Phylogenetic reconstruction of Syntermitinae (Isoptera, Termitidae) based on morphological and molecular data

The subfamily Syntermitinae comprises a group of Neotropical termites with 18 genera and 101 species described. It has been considered a natural group, but relationships among the genera within the subfamily remain uncertain, and some genera appear to be non-monophyletic. Here, we provide a comprehensive phylogeny including six Neotropical species of Termitinae as outgroup, 42 Syntermitinae species as ingroup, 92 morphological characters (from external and internal anatomy of soldier and worker castes) and 117 molecular sequences (109 obtained for this study and 8 from GenBank) of 4 gene regions (41 and 22 from Cytochrome Oxidase I and II respectively, 19 from Cytochrome b, and 35 from 16S rDNA). Morphological and molecular data were analyzed in combination, with the Bayesian inference method, and the important aspects of termite biology, defense and feeding habits are discussed based on the resulting tree. Although useful for providing diagnostic characters, the morphology of the soldier caste reveals several cases of convergence; whereas the feeding habit shows indications of evolutionary significance.


Introduction
The subfamily Syntermitinae comprises a group of Neotropical termites that ranges from southern Mexico (Cahuallitermes) to northern Argentina (Cornitermes, Procornitermes, Rhynchotermes, Syntermes), with the richest generic and specific diversity in the Brazilian Cerrado biome. Fifteen syntermitine genera occur in the Cerrado, where several species of Cornitermes, Silvestritermes and Syntermes construct conspicuous epigeal nests that characterize this savanna-like landscape. Cornitermes cumulans can reach a nest density of 55/ha, and is considered a keystone species in the Cerrado [1]. These termite nests may harbor many other termite species as well as other groups of invertebrates.
The feeding and nesting habits of syntermitine species are diverse. The group includes grass/litter-feeders, intermediate feeders, and humus-feeders. The nests are variable; some species build earthen nests; most are commonly epigeal, but arboreal and subterranean forms are a1111111111 a1111111111 a1111111111 a1111111111 a1111111111

Material and methods Taxon sampling and outgroup selection
We included a total of 42 syntermitine species as ingroup, representing the diversity of the 18 currently described syntermitine genera; and 6 species of Termitinae as outgroup, chosen for their established relationships to Syntermitinae [21][22][23] and also based on our experience with Neotropical termites. Morphological studies were carried out on termite specimens deposited in the Isoptera collection of the Museu de Zoologia da Universidade de São Paulo, São Paulo, Brazil (MZUSP). A representative sample of each lot used to perform the molecular studies was formally deposited in the MZUSP as well and appropriately registered for public consult.

Morphological characters
We included a total of 92 characters, 40 of the soldier external morphology, 42 of the coiling gut in situ and the configuration of the different parts of the worker digestive tube, and 10 of worker external morphology. The morphological character data are expanded from our previous study [3]; most characters are referenced in Figs 1-17. (1) lanceolate ( Fig 1C); (2) obtuse ( Fig 1D and 1F).
Characters based on external morphology of workers. 83. Size proportion of head to thorax: (0) head much larger than thorax ( Fig 16A); (1) head size similar to thorax (Fig 16B  and 16C).

Molecular protocols
We chose four regions of the mitochondrial genome, Cytochrome Oxidase I and II (COI~600 bp, COII~660 bp), Cytochrome b (Cyt B~340 bp) and 16S rDNA (~430 bp). The DNA was extracted preferentially from the head and thorax of a single soldier individual preserved in 95% ethanol (Table 1), with the set of reagents from the DNeasy Blood & Tissue Kits (Qiagen), supplemented with 20 mg/ml proteinase K, following the manufacturer's protocol. The homogenates were incubated at 55˚C for 3 h. The gene fragments were then amplified by polymerase chain reaction, PCR [26]. The primers and the amplification conditions are listed in Table 2. PCR was performed in 25 μL reactions (12.5 μL PCR master mix Promega1, 0.6 μM of each primer, 3.0 μL of total DNA, and 3.5 μL deionized water). The amplified PCR products were determined by gel electrophoresis on a 1% agarose gel diluted in TAE buffer (1X) (0.04 M Tris base, 0.02 M acetic acid, and 1 mM EDTA). This same buffer was also used in 1-h electrophoresis runs in an 8-V/cm length gel. All reaction products were purified with Wizard1 SV Gel and PCR Clean-Up System (Promega), following the manufacturer's protocols. Purified PCR products were sequenced with the same primers used in the original PCR reactions and the BigDye1 Terminator v3.1 Cycle Sequencing Kit, under the same conditions of PCR. The sense and antisense sequences obtained from each amplicon were assembled, and a consensus sequence for each gene was generated with Geneious v.8.1.7 analysis tools [27]. The nucleotide sequences reported here were submitted to the GenBank database under the accession numbers indicated in Table 1.

Analyses
The saturation of the molecular data was assessed with DAMBE v.6.0.48 [33] using the test of substitution saturation by [34,35]. The saturation test showed little saturation, indicating that the data were suitable for phylogenetic analysis (Iss < ISSc; p < 0.05).
To evaluate the most useful data set, we made several tests combining different sets of sequences (the four gene sequences, COII + 16S rDNA + COI, COII + 16S rDNA + Cytb and only COII + 16S rDNA), with and without the morphological data, and with the protein-coding sequences partitioned either by genes or by the codon position. The results are summarized in the Table 3.
The models for each DNA data partition were determined using the JModel Test 2 [36] and PartitionFinder v1.1.0 [37], for the morphological data partition the states were unordered. The Bayesian inference analyses were performed with MrBayes version 3.2.1 [38], in the CIPRES Science Gateway V. 3.3 [39]; in all analyses, four chains were run for 50 million generations and sampled every thousand generations (two runs). In all cases the burn-in limitation was determined by visual inspection of the trace-plot and evaluation of the effective sample size value (ESS) of the combined runs, using Tracer v1.6 [40]. The burn-in of 1% was sufficient.
Ancestral character states were reconstructed with the help of Mesquite v.3.04 [24], by the parsimony criterion.

Results and discussion
From the total of 48 species used in our analyses, we obtained DNA data from three or four different gene for 25 species, two different sequences for 16 species and only one sequence for 4 species (Table 1). Two taxa are represented only by morphological data. About one third of the sequences information is absent. Although the poverty of sequences may compromise the results, the majority of taxa share COII and 16s rDNA information (The information for COII sequences is absent only in four taxa and for 16S rDNA in 10), the major part of lacking information is concentrated in COI and Cytb sequences.
Among the trees obtained, three of them present informative topology (with few polytomies) and high posterior probabilities (especially in the basal nodes); 1)The result of an analysis with Morphology + COII + 16S rDNA, partitioned by genes (COII: GTR +I + G, 16S rDNA: GTR +G) represented in Fig 18 and Fig 20A (29 nodes more than 0.9, 2 nodes equal  Their respective traceplots are illustrated in S1-S3 Figs. Considering all obtained trees (Figs 18-20 and S4-S16 Figs), the exclusion of morphological data from the analysis result in large pectinate nodes (S5, S6, S9, S10, S12, S13, S15 and S16 Figs). The position of Acangaobitermes krishnai, Cahuallitermes intermedius, Macuxitermes triceratops and Noirotitermes noiroti represented by only one sequence and Armitermes armiger and Ibitermes tellustris, with no sequences, remains stable in the three more consistent trees (Fig  21), and considering just morphological characters, mainly from internal morphology, the position among them seems reliable.
Comparing the three most robust results (Fig 21A-21C), their topologies have a few divergences; the groups of genera (delimited by the best-supported basal nodes, and indicated by the same colors in each tree) were recovered with identical taxa compositions.
The more notable differences among the selected trees are: The position of the genus Syntermes (the red branch, with taxa initiated by the acronym "SY"), not resolved in tree A and positioned in trees B and C as a sister group of the yellow branch, composed of Rhynchotermes (indicated by the acronym "RH"), Procornitermes (PR), Cahuallitermes (CA) and Cornitermes (CO). The relationships among the taxa are indicated by dark-green branches; Genuotermes (GESP), Curvitermes (CUOD), Embiratermes (EM), Paracurvitermes (PAMA), Cyrilliotermes (CYAN) and Silvestritermes (SI); possible paraphyletic in trees A and B, and recovered as a monophyletic group in tree C.
Although these results do not contradict each other, for prudence we opted to reconstruct and discuss the ancestral character states using tree A, which is less resolved but more conservative.

Taxonomic discussion
The sample of species is sufficiently comprehensive to allow a discussion that previous studies did not attempt. Four genera appeared as paraphyletic in our analysis: Armitermes (possibly), Procornitermes (possibly), Embiratermes, and Ibitermes.
The cases of Armitermes (pink branch, Fig 21) and Procornitermes (part of the yellow branch, Fig 21) can be resolved with reallocation of a few species: for Armitermes the most conservative solution is include all taxons (Armitermes, Macuxitermes, Uncitermes and  With morphological data Protein-coding S11 Fig  S8 Fig  S14 Fig  Fig 18   sequences  Embiratermes heterotypus) into a single genus, since the node that it is grouping the five taxa has a high posterior probability. Nevertheless, considering we only obtain one sequences for Macuxitermes, and also the support for the three more internal nodes are low, we think more studies are necessary to infer consistently the relationship among the taxa before introducing nomenclatural changes; for Procornitermes, resurrecting it from Triacitermes Emerson, now including Procornitermes triacifer and P. araujoi is a admissible solution, however the paraphyly of the genus is not a consensus among the results. More detailed studies for these cases are necessary before formal proposals for nomenclatural changes can be made. A revisionary work is necessary in the next future to reassess generic and specific limits as well as the intergeneric relationships of Embiratermes and Ibitermes within other members of the subfamily, since their named species used herein as terminals are spread all over the tree.
A surprising result is the position of Genuotermes, deeply inserted in our tree. Although the association with Syntermitinae is unintuitive, some unique characteristics are shared with Syntermitinae. The soldier frontal-gland aperture is at the tip of a large projection located in the frontal region of the head; the soldier mandibles have a clearly recognizable molar plate and prominence, as in Silvestritermes [3], Cyrilliotermes [5], and Curvitermes [6]; and the worker gut morphology is very similar, including the characteristic dilated P1 of Syntermitinae [41]. Considering these points, the reallocation of the genus to Syntermitinae is expected, following comprehensive studies of other Neotropical termitine genera.

Defense and feeding behavior in Syntermitinae
Two aspects stand out in termite research: defense and feeding habits. The first aspect relates to the soldier caste in Isoptera, which comprises a very particular case for evolutionary biology. Soldiers are a "burden" on the colony maintenance, since they need to be fed by the workers, and the effective contribution of a very specialized caste for the colony defense is not clear. The proportion of soldiers and workers varies widely among species [42] and nearly 10% of termite species do not have soldiers (mainly Apicotermitinae). The second aspect relates to the central role of termites as decomposers in tropical climates; they can comprise as much as 95% of the soil insect biomass [43]. Termites can obtain nourishment from a variety of plant biomass sources, including wood, rotting wood, grass, cultures of fungi, lichen and humus; and this diversification of feeding habits appears to be linked to termite species diversification [21].
For the reconstruction of the defense behavior, each taxon was classified according to the categories of primary individual defense mechanisms summarized in [44]. Three categories of defense were recognized: "Biting/Crushing" (example in Fig 3A) "Piercing" (examples in Fig  3B and 3F), and "Slashing" (examples in Fig 3C, 3D and 3E). Orthognathotermes sp. is formally classified as "Slashing/Snapping", but this is not relevant to the present discussion; the result is represented in Fig 22. The reconstruction showed that equivalent categories of defense evolved independently several times in syntermitine history: "Slashing" mandibles appeared two or three times independently (Fig 22, black branches), "Piercing" (Fig 22, blue branches) two or three times, and "Biting/Crushing" (Fig 22, white branches) five or six times. Fig 22 shows two cases that well represent the degree of convergence of form. The soldier head of Rhynchotermes nasutissimus (Fig 22B) is clearly similar to Uncitermes teevani ( Fig  22D), although U. teevani is more kindred to Labiotermes labralis (Fig 22C), which itself shares several traits with a distantly related species, Syntermes molestus ( Fig 22A). These are not the only cases of convergence; the species of Embiratermes share a variety of external traits and are spread among four clades. It is unnecessary to exhaustively discuss all the cases, which would excessively lengthen this article. The soldier external morphology of each genus can be found in revisionary studies, or in the identification keys of Constantino [45,46].
Other cases of convergence in termite soldier morphology were discussed by Inward and collaborators [21], who found that the asymmetrical snapping mandibles, a very specialized type of termite defense, evolved independently four times among all Isoptera. In the present case, we found a high degree of convergence in the soldier types of defense inside a much more restricted group.
The means of establishing the diet of each termite species can be controversial. Some specialists have proposed using analyses of the gut contents [47] or nitrogen stable-isotope ratios [48], but no discrete criteria have been developed to classify the termite diet precisely. Despite this, the resources consumed by termites can be organized in a continuous humification gradient, from wood and grass, which are non-humified resources, at one extreme; and very humified resources, such as humus and stercoral material from other nests, at the other [49]. This Phylogenetic reconstruction of Syntermitinae gradient can be correlated and recognized in the worker mandible morphology [50,51]. Species that feed on non-humified resources have the molar region with conspicuous ridges and a relatively small apical tooth, which is termed "xylophagous morphology" (Fig 17A and 17B, for example). Species that feed on humified resources have the molar region without ridges and a prominent apical tooth, termed "intermediate/geophagous morphology" (Fig 17C-17E, for example).
The reconstruction of these two characters (Fig 23), relative size of the left apical tooth (85) and the molar region (92), showed the expected overlap between these characteristics; xylophagous traits are traced in yellow and geophagous in black. Our topology indicated that in the syntermitine evolutionary history, a very early split occurred between lineages that tend to feed on non-humified resources and species that tend to feed on very humified resources. The change in the species' diet was reflected in more than the mandible shape, and a complex change in the digestive apparatus and the associated symbionts would be expected; however, knowledge of Termitidae digestive processes and their correlation with the gut morphology is presently limited. We expect that the Syntermitinae will provide a useful and more practical case for future studies.
Unfortunately, the lack of syntermitine fossils limits the dating and evolutionary interpretations of these characteristics. The oldest record is an ichnofossil, described as a Syntermes-like nest [52], from southern Argentina and dating from the late Early Miocene; all other syntermitine fossil records in the literature are much more recent [53,54].