serum for testing, or diluted in 450 µl BA-1 diluent (to achieve a 1∶10 dilution of serum). Blood samples were allowed to coagulate at room temperature for up to 30 minutes before placing them on ice.

Blood samples taken by syringe and mosquito were titrated simultaneously using the double-overlay Vero cell plaque assay [9]. Vero cell monolayers in 6-well polystyrene culture plates (Costar Inc, Cambridge, MA) were inoculated with 0.1 mL of diluted serum in duplicate. For syringe samples, 10-fold dilutions from 10⁻¹ to 10⁻⁹ were tested, and for homogenized mosquito samples, 10-fold dilutions from undilute to10⁻⁵ were tested. The second overlay, containing neutral red, was added on the second day post-inoculation (dpi) for HJV, and the third dpi for WNV. HJV plaques were counted at 2 and 3 dpi, and WNV at 3 and 4 dpi to ensure no plaques were missed. For titer calculations, 2-µl blood meals (1 µl serum) were assumed [15]–[17]. Mean daily titers were calculated by averaging the log titers for multiple mosquitoes fed on the same animal, then averaging the log titers for animals each day.

To ensure that any differences in titer between mosquito and syringe samples were not due to virus binding to mosquito tissues during blood feeding or grinding, six replicates each of 1 mL BA1+20% FBS containing approximately 10² pfu HJV or WNV were incubated at 28°C for 45 minutes with or without a homogenized mosquito. Virus titers of paired samples were compared by Vero cell plaque assay.

Three questions of interest concerning the association between virus log titer measurements obtained via syringe sampling and mosquito sampling of blood from a given animal were addressed: 1) as a function of the syringe-sampled log titer, what is the probability that the mosquito-sampled log titer would be positive, i.e., would detect virus? 2) among animals with positive log titer measurements using both mosquito and syringe sampling, at what syringe-sampled log titer would there be a 25%, 50%, or 75% probability that the corresponding mosquito-sampled log titer would be measurable, and what is the expected mosquito-sampled log titer as a function of the syringe-sampled log titer? 3) for each of the questions 1 and 2, do these associations vary by animal host (chicken, hamster, sparrow) or virus (HJV, WNV)?

A mixed discrete and continuous regression model was used to evaluate all three of these questions simultaneously. The discrete component of the model consisted of a logistic regression for log titer positivity, where the logit of the probability of a positive mosquito-measured log titer was modeled as a linear function of the syringe-measured log titer. The continuous component of the model was a linear model of the mosquito-measured log titer as a function of syringe-measured log titer, though no intercept was included in the model to reflect the fact that no mosquito-sampled values would be positive when the syringe-sampled values were 0. For both discrete and continuous components of the model, variables were included to permit different intercepts (discrete only) and slopes for animal and virus, including interactions. The models were fit using maximum likelihood, 95% profile confidence intervals (CI) for the model parameters were computed, and models were compared and selected using the likelihood ratio test and 5% significance. Inversion of the logistic model component provided estimates and 95% CIs of syringe-measured log titer values expected to yield positive mosquito-measured log titer values with probabilities of 25%, 50%, 75% and 90%. Standard regression diagnostics were used to evaluate model assumptions.

Analyses were carried out and graphs produced in the R statistical software package [18], and maximum likelihood (ML) estimation and associated inference was done using the bbmle package in R [19].

For analysis of WNV and HJV viremia profiles (log titer over time), we took two approaches. For a direct comparison of mosquito-sampled versus syringe-sampled log ‘titers at each dpi separately, we used the paired t-test, which accounts for the paired observations on individual house sparrows. These comparisons by dpi within virus were adjusted for multiple comparisons using the Bonferroni adjustment. We then used mixed linear models to evaluate potential trends in the association between dpi and titer, including collection type (mosquito or syringe) as predictors; interaction of collection type with dpi was also included. Because titers were only measured at four time points, we interpret results of this modeling as indicative of general trends in the titers over time, rather than as definitive descriptions of the profiles. Estimation was done using ML when comparing fixed effect model parameters using the likelihood ratio tests, while final model parameters were estimated with restricted ML (REML). For these analyses, there were minimal 0-valued titers, so we did not model the probability of obtaining a 0 titer as a separate model component. Models were fit in R using the nlme package [20].

Starved Ae. albopictus mosquitoes usually fed within a few minutes of exposure to the animal, whereas Cx. quinquefasciatus did not feed as readily. Therefore, feedings with Cx. quinquefasciatus were stopped after the first few days and continued with only Ae. albopictus. Aedes mosquitoes fed to repletion more quickly when starved 48 h and fed on anesthetized animals as compared with manually-restrained animals.

All birds were >3 months old at the time of capture. Initial sampling indicated that 18% (13/71) of sparrows were positive for neutralizing antibodies to WNV, and 11% (8/71) were positive for alphavirus-neutralizing antibodies. Birds with prior exposure to WNV were excluded from experimental infection with WNV. Similarly, birds with alphavirus neutralizing antibodies were excluded from experimental infection with HJV.

The mixed discrete and continuous regression model fits and model comparisons using the likelihood ratio test resulted in a final model with no statistically significant differences in the logistic component’s intercept and slope by either animal or virus. In contrast, a statistically significant difference was found in the slopes of the linear component by virus, though not by animal (Table 2). Model parameter estimates and 95% CIs for the final model are given in Table 3.

Chicken, hamster, and house sparrow all developed detectable viremia for both WNV and HJV infections. The probability of detecting virus in a mosquito blood meal increased with virus titer (as syringe-measured; Fig. 1). Detection of virus in a single mosquito blood meal was limited to titers >10³ pfu/mL serum, (approximately one pfu in one microliter of serum in a blood meal) because of volumetric constraints of the mosquito blood meal size. However, this obstacle was overcome by testing multiple individual mosquitoes fed on the same viremic animal. For a 25% probability of detecting virus in a single mosquito blood meal, the syringe log titer needed to be ≥2.72 log₁₀ pfu/mL (95% CI 2.19–3.27), while for a 50% probability of detection, the syringe log titer needed to be ≥3.64 log₁₀ pfu/mL (95% CI 3.20–4.08). Corresponding syringe log titers for 75% and 90% probabilities of detection were ≥4.56 log₁₀ pfu/mL (95% CI 4.02–5.10) and ≥5.48 log₁₀ pfu/mL (95%CI 4.71–6.24), respectively. These reference log titers are shown in Figure 1, panel (c).

In general, good correlation was seen between mosquito and syringe samples at titers ≥5.0 log₁₀ pfu/mL serum as compared with titers <5.0 log₁₀ pfu/mL serum (Fig. 1). Both estimates for the slopes by virus in the linear component of the regression model were essentially 1, indicating that when the virus was detected via mosquito sampling, the log titer results matched the syringe-measured results (Table 3). The mean difference in log titer between mosquitoes and syringe was not significantly different by virus or by day post inoculation, except for HJV 2 dpi(Table 4). Ninety-two percent (24/26) of sparrows with virus titers >10⁵ pfu/mL serum had mosquito- and syringe-derived titers within one log of each other. Of the remaining two birds with virus titers >10⁵ pfu/mL, the average mosquito titer was 1.1 log₁₀ pfu/mL (SE+0.06) greater (HJV 1 dpi), and 1.8 log₁₀ pfu/mL (SE±0.1) less (WNV 3 dpi) than the corresponding syringe titers, but the mean difference between mosquito- and syringe-derived titers at these time points across all birds and mosquitoes was not significant (Table 4).

There were 38 instances where multiple mosquitoes fed on the same viremic sparrow (HJV or WNV) and each mosquito was analyzed separately. When sparrow virus titers were >5.0 log₁₀ pfu/mL serum, mosquito titers varied on average by 0.44 log pfu/mL serum from each other (range 0–1.5 log difference, n = 22). When sparrow titers were <5.0 pfu/mL serum, mosquitoes fed on the same animal had titers that varied on average by 2.7 log pfu/mL serum (range 0.2–4.7 log difference, n = 16).

Viremia profiles (mean daily viremia) were generated for house sparrow infected with HJV (Fig. 2) or WNV (Fig. 3). Comparison of the differences between mean log-titers derived by mosquito versus syringe at each dpi using the paired t-test adjusted for multiple comparisons within virus yielded no statistically significant differences for either WNV or HJV for any dpi. There were no observations for HJV at 4 dpi for comparison. We fit a linear mixed effects model for log-titer as a function of dpi for the HJV group, and a quadratic for log-titer as a function of dpi for the WNV group. The model for HJV indicated statistically significantly different slopes for the mosquito (−3.1; 95% CI −3.6, −2.6) and syringe collection methods (−2.0; 95% CI −2.6, −1.4). The quadratic model for WNV indicated no statistically significant difference in regression parameters by collection method. As noted in the methods section, we consider these modeling results as generally capturing the trends in these viremia profiles over this timeframe, but caution that we do not interpret these functional (i.e., linear and quadratic) relationships as definitive viremia profiles, but rather descriptive of trends.

Mean virus titers were similar for WNV and HJV incubated with and without a homogenized Ae. albopictus mosquito (WNV 2.8 log₁₀ pfu/mL vs. 2.9 log₁₀ pfu/mL, HJV 3.0 log₁₀ pfu/mL vs. 3.0 log₁₀ pfu/mL, respectively). Therefore, any differences in virus titer between mosquito- and syringe-derived blood samples (i.e. day two titers for HJV in house sparrows) could not be explained by virus adsorption onto mosquito tissues during feeding or grinding.