MLNWR (elevation 590 m) is located in Sheridan County in northeastern Montana (48°27′N, 104°23′W). The refuge covers 13,000 hectares, including Medicine Lake (3,320 hectares), the largest natural lake in eastern Montana. Extensive wetlands provide suitable breeding habitat for mosquitoes and aquatic birds. Cropland and short-grass prairie surround the lake and provide nesting grounds to ≈125 species of birds. Approximately 4,000 breeding pairs of pelicans nest on a narrow peninsula (length ≈500 m) (7).

Data for human WNV disease cases were obtained from ArboNET, an internet-based passive surveillance system maintained by the Centers for Disease Control and Prevention (Fort Collins, CO, USA) in collaboration with state and local health departments. Locations of colonies were obtained from King and Anderson (8). Wildlife Mortality Quarterly Reports published by the US Geological Survey National Wildlife Health Center provided locations and dates of WNV-related pelican deaths during 2003–2007 (www.nwhc.usgs.gov/publications/quarterly_reports/index.jsp). We used an odds ratio to compare incidence of WNND cases in counties with reported WNV-associated deaths at pelican-nesting colonies to that in all other counties with pelican colonies during 2003–2007. Positive and negative predictive values were defined as the percentage of counties with pelican-nesting colonies in which a pelican WNV die-off and human WNND case(s) occurred, and in which neither a pelican WNV die-off nor a human WNND case occurred, respectively, during a given year.

Although neuroinvasive and nonneuroinvasive disease cases are reportable, reporting of nonneuroinvasive WNV disease has varied substantially by jurisdiction and over time. Therefore, only WNND cases were considered. For each year and county that pelican and human WNV disease were observed, we determined the interval between the earliest collection date of WNV-positive pelicans and the earliest onset date of human WNV disease.

Mosquitoes were collected from MLNWR by using battery-powered miniature light traps (J.W. Hock, Gainesville, FL, USA) supplemented with CO₂ from a 9-kg compressed gas tank. In 2006, seven traps were placed at 5 locations on the northeast perimeter of Medicine Lake; 3 were at Bridgerman Point, 10–200 m from pelican-nesting and -congregation sites (9). In 2007, all 5 traps were placed at Bridgerman Point. Traps were generally operated for 2 consecutive nights each week from mid-May through August in 2006 and 2007, and collections were stored at –20°C for >24 h before transport on dry ice. Collections were processed on a chill table and the light trap index (LTI) was calculated for each week as the number of trapped Culex tarsalis mosquitoes per trap night.

For WNV testing by reverse transcription–PCR (RT-PCR), weekly trap collections were sorted by species and location. Pools of <50 adult female mosquitoes were homogenized in vials containing 4.5 mm-diameter copper-coated steel beads (BB pellets) in BA-1 medium (medium 199 with Hanks balanced salt solution, 0.05 mol/L Tris buffer, pH 7.6, 1% bovine serum albumin, 0.35 g/L of NaHCO3, 100 mg/L streptomycin, 100 U/mL penicillin G, 1 μg/mL amphotericin B) and clarified by centrifugation. RNA was extracted from the supernatant and purified through an EasyMag extractor (bioMérieux, Durham, NC, USA) by using automated magnetic silicon extraction. Purified RNA was transcribed into cDNA and amplified by using specific WNV primers as described (10) in an EasyQ thermocycler (bioMérieux). Detection of WNV cDNA was achieved by agarose gel electrophoresis. Prevalence of WNV infection in mosquito populations was estimated by using PooledInfRate (www.cdc.gov/ncidod/dvbid/westnile/software.html). Vector index was calculated as the product of the LTI and infection rate (11).

Blood-engorged Aedes vexans and Cx. tarsalis mosquitoes were stored individually at –20°C and processed for blood meal identification as described (12). Briefly, each mosquito was homogenized and DNA was extracted. A portion of the mitochondrial cytochrome B gene was amplified and sequenced, and the resulting sequence was compared with sequences in a database for species identification.

Moribund pelican chicks (≈6–12 weeks of age) showing clinical signs suggestive of WNV infection (e.g., ataxia, torticollis, reluctance or inability to move) (Figure) were killed by cervical dislocation and stored frozen until necropsy. In 2006, carcasses were collected from July 25 through August 11 (n = 8) during a period of maximum chick deaths, and in 2007 from July 11 through August 1 (n = 24) after confirmation of WNV in mosquito pools. Also in 2007, oropharyngeal and cloacal swab samples, skin, and feather samples were collected from 23 carcasses in the field before they were frozen for comparison with samples collected from the same animals in the laboratory during necropsy.

For swab samples, dacron-tipped applicators were inserted into the oropharyngeal cavity (behind the pouch) or into the cloaca and then submerged and swirled in vials containing 1 mL BA-1 medium and discarded. In 2007, eye swab samples were collected from 17 carcasses by placing the applicator tip between the inner membrane of the eyelid and the eye. In addition, pouch lice (Piagetiella peralis) were individually removed from the inner lining of the pouch of each pelican and pooled in cryovials (<40 lice/pool). Four flight feathers were removed from each carcass (2/wing). Feathers were removed from the follicle, and the calamus (quill tip) was aseptically cut and placed with the associated pulp into a vial containing 1 mL BA-1 medium. Approximately 0.5 cm³ each of skin, kidney, spleen, heart, lung, and brain was aseptically collected and placed in cryovials containing 1 mL BA-1 medium for cryopreservation.

Tissues and chewing lice, obtained from the inside of throat pouches, were homogenized in a mixer mill (5 min at 25 cycles/s; Retsch GmbH, Haan, Germany) in 1 mL of BA-1 medium containing 20% fetal bovine serum and a BB pellet. Homogenates were clarified by centrifugation (12,000 × g for 3 min), and an aliquot was removed for immediate testing. Remaining supernatants were stored at –80°C.

Virus isolation was performed for tissues, lice homogenates, and swabs by using a Vero cell plaque assay as described (13). Viral plaques were confirmed as WNV by RT-PCR or VecTest WNV Antigen Detection Assay (Medical Analysis Systems, Camarillo, CA, USA) as described (2). RT-PCR and plaque assay detection methods were compared within tissue types by using the Fisher exact test with Bonferroni correction for 9 comparisons (α = 0.0056). For specimens collected in the field and their carcass-matched controls collected in the laboratory, test results were compared by using the κ statistic for concordance.

RT-PCR methods for detection of WNV RNA in tissues were according to those of Lanciotti et al. (14), except for use of the Viral RNA Minikit (QIAGEN, Valencia, CA, USA) for RNA extraction and the Bio-Rad Icycler IQ Real-time Detection System (Bio-Rad, Hercules, CA, USA) for cDNA amplification. A cycle threshold value <37 was considered positive for target sequence amplification. Samples were screened with 1 pair of primers (genome positions were 10668 for forward primer, 10770 for reverse primer, and 10691 for probe) and positive results were confirmed with a second pair of primers (genome positions were 1160 for forward primer, 1229 for reverse primer, and 1186 for probes).

The probability of human WNND cases in counties with pelican nesting colonies increased 5× when WNV-associated deaths occurred among the pelicans (odds ratio 5.0, 95% confidence interval 1.9–13.0, n = 135 county-years). The positive predictive value of pelican deaths for human WNND cases was 55%, and negative predictive value was 81%. Pelican deaths were observed an average of 23.1 days (median 13.5 days) before human case onset and occurred before human disease onset in 12 (75%) of 16 county-years (Table 1).

In 2006, a total of 414 Cx. tarsalis mosquitoes were captured in 67 light trap-nights from July 1 through August 5. Weekly LTI values for Cx. tarsalis mosquitoes ranged from 4.1 to 9.0 during July and early August (Table 2) when mosquito populations were low because of severe drought (larval production sites were dry or contained ephemeral water). None of 12 pools of Cx. tarsalis mosquitoes assayed were positive for WNV RNA.

In 2007, a total of 25,291 Cx. tarsalis mosquitoes were captured in 42 light trap-nights from June 30 through August 8 (Table 2). Weekly LTI values during this period ranged from 169 to 1,643. WNV was detected in 28 (32.2%) of 87 mosquito pools by RT-PCR, with the first positive mosquito samples collected during the week of July 8, 2007. WNV was detected in 10 of 20 pools of mosquitoes collected during the last week of July; we observed an estimated infection rate of 10.7/1,000 mosquitoes.

Vertebrate DNA sequences were obtained from blood-engorged abdomens of 22 mosquitoes collected in 2007, including 8 Ae. vexans and 14 Cx. tarsalis. All 22 mosquitoes had fed on American white pelicans.

Twenty-seven (84.4%) of 32 pelicans sampled had >1 tissues positive for WNV by plaque assay compared with 7 (87.5%) of 8 positive by RT-PCR (Table 3). Pelicans with WNV-positive tissues were collected from July 25 through August 11, 2006, and July 11 through August 1, 2007. Because pelicans were not collected before these dates, the timing of initial onset of WNV outbreaks in pelicans is unknown.

Skin was the most efficacious tissue for WNV detection in pelican carcasses. Viral loads were greatest in feather pulp, brain, heart, and skin. RT-PCR and plaque assay results were similar; detection rates did not differ among specific tissues or between field-collected vs. laboratory-derived samples (Table 4). Concordance (i.e., test agreement) was 82% (κ = 0.82) among matched field and laboratory samples from the same carcasses. All pouch lice samples were negative for WNV.