The morphological data for the new species were obtained from 96 soldiers, 141 alates, and 20 soldier mandible pairs. The number of soldier and alate specimens examined for the four previously described species ranged from 32 to 431 [29]. Genetic data for the new species were obtained from 156 specimens collected from Sapelo Island, GA (McIntosh Co.), Thomasville, GA (Thomas Co.), Havelock, NC (Craven Co.) and Branford, FL (Suwannee Co.).

All specimens were preserved in 70–100% ethanol. The number of specimens examined and collection information for the new species description are listed in Table 2.

Accurate Reticulitermes species attributions are best determined using both soldiers and alates from the same collection site supported by genetic and behavioral information. The specific measurement points used for the quantitative dataset are shown in Figure 1 and Figure 2.

Four characters were measured on soldier head capsules: length without mandibles (sl), width (sw), and two separate angles of curvature for the right mandible (smac1, smac2). A fifth character, ratio of length:width (sl:sw), was calculated to determine its usefulness as a character for species separation.

Soldier head capsules were removed from the body and mounted by placing a minuten pin into the occipital foramen. The opposite end of the minuten was positioned into a cube of foam mounted on a standard size # 2 insect pin. Soldier head capsule length (sl) was measured from the clypeal sulcus to the posterior edge as seen from a dorsal view (near the occipital foramen), and width (sw) was measured at a 90° angle from the mid-point of sl (Figure 1).

Soldier mandibles were dissected from head capsules and mounted on two-sided tape positioned inside a 2 mm × 2 mm grid box. Mandibles were positioned dorsal side up and parallel to the bottom line of the grid box to establish a 90° vertical line for the right mandible angle of curvature measurement (Figure 2a). The soldier right mandible angle of curvature (smac) was measured from two positions: the dorsal condyle (smac1) and external curvature inflexion point (smac2) (Figure 2b).

Alates were mounted between a glass slide and cover slip in 100% ethanol [29]. Qualitative characters included body color (abc) and wing pigmentation (awp). Quantitative characters included body length (abl), body length including wings (ablw), average forewing length (afw) and average hind wing length (ahw) (Figure 3a–c).

Soldier and alate specimen were prepared under a binocular dissecting microscope (CIT-OVAL2, Carl Zeiss aus Jena, Jena, Germany and Leica WILD M10, Wetzlar, Germany). Images were taken with a Sony DKC-5000 camera attached to a Leica WILD M10 stereomicroscope (Wetzlar, Germany) using Adobe Photoshop v.8.0 (Adobe Systems, San Jose, CA, USA). All soldier and alate images were taken at 25× and 20× magnification respectively, and calibrated with a micrometer using the internal preset calibration setting in AutoMontage Pro, v.5.0.1 (Cambridge, UK). Morphometric measurements were recorded using the AutoMontage Pro, v.5.0.1 and exported to Microsoft Office Excel (Redmond, Washington, USA). All statistical analyses of mean, standard deviation and simulation of sample size were performed using SAS v.9.2 (SAS Institute Inc., Cary, NC, USA), one-way analyses of variance (ANOVA) performed on each character state to determine if it contributed significantly to species separation. Sequential t-tests with LSD (protected least square deviation) pair-wise comparison were used to determine species differences for all characters measured. Step-Wise Discriminant Analysis (SWDA) was used to determine which morphological characters were most useful in species separation. The reliability and accuracy of each character state was determined using multiple Discriminant Function Analyses (DFA) [29].

Fully developed, not-yet flown, winged alates of the new species were collected from infested wood on Sapelo Island on three separate dates.

A dichotomous key for the Reticulitermes species of Georgia was constructed using morphological and behavioral (flight phenology) data.

Sequence data for two mtDNA genes obtained from workers, soldiers and alates were employed to provide additional support for the new species. Genomic DNA was extracted from selected specimens using either Promega’s Wizard Genomic DNA Purification Kit or Qiagen’s DNeasy Extraction Kit, following a modified protocol [22]. Primers used for amplification of the entire length of the mitochondrial COII and partial COI genes are listed in Table 3. Amplified PCR products were sequenced at Molecular Cloning Laboratories (South San Francisco, CA, USA) or Eurofins MWG Operon (Huntsville, AL, USA).

Sequences were curated with Sequencher 4.5 (Gene Codes Corp., Ann Arbor, MI, USA) and aligned with MUSCLE (MEGA 5 [30], or Phylogeny.fr: Robust Phylogenetic Analysis For The Non-Specialist [31]) using the default settings. Gaps were coded as missing. An estimate of net evolutionary divergence was calculated between the five Southeastern Reticulitermes using MEGA 5 by determining the number of base substitutions per site. The sequence data were used to infer optimal phylogenetic trees employing the following tree estimation methods as implemented by the listed software package: Maximum Likelihood (PHYML) and Maximum Parsimony (MEGA 5) [26,27,28,29,30,31,32]. ML (Maximum Likelihood) analysis for COI and COII was performed with PHYML 3.0 on the Phylogeny.fr: Robust Phylogenetic Analysis For The Non-Specialist [31]web server using the GTR+G+I model. The MP (Maximum Parsimony) analysis of COI and COII sequence used MEGA 5 with Close-Neighbor-Interchange (CNI) search and 1000 bootstrap replicates. Graphical representations of the resulting trees were improved using FigTree [32].

Phylogenetic relationships of Reticulitermes from the southeastern USA were compared to GenBank sequences from other regions of the world to support the hypothesis that the new species was not previously described. Primary molecular voucher specimens were deposited at the University of Georgia Collection of Arthropods (UGCA), Georgia Museum of Natural History, Athens, Georgia, USA. DNA extraction vouchers were deposited in HSERP Laboratory, University of Georgia, Athens, GA, USA.

The new species had the smallest range of measurements for both alate and soldier samples for all morphological characters examined with the exception of soldier head capsule ratio (sl:sw), and soldier right mandible angles of curvature (smac1 and smac2) (Table 4a). The range of sl:sw was similar to R. virginicus, and smac1 and smac2 were similar to R. flavipes (Table 4a).

The range of soldier head capsule measurements for the new species was 1.14–1.72 mm for sl and 0.70–0.99 mm for sw (Table 4a). The sl:sw for soldier head capsules ranged from 1.52–1.98 (Table 4a). The range for the soldier right mandible angle of curvature was 7.2–14.6° for smac1 and 24.1–34.0° for smac2 (Table 4a). Alate body length without wing (abl) ranged from 3.26–4.63 mm, and alate body length including wing (ablw) ranged from 6.53–7.88 mm (Table 4b). The length of average forewings (afw) was slightly longer than the average hind wings (ahw), ranging from 4.94–5.98 mm, and 4.81–6.21 mm, respectively (Table 4b).

Alate samples were collected from Sapelo Island on 12 May 2005, 6 February 2007 and 6 March 2007. All samples of fully developed, sclerotized and winged, alates were collected directly from sampling devices prior to flight. We predict that flights of the new species would occur during the same time frame as the alates were fully sclerotized and winged.

Keys to the soldiers and alates of the Reticulitermes species of Georgia, USA, were constructed (Appendix 1, Appendix 2, respectively) based on values shown in Table 4a-Table 4b in shaded boxes representing the mean ± 1 std. dev. values for each character, with the exception of sl:sw. A minimum of 9 soldier head capsules are recommended to obtain a 95% confidence level for correct identification of R. flavipes, R. virginicus, and the new species, and 29 specimens for separating R. malletei from R. hageni. The minimum number of alate specimens recommended to obtain a 95% confidence level for discriminating all five species is 6. A more detailed discussion of the statistical analyses and generation of minimum sample size can be found in Lim (2011) [29].

A total of 102 COII and 49 COI sequences were analyzed providing 20 new haplotypes for COII and 21 new haplotypes for COI. The estimated net evolutionary divergence among the five Southeastern Reticulitermes is shown in Table 5a and Table 5b for COII and COI respectively. The values indicate that R. nelsonae evolutionary divergence is comparable to the net evolutionary divergence observed between other described species from that region (Table 5a and Table 5b). Similar clades were recovered in all phylogenies from the COII and COI sequence data (Figure 4, Figure 5, Figure 6 and Figure 7). There are five reference sequences (labeled as α in Figure S1, Figure S2, Figure S3 and Figure S4) obtained from specimens where both alates and soldiers from that same collection point and date matched the described morphological criteria for the respective species (Figure 4, Figure 5, Figure 6 and Figure 7). Two sequences in the R. malletei clade for COII and COI were labeled β (Figure S1, Figure S2, Figure S3 and Figure S4) to signify that these sequences were corroborated with the original 16SrRNA haplotype sequences in Austin et al. (2007).

The COII alignment included 665 bp of the 685 bp full length to size-match the length of some sequences retrieved from GenBank. The comparisons yielded 216 variable sites, of which 178 were parsimony-informative. Maximum likelihood (ML) and maximum parsimony (MP) were employed to construct phylogenetic trees that included selected GenBank sequences (as of 8 March 2011) for Reticulitermes. All inferred phylogenetic trees were well resolved for the Southeastern Reticulitermes showing consistent and strongly corroborated topologies with the exception of two haplotypes in the R. nelsonae clade (JF796236 and AF525328) from the MP phylogeny that are recovered as polytomies containing R. nelsonae, R. malletei, R. virginicus and R. hageni (Figure 4, Figure 5, Figure S1 and Figure S2). The inferred ML tree shows high branch support (>84) for all species clades within the Southeastern Reticulitermes group with Ln likelihood = −5075.935 (Figure 4). The MP analysis resulted in 42 most parsimonious trees (length = 854, CI = 0.33, RI = 0.82) (Figure 5). The composite index for all sites = 0.300, and a composite index for parsimony-informative sites = 0.270 that used 1000 bootstraps to generate the consensus tree (Figure 5). COII sequences were deposited in GenBank (EU689013, JF796229-JF796233, and JF796235-JF796236) and are listed in bold in the supplemental phylogenies (Figure S1 and Figure S2). The ML and MP phylogeny for COII showed separation between Reticulitermes from the southeastern USA and Reticulitermes from other regions (Figure 4 and Figure 5).

The alignment of COI included 767 bp of the 801 bp partial length to size-match with some of the GenBank sequences retrieved for this analysis, Results showed 146 variable sites of which 113 were parsimony-informative. Inferred ML tree shows high branch support for most clades with Ln likelihood = −2905.620 (Figure 6). MP analysis resulted in 33 most parsimonious trees of 366 steps with a consistency index = 0.556, a retention index = 0.892, a composite index for all = 0.565, and a composite index for parsimony-informative sites = 0.496 that used 1000 bootstraps to generate the consensus tree that was similar in tree topology to the ML analysis (Figure 6 and Figure 7). COI sequences also were deposited in GenBank (JN207486-JN207491) and are listed in bold in the supplemental phylogenies (Figure S3 and Figure S4).

The phylogenetic trees from ML and MP analyses using the COI data also recovered genetic separation between Reticulitermes from the southeastern region of the USA and Reticulitermes from other parts of the world (Figure 6 and Figure 7). No COI sequences were found in GenBank for Reticulitermes from the western USA.

Soldier (Table 4a, Figure 8b and 8c): Head capsule small (mean sl 1.407 mm, mean sw 0.054 mm, mean sl:sw 1.793 mm) (Table 4a).

Alate (Table 4b, Figure 8a, Figure 9 and Figure 10): Body length, with (ablw) and without (abl) wing, small (mean abl 3.93 mm, mean ablw 7.080 mm), body color pale brown and wings non-pigmented (Figure 8a, Figure 9 and Figure 10).

There is little sexual dimorphism for soldier or alates, males and females can however, be differentiated by the form of the 8th sternal plate (Zimet & Stuart, 1982).

Soldier Head capsule rectangular in shape, typically longer than wide (Figure 8b). Head capsule color yellowish with dark brown to black mandibles (Figure 8b). Body color light yellowish to white. Mean head capsule length (sl) 1.41 mm ± 0.13 (Table 4a), mean width (sw) 0.78 mm ± 0.05 (Table 4a), mean ratio length-to-width (sl : sw) 1.793 ± 0.09 (Table 4a). Mean mandible curvature angles: smac1 = 10.7° ± 2.21, smac2 = 27.27° ± 2.65 (Table 4a).

Alate Body color pale brown with 14 antennal segments (Figure 8a and Figure 10). Wings non-pigmented (Figure 8a and Figure 10). Legs light to dark brownish with 14 antennal segments (Figure 8a and Figure 10). Mean body length without wing (abl) 3.93 mm ± 0.24 (Table 4b). Mean body length with wings (ablw) 7.08 mm ± 0.29 (Table 4b). Mean forewing length (afw) 5.43 mm ± 0.21. Mean hind wing length (ahw) 5.32 mm ± 0.30 (Table 4b).

This patronym was established to honor Lori J. Nelson (USDA Forest Service, Buchanan, CA, USA) who realized in 1996 that specimens collected on Sapelo Island, a barrier island off the Atlantic coast of Georgia, were notably different from all previously described Reticulitermes species based on analysis of cuticular hydrocarbons [17,18,19].

Behavior

Reticulitermes nelsonae is expected to swarm from February to May. Reticulitermes hageni swarms from August to October. Reticulitermes flavipes flights have been recorded from November through April thus overlapping with R. nelsonae [29]. Reticulitermes nelsonae flight times also overlap with R. virginicus and R. malletei, both of which have been recorded in May [10,13,33].

Reticulitermes nelsonae is found in the southeastern region of the United States, in the Atlantic Coastal Flatwoods and South Coastal Plain soil provinces (Table 2, Figure 11). The species has not been detected in the Piedmont soil province despite sampling in that region. In addition to the type locality on Sapelo Island, GA, this species has been collected in Croatan National Forest in Havelock, NC, Greenwood Plantation in Thomasville, GA, and Branford, FL (Table 2, Figure 11).

Soldier Reticulitermes nelsonae head capsule length (sl) is at least 0.2 mm shorter than R. flavipes and R. virginicus (Table 4a). Reticulitermes nelsonae head capsule width (sw ) at least 0.1 mm smaller than R. flavipes and R. virginicus (Table 4a). The sl and sw are also more than 0.1 mm smaller than R. malletei (Table 4a). The R. nelsonae smac2 is more than 24° (typically 24°–30°), while R. hageni smac2 is less than 25° (typically 22°–25°) (Table 4a).

Alate Reticulitermes nelsonae abl and ablw are typically 3.7 mm–4.2 mm and 6.8 mm–7.4 mm, whereas in R. flavipes those same characters are more than 4.4 mm and 8.6 mm, respectively (Table 4b). Reticulitermes nelsonae afw and ahw are 1.0 mm shorter than R. flavipes (Table 4b). Body color differs from that of R. hageni’s yellowish to yellowish-brown body color (Figure 10). Reticulitermes nelsonae wings are non-pigmented, while R. malletei has pigmented wings (Austin et al., 2007) and forewing (afw) measurements are 0.4 mm shorter than R. malletei. Reticulitermes nelsonae and R. virginicus share similar morphometric ranges, but differ in body color, with R. nelsonae having light brown color and R. virginicus dark brown. The ratio of mean body length including wings to mean forewing length (ablw: afw) is typically 1.27–1.31 for R. nelsonae, while for R. virginicus it is 1.32–1.37 (Table 4b).

Sequences from cytochrome oxidase I (COI) and cytochrome oxidase II (COII) genes generated data that showed R. nelsonae was genetically unique as its haplotypes consistently formed a separate clade from the other described species (Figure 4, Figure 5, Figure 6 and Figure 7).

An examination of Haverty et al. (1999) [18] and Jenkins et al. (2000) [19] indicates that the hydrocarbon phenotypes GA-L and GA-I most likely correspond to R. nelsonae. Jenkins et al. (2000) [19] observed that two of their collections were “different morphologically, chemically and genetically”. We now believe these samples, identified as haplotypes BH25 (JF796235) and HH11 (JF796236), are R. nelsonae because we recovered those haplotypes in the R. nelsonae clade for the COII gene (haplotypes are underlined in Figure S1 and Figure S2). We, therefore, reason that the cuticular hydrocarbon phenotypes GA-L and GA-I reported by [17,18] belong to R. nelsonae.

Morphological separation of Reticulitermes species is notoriously difficult. Our morphometric measurements provided a range of overlap consistent with past reports for the genus [13,14,16,17,21,34]. Dichotomous keys for soldiers and alates of Reticulitermes species collected in Georgia (Appendixes 1 and 2) were prepared to distinguish the five species endemic to the southeastern USA. The measurements used to build the soldier key can separate R. nelsonae soldiers from all previously described species with the exception of R. hageni, which overlap on all measures at the upper range for R. nelsonae (Table 4a and Appendix 1). Alates of all species can be separated based on the combination of body color, morphometric measurements, and flight times (Table 4b and Appendix 2).

The first author has prepared a more extensive study of the literature on morphological variation in Reticulitermes species from the southeastern USA [29]. We recommend that 6 alates and/or 29 soldiers specimens be used to achieve a 95% confidence in morphometric-based species diagnosis [29].

Congruent and similar phylogenies were obtained from ML and MP analyses for both COII and COI sequences (Figure 4, Figure 5, Figure 6 and Figure 7). Genetics of Reticulitermes from the southeastern USA were consistently differentiated from that of Reticulitermes from other regions of the world and haplotypes of Reticulitermes nelsonae were further differentiated within the ‘Southeastern Reticulitermes’ grouping and thus are genetically unique from previously described Reticulitermes (Figure 4, Figure 5, Figure 6 and Figure 7). The net evolutionary distance dataset demonstrates that the ‘genetic uniqueness’ of R. nelsonae is within the range of the already described species (Table 5a and Table 5b).

Three Reticulitermes species (R. virginicus, R. hageni and R. malletei) from the southeastern USA have clear separation between COII haplotype designations within their respective species clades, while mixed haplotype designations were observed in the R. flavipes and R. nelsonae clade (see supplementary Figure S1 and Figure S2). The R. flavipes clade includes R. santonensis (which has been synonymized as R. flavipes) and R. arenincola haplotypes (Figure S1 and Figure S2). Previous reports on R. arenincola support our findings showing R. arenincola is genetically similar to R. flavipes [35,36,37].The R. nelsonae clade includes four GenBank sequences that were designated as R. hageni (NC009501, AY808088, AY808089, and AF525328) (Figure S1 and Figure S2). Voucher specimens for these GenBank accessions have been requested and to date two examined (AY808088, AY808089). The mistaken GenBank accessions are most likely the result of the fact that only 3 of the 5 species endemic to the southeastern USA are listed in published keys and the available taxonomic keys would have identified R. nelsonae specimens as R. hageni [13,14,15].

The 767 bp COI gene phylogenies also indicate that haplotypes of Reticulitermes species from the southeastern USA are different from Reticulitermes found in other areas of the USA (Figure 4, Figure 5, Figure 6 and Figure 7). Reticulitermes santonensis was again recovered within the R. flavipes clade further supporting junior synonym status (Figure S3 and Figure S4). One GenBank R. hageni sequence (EF 206320) was recovered within the R. nelsonae clade (Figure S3 and Figure S4) which is not surprising, and as mentioned, the specimen would have been identified as R. hageni based on published keys [13,14,15].

Molecular phylogenies are an estimation of plausible species relationships and therefore detailed research and comparison is warranted to accurately identify species designations for gene sequence data. Cytochrome oxidase II (COII), with a length of 685 bp, has been a valuable marker for identification of Reticulitermes species [19,22,38,39] as evidenced by Sillam-Dusses and Forschler (2010) who reported an undescribed species based on genetics alone [22]. We recovered R. arenincola Goellner 1931 in the R. flavipes clade lending support for synonymy with R. flavipes (Figure 4 and Figure 5) [12,35,36,37,40].

We suggest that future genetic analysis provide reference sequences for COII and COI genes that are from specimens corroborated with morphological descriptions for Reticulitermes species (as denoted by α in Figure S1, Figure S2, Figure S3 and Figure S4).