Isolates from New York State and the Eastern Seaboard were sequenced for this study (online Appendix Table, available from http://www.cdc.gov/EID/content/14/3/454-appT.htm). Strains originating outside of New York were isolated from avian serum samples, which were collected during a study conducted by the US Geological Survey and stored at –80°C until inoculation onto cell culture. All EEEV strains from within New York were collected from mosquito, avian, and equine samples that were submitted to the Wadsworth Center’s Arbovirus Laboratories as a part of surveillance efforts by the New York State Department of Health. EEEV strains isolated during 1971–1975 were obtained from our archives.

To obtain mosquito-derived EEEV strains, mosquitoes were collected from May through October by local county health department staff, who used standard miniature light or gravid traps. Mosquitoes were identified to the species level, and pooled samples of 10–50 mosquitoes in 2-mL microfuge tubes were submitted to the Arbovirus Laboratories for analysis. Tubes contained a steel ball-bearing (Daisy Brand, Rogers, AR, USA), and to each tube we added 1 mL of mosquito diluent (20% heat-inactivated fetal bovine serum [FBS] in Dulbecco’s phosphate-buffered saline with 50 μg/mL penicillin/streptomycin, 50 μg/mL gentamicin, and 2.5 μg/mL amphotericin B). Pools were homogenized by using a mixer mill (Retsch, Haan, Germany) at 24 cycles/s for 30 s and centrifuged for 4 min at 6,000 rpm at room temperature. The clarified homogenate was transferred to a new microcentrifuge tube and stored at –80°C until testing.

To obtain vertebrate-derived EEEV strains, samples of horse brains were submitted by the Wadsworth Center’s Rabies Laboratory, and samples of avian kidney, heart, and brain were submitted by the New York State Department of Environmental Conservation Wildlife Pathology Unit. Each horse brain was excised into 3 separate 1–3 mm³ sections, which were pooled for testing for each horse. Avian tissues were tested by excising the same-size portion from 1 of the tissues or by pooling sections of all 3 tissues. Excised tissues were placed in 2-mL microfuge tubes containing a ball-bearing and 1 mL of BA-1 virus diluent (M199 with Hanks’ salts and L-glutamine; [Mediatech, Herndon, VA, USA] in sterile distilled water with 0.05 M Tris(hydroxymethyl)aminomethane,1% bovine serum albumin, 0.35 g/L sodium bicarbonate, 100 U/mL penicillin, 100 U/mL streptomycin, 1 µg/mL amphotericin B, and 20% FBS). Hydrogen chloride was added to the diluent to bring the pH to 7.4. Samples were homogenized on the mixer mill at 24 cycles/s for 4 min and centrifuged for 5–8 min at 10,000 rpm at 4°C.

Virus was isolated by inoculating 100 mL of supernatant from mosquito pools, vertebrate tissues, or avian serum onto confluent monolayers of African green monkey kidney (Vero) cells grown in 6-well plates. Plates were incubated for 1 h at 37°C in 5% carbon dioxide, with gentle rocking every 15 min. To each well, 3 mL of maintenance medium (1× minimum essential medium with Earle’s salts [Invitrogen, Carlsbad, CA, USA], 2% heat-inactivated FBS, 1% 100× L-glutamine, 0.15% sodium bicarbonate, 1% penicillin–streptomycin, 0.1% amphotericin B, and 0.1% gentamicin diluted in sterile distilled water) was added, and plates were returned to the incubator and observed daily. If cytopathic effect was observed, the infecting virus was identified by either immunofluorescence assay (14) or reverse transcription PCR (RT-PCR) by using One-Step RT-PCR kit and protocol (QIAGEN, Valencia, CA, USA) as directed by the manufacturer. Primer sequences and cycling parameters are described elsewhere (15). Target bands were examined under UV light after electrophoresis on a 1.5% agarose gel. Samples that were positive for EEEV were stored at –80°C until use.

RNA was extracted from the Vero cell culture supernatant after a single passage, the original sample (no passage), or isolates with an unknown passage history by using the RNeasy kit (QIAGEN) as directed by the manufacturer. The entire E2 coding region was amplified in 2 separate reactions. To produce overlapping fragments, primers EEE8460 (5′-AGAATCCACACGAAACACTCACCA-3′) and EEE9200c (5′-ATCCGTGCAGGTGGTTGTATGGTC-3′) were used for the first reaction, and primers EEE9105 (5′-TCCACAGTGCCAAGGTGAAAA-3′) and EEE9887c (5′-CTGCAAGTGGGATAAGCGTCTG-3′) were used for the second reaction. The partial NSP3 coding region was amplified by using primers EEE4836 (5′-CAGAGCGAGTTTACAGATTACG-3′) and EEE5477c (5′-AACGGCGAACGACTGAA-3′). Sample RNA (5 μL) was added to 45 μL of One-Step RT-PCR master mix (QIAGEN) prepared according to the manufacturer. Several drops of mineral oil were added on top of each reaction. Samples were reverse transcribed for 30 min at 55°C and heat inactivated at 95°C for 5 min. To eliminate residual RNA, RNase was added to each reaction after reverse transcription. Samples were amplified by PCR according to the following thermocycler conditions: 94°C for 10 min; 39 cycles of 94°C for 30 s, 55°C for 30 s, 72°C for 45 s; and 72°C for an additional 10 min. PCR product (40 μL) was added to 4 µL of BlueJuice loading dye (Invitrogen, Carlsbad, CA, USA), and loaded onto an agarose gel containing 0.4 μg/mL of ethidium bromide. DNA underwent electrophoresis and was examined under UV light. Samples were sequenced on either an ABI 3100 or ABI 3700 automated DNA sequencer (Applied Biosystems, Foster City, CA, USA) at the Wadsworth Center Molecular Genetics Core facility. (Sequencing primers are available from the authors upon request.)

Trace files were compiled by using the SeqMan module of Lasergene (DNAstar, Madison, WI, USA), with a minimum of 2-fold base-call redundancy required for all sequences. Consensus sequences for each strain sequenced in this study and reference strains obtained from GenBank were aligned by the ClustalV method (16) in the MegAlign module of Lasergene. Two phylogenetic trees were produced for the E2 coding region analysis: the main E2 tree, which included all strains included in this study, and the E2 subset tree, which contained only strains sequenced in this study and only those sequences from GenBank for which both E2 and NSP3 sequence data were available. The main E2 tree was generated by maximum likelihood in PAUP version 4.0b10 (Sinauer Associates, Sunderland, MA, USA) by using the HKY85+G model with relevant parameters estimated from the data. The robustness of the branching pattern was estimated by performing 1,000 neighbor-joining bootstrap replicates under the maximum-likelihood substitution model, also using PAUP; these values are presented on the maximum likelihood tree. The E2 subset and NSP3 trees were generated by neighbor-joining analysis with 1,000 bootstrap replicates by using the Kimura 2-parameter model in MEGA2 (17). Nucleotide sequences of newly sequenced strains were deposited in GenBank (see online Appendix Table for accession numbers). Nucleotide diversity (π) and the Tajima D statistics were computed by using DnaSP (18).

We sequenced 42 EEEV strains, which represented various geographical locations, hosts, and isolation dates (online Appendix Table). Of the 42 strains, 29 were from central upstate New York. Of those 29 strains, 13 were isolated during the 1970s epizootic, and the remainder were isolated during the current epizootic. The remaining strains sequenced were from various locations in the southeastern United States (Louisiana, Florida, and Virginia), New Jersey, and New York outside the central upstate focus (counties of Suffolk, Orange, Sullivan, Ulster, and Chemung) (Figure 1).

Phylogenetic analysis of the E2 coding region demonstrated that all isolates sequenced in this study belonged to lineage I (Figure 2), and showed strong spatiotemporal clustering. The 3 strains isolated from Oswego County in 1971 (NY71a, NY71b, and NY71c) clustered together as a result of almost identical E2 coding regions. All 8 strains isolated from Oswego County in 1974 (NY74a–h) and the only strain isolated in 1975 (NY75) grouped together strongly and formed the Oswego74 clade (Figures 2–4). Strong clustering was also evident among strains isolated mainly from Onondaga County during 2003–2005 (the Onondaga03 clade, Figures 2–4). Of the 16 strains in this clade, 13 had identical E2 coding regions (data not shown). Both NY04g and NY04j grouped together and were isolated in close geographic proximity in Sullivan and Ulster Counties, respectively, in the lower Hudson Valley.

Examination of π, a measure of sequence diversity, confirmed the close relationships of sequences sampled during spatiotemporally defined epizootics. The π values for the Oswego74 and Onondaga03 clades were 0.00035 and 0.00030, respectively. The π value for the entire set of US strains included in our analysis was 0.01058, ≈30× greater than intraepizootic values. The Tajima test failed to reject neutrality in either the Oswego74 or Onondaga03 clade because of extremely low genetic diversity: only a few mutations were present in each grouping. Thus, the EEEV collected during New York epizootics is generally characterized by a high degree of sequence conservation with little genetic variation among spatially and temporally related strains.

However, spatiotemporal conservation was not absolute. NY73, isolated from a horse in Onondaga County in 1973 (NY73) was most genetically similar to NY69, isolated from a pheasant in Suffolk County on Long Island in 1969. Similarly, a strain isolated from a horse in Chemung County, New York (NY04k), fell into the Onondaga03 clade, which further demonstrated occasional relaxation of the otherwise strong time-space clustering of the strains studied. Consideration of strains from outside of New York provided additional instances in which the pattern of spatiotemporal clustering was broken. Well-supported subclades frequently contained southern progenitor strains that had been isolated years before they appeared in New York or New Jersey (Figures 2–4). Examples of this trend include the following: VA03 linked with NY03b and NY04f, GA97 linked with NJ03a and NJ03b, and FL02a linked with NY04g and NY04j. In addition, strain FL02b, isolated from an ovenbird (Seiurus aurocapillus) collected in Florida during 2002 was highly similar to the Onondaga03 clade.

To evaluate the possibility that analysis of different coding sequences would yield different results, we studied the NSP3 coding sequences of all strains for which sequence data were available or further sequencing was possible. GenBank sequence data for the NSP3 coding region was limited; therefore, another phylogenetic tree for the E2 coding region was produced by using a subset of lineage I strains for which the NSP3 sequences were available (Figure 3). This E2 subset tree was used for comparisons with the NSP3 tree to determine whether the trees shared similar topology. Phylogenetic analysis of the NSP3 coding region produced similar overall topologies (Figure 4); both trees recognized the major clades (Oswego74 and Onondaga03) and most of the minor subclades.