Seeking quantitative morphological characters for species identification in soldiers of Puerto Rican Heterotermes (Dictyoptera, Blattaria, Termitoidae, Rhinotermitidae)

Abstract Subterranean termites in the genus Heterotermes Froggatt (Rhinotermitidae: Heterotermitinae) are pantropical wood feeders capable of causing significant structural damage. The aim of this study was to investigate soldier morphological attributes in three Puerto Rican species of Heterotermes previously identified by sequencing of two mitochondrial genes and attributed to Heterotermes tenuis (Hagen), H. convexinotatus (Snyder) and H. cardini (Snyder). Soldiers (n = 156) were imaged and measured using the Auto-Montage image-stacking program. We demonstrated that Puerto Rican Heterotermes soldiers could not be identified to species level based upon seven morphometric indices or any combination thereof. Nor could differences in soldier head pilosity be used to discriminate species, in contrast to previous findings. However, previously described characters of the soldier tergal setae were reported to be useful in discriminating H. tenuis from both of its Puerto Rican congeners.


Introduction
Heterotermes Froggatt, 1897 (Rhinotermitidae: Heterotermitinae) is a pantropical genus of subterranean wood feeding termites (Constantino 2000). Seventeen species have been reported as pests that damage human structures (Scheffrahn and Su 2000). The Heterotermes fauna in the Caribbean Region (the Bahamas, Greater Antilles, and Lesser Antilles) is thought to consist exclusively of pest species that have been introduced from the South American mainland (Constantino 1998, Evans et al. 2013. Caribbean Heterotermes are consequently of interest in that they are both invasive and economically significant. In the Caribbean Region, the Puerto Rican archipelago is situated at the eastern end of the Greater Antilles and is adjacent to the younger, actively volcanic Lesser Antilles, hence providing a biotic link to the South American mainland. Our understanding of Heterotermes species' identity and distribution in the Caribbean Region has fluctuated over time, with the species composition of the Heterotermes fauna in Puerto Rico and its associated islands being quite ambiguous. Snyder (1956) reported Heterotermes tenuis (Hagen, 1858) and Heterotermes convexinotatus (Snyder, 1924) from the archipelago, while Scheffrahn et al. (2003) left Puerto Rican Heterotermes specimens unidentified to species level due to taxonomic uncertainty. Using a phylogenomic approach conjunct with some morphological and biogeographical data, Szalanski et al. (2004) concluded that Heterotermes cardini (Snyder, 1924) was present in the Caribbean Region in addition to H. tenuis and H. convexinotatus, along with one or more undescribed species. Three samples from Puerto Rico were included in their study, all identified as H. convexinotatus. In contrast, Eaton et al. (2016) reported H. tenuis, H. convexinotatus, and H. cardini in Puerto Rico based on molecular phylogenies of two mitochondrial loci from 76 Puerto Rican samples, with morphological confirmation of species identification. Furthermore, their phylogenomic data provided strong evidence that the proposed undescribed Heterotermes (Szalanski et al. 2004) were consistent with H. cardini.
Heterotermes soldiers from the Caribbean remain difficult to reliably distinguish due to non-robust diagnostic morphological characters: Snyder (1924) asserted that soldiers of the three species differed in cephalic and pronotal pilosity, body coloration, and relative size. However, due to morphological ambiguity, Snyder (1924) suggested in his original descriptions of H. convexinotatus and H. cardini (the latter described from the Bahamas) that they might be synonymous with H. tenuis. Consequently, alates (winged reproductives) are essential for reliable species identification of those Heterotermes spp. putatively present in Puerto Rico (Snyder 1924;Szalanski et al. 2004;Eaton et al. 2016). Since alates are produced only seasonally, they are difficult to obtain and are seldom properly associated with their parent colonies. Therefore, robust soldier-based identification of Puerto Rican Heterotermes is of practical taxonomic interest.
The purpose of our study was to measure morphometric parameters in the soldier caste from a comprehensive sample of the Puerto Rican Heterotermes fauna, and to determine what parameters, if any, were most useful in identifying Heterotermes to species level. Additionally, pilosity of the head capsule (Snyder 1924) and abdominal tergites of Heterotermes soldiers (Constantino 2000) were examined as diagnostic characters. The possibility of an additional, undescribed species of Heterotermes in the Caribbean Region (as per Szalanski et al. 2004) was also herein analyzed.

Materials and methods
Our study is supplementary to that of Eaton et al. (2016) in that it largely uses a subset of the same Puerto Rican Heterotermes samples (n = 60 of 76 samples). These were assigned to H. tenuis, H. convexinotatus, or H. cardini on the basis of their 16S rRNA and cytochrome oxidase subunit II (COII) phylogeny. Samples were collected by SCJ from different locales on the main island of Puerto Rico and adjacent Culebra Island in 2002, 2004, 2006, or 2010. Each sample consisted of termites collected from a single access point in a given colony and placed in individual vials filled with absolute alcohol.
In total, 156 individual soldiers were examined morphometrically (see Suppl. material 1), with 3 being selected, when available, from each sample, providing 40 specimens of H. tenuis, 55 H. convexinotatus, and 61 H. cardini. We investigated mandible length in combination with the same axis of the head capsule, along with head width (Figure 1), in order to provide additional commonality with the data presented in Constantino (2000) and Szalanski et al. (2004). In addition, multiple pronotal metrics ( Figure 2) were examined, two of them novel to this study (see Suppl. material 1). A total of three morphometric indices recommended by Roonwal (1969) for use in termite taxonomy were derived from a subset of these metrics (see Suppl. material 1). Cephalic setae were surveyed in 79 of the 156 soldiers, plus an additional 7 H. convexinotatus soldiers examined without morphometric analysis. Also, setae within a 300-μm radius of the soldier fontanelle were censused in 66 of the 156 soldiers. A subset of 45 of the 156 soldiers was used for characterization of tergal setae.
A Z16 AP0 stereomicroscope (Leica Microsystems, Buffalo Grove, USA) with a KY-570B camera (JVC, Wayne, USA) was used to image specimens for morphometric analysis and examination of tergal setae, with images being stacked into montage microphotographs using Auto-Montage Pro software (ver. 5.01.0005, Synoptics Ltd., Cambridge, UK).
We performed analyses of variance (ANOVA), discriminant analyses, and k-means cluster analyses on morphometric data using SPSS Statistics 24 (2016, International Business Machines, Armonk, USA). The level of significance for all analyses was set at α=0.05. Based on the results of ANOVA, a discriminant analysis and three k-means cluster analyses were performed using all statistically significant parameters. Discriminant analyses determine the efficacy of any array of variables in assigning group membership (Green et al. 2008).
K-means cluster analyses do not assume a priori assignments of specimens to groups, instead attempting to iteratively delineate groups de novo from the data given according to the number of groups provided by an a priori hypothesis. Our k-means cluster analyses specified 3, 4, and 2 morphometric-delineated groups within the sampled Heterotermes fauna. The first analysis was premised according to the conclusion of Eaton et al. (2016) that there are 3 valid Heterotermes spp. in the Puerto Rican archipelago. The second was premised according to the conclusion of Szalanski et al. (2004) that 4 Heterotermes spp. occur in the Caribbean Region. The third was premised according to the phylogenetic hypothesis that H. tenuis is sister to a clade including H. convexinotatus and H. cardini (Szalanski et al. 2004;Eaton et al. 2016): if this were  the case, one would expect specimens to group into two clusters. It was not possible to delineate one or more metrics on a total of 40 specimens (5 H. tenuis, 8 H. convexinotatus, and 27 H. cardini), so these were excluded from k-means cluster analysis.

Results and discussion
ANOVA concerning all parameters demonstrated that all but one of the metrics surveyed -the depth of the posterior pronotal notch, which was novel to this studydisplayed statistically significant variation across all three species (Table 1). Only one of the three morphometric indices we derived -the head-mandible index (Roonwal 1969) -exhibited statistically significant variation across all three species (Table 1).
Discriminant and k-means cluster analyses used only statistically significant parameters. Of all 116 specimens with sufficient morphometric data for their inclusion in discriminant analysis, our discriminant analyses correctly classified 87.1% to species. Thus, while the majority of Puerto Rican Heterotermes with sufficient morphometric data could be accurately identified to species level using statistically significant morphometric parameters, these data were not invariably reliable for that purpose; nor was this mode of identification concise given that a total of seven parameters was necessary.
The results of morphometric 3-and 4-cluster analyses did not consistently conform to the phylogenetic hypotheses of Eaton et al. (2016) and Szalanski et al. (2004) for Caribbean Heterotermes, respectively, nor did any cluster analysis support placement of H. tenuis as a sister-group to the remaining two putative species of Heterotermes. Furthermore, no morphometric consensus, novel or otherwise, could be made across all three k-means cluster analyses (Table 2). These results were likely influenced by the fact that many specimens (n = 40) provided insufficient data to be included in cluster analyses.
We found short to medium-length (uniformly <100 μm) setae on the head capsules of all H. convexinotatus specimens examined with respect to this character, but pilosity of this species and H. tenuis was comparable (Table 3). Furthermore, there was considerable overlap in head capsule pilosity among all three species (Table 4). Likewise, the area within a 300-μm radius of the fontanelle of a subset of the 156 specimens displayed considerable overlap in seta quantity and no apparent pattern among all three species (Table 5). We therefore conclude that soldier head pilosity is not a reliable character for discriminating among Puerto Rican Heterotermes species, contrary to Snyder (1924).

Conclusions
Our findings demonstrated that examined morphometric characters of the soldier pronotum and head could not reliably differentiate Heterotermes species in the Puerto Rican archipelago. These data did not provide any well-resolved distinction between the three putative species in question, such as is supported by rigorous phylogenomic investigation (Eaton et al. 2016). We also found that H. convexinotatus and H. cardini soldiers could not be unequivocally discriminated by cephalic seta counts, contrary to Snyder (1924). However, H. tenuis soldiers could be readily identified by means of tergal setal distribution, as reported by Constantino (2000). We furthermore report that soldier tergal seta distribution is useful to distinguish H. tenuis from H. cardini. and by Research Scholar Award funds disbursed by the Undergraduate Research Office of The Ohio State University. The authors would like to thank Abhijoy Saha for his statistical consulting services; Dr David Shetlar for his assistance proofreading application materials for the Research Scholar Award; Huayan Chen and Dr Hans Klompen for their assistance in the imaging process; and Alex Tyrpak, Kara Baker and Nina Bogart for their past and present aid and support at the Jones Laboratory.