Molecular Phylogenetics and Evolution 98 (2016) 201–209
Contents lists available at ScienceDirect
Molecular Phylogenetics and Evolution
journal homepage: www.elsevier.com/locate/ympev
A global molecular phylogeny and timescale of evolution for Cryptocercus woodroaches q ⇑ Yanli Che a,c, Dong Wang b, Yan Shi a, Xiaohong Du a, Yongquan Zhao a, Nathan Lo c, Zongqing Wang a, a College of Plant Protection, Southwest University, Beibei, Chongqing 400716, PR China b Chongqing International Travel Healthcare Center, Jiangbei, Chongqing 400020, PR China c School of Life and Environment Sciences, University of Sydney, Sydney, NSW, Australia article info abstract
Article history: Cryptocercus is a genus of sub-social wood-feeding cockroaches that represents the sister group to the Received 9 June 2015 eusocial termites. We generated mitochondrial (12S + 16S rRNA, COII), nuclear (28S rRNA) and Revised 27 January 2016 Blattbacterium endosymbiont (16S + 23S rRNA) sequence data for 8 new Chinese species, and combined Accepted 7 February 2016 these with previously available data to undertake the most extensive analysis of phylogenetic relation- Available online 11 February 2016 ships within the genus to date. As expected, phylogenetic relationships among Blattabacterium strains were found to be congruent with those of their hosts. Three major clades were found to exist in Asian Keywords: populations, one representing taxa from the Hengduan mountains in Southwestern China, a second Wood burrower including taxa from Russia, Korea, Northeastern China, and Yunnan in the Hengduan Mountains, and a Molecular phylogeny Biogeography third including taxa from the Qinling Mountains and Daba Mountains in Central China. A molecular dat- Molecular clock ing analysis using 7 termite fossils to calibrate the molecular clock indicated that the divergence of American and Asian Cryptocercus occurred 55.09 Ma (41.55–72.28 Ma 95% CI), and that the radiations of American and Asian taxa occurred 28.48 Ma (20.83–37.95 Ma 95% CI) and 20.97 Ma (15.78– 27.21 Ma 95% CI) respectively. Reconstruction of ancestral geographic distributions using S-DIVA sug- gested Cryptocercus was originally distributed across both continents, as opposed to ancestral migration of Cryptocercus from one continent to the other. The last common ancestor of Asian Cryptocercus was inferred to have existed in Central China. An examination of male chromosome numbers in Asian Cryptocercus showed that diploid numbers vary from 2n =15to2n = 41, and indicates the presence of eight new species. Our study represents the most comprehensive phylogenetic and biogeographic study yet performed for this important group of cockroaches. Ó 2016 The Authors. Published by Elsevier Inc. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
1. Introduction Klass et al., 2008). Similar to the case for most other cockroaches and the termite Mastotermes darwiniensis, Cryptocercus species har- The wingless and wood-feeding cockroach genus Cryptocercus bor the intracellular bacterial symbiont Blattabacterium cuenoti contains representatives with a uniformly stocky and compact (hereafter Blattabacterium) in their fat-bodies. Previous studies body shape. They have a thick, scoop-shaped pronotum used as a have suggested that Blattabacterium is passed on from mother to tool in defense or digging, and their legs are powerful and fes- offspring in a strictly vertical fashion, and that the phylogenies of tooned with stout, articulated spines. Members of Cryptocercus live host and symbiont are congruent with each other (Clark et al., in biparental family groups in logs and are considered to be 2001; 2003; Lo et al., 2003; Maekawa et al., 2005). monogamous and subsocial insects (Nalepa et al., 2001). They live Before 1997, only three species of Cryptocercus were reported: in temperate forests of the Nearctic, Palaearctic (Clark et al., 2003) C. punctulatus Scudder, 1862, C. relictus Bey-Bienko, 1935 and C. and Oriental Regions (Wang, 2013). Cryptocercus is the sister group primarius Bey-Bienko, 1938. Additional species were subsequently of termites (Lo et al., 2000; Inward et al., 2007), and representa- recognized by Nalepa et al. (1997) (C. clevelandi), Burnside et al. tives of the genus are considered important models for under- (1999) (C. darwini, C. garciai and C. wrighti (but see Nalepa et al., standing the evolution of termite eusociality (Nalepa et al., 2001; 2002)) and Grandcolas et al. (2001) (C. kyebangensis), mainly on the basis of sequence divergence in mitochondrial DNA and
q chromosome numbers. Some researchers distinguished additional This paper was edited by the Associate Editor Alfried Vogler. species on the basis of female genitalia (C. matilei, Grandcolas, ⇑ Corresponding author. 2000; C. hirtus, C. meridianus and C. parvus, Grandcolas et al., E-mail address: [email protected] (Z. Wang). http://dx.doi.org/10.1016/j.ympev.2016.02.005 1055-7903/Ó 2016 The Authors. Published by Elsevier Inc. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). 202 Y. Che et al. / Molecular Phylogenetics and Evolution 98 (2016) 201–209
2005; C. convexus and C. arcuatus, Wang et al., 2015). Currently the to increase the accuracy of inferred evolutionary timescales (Ho genus comprises 14 species (Beccaloni, 2014; Wang et al., 2015), and Lo, 2013). eight of which are from China. The phylogenetic relationships and divergence times within 2. Materials and methods and between Palaearctic and Nearctic lineages based on molecular data have been examined on multiple occasions, with variable 2.1. Taxon sampling results. Grandcolas (1999) investigated Palaearctic and Nearctic taxa and hypothesized that the American lineage evolved from The taxa used in this study include all known Cryptocercus spe- the Asian lineage. Grandcolas et al. (2001) inferred the splitting cies. Specimens used in the analysis, and their collection localities event between Asian and American Cryptocercus group to have are shown in Table S1 and Fig. 1. Specimens of Cryptocercus from occurred 18–2 Ma based on mitochondrial 12S and 16S rRNA. A China were collected during June 2011 and October 2014. All spec- molecular clock analysis based on Blattabacterium endosymbiont imens studied were deposited in the Institute of Entomology, DNA performed by Clark et al. (2001) indicated that the divergence Southwest University (SWU), Chongqing, or Museum of Hebei times between Asian and American species were much older, at University (HBU), Baoding, China. Specimens were preserved in 115–70 Ma. The divergence between Asian and American lineages 100% ethanol (SWU) or pinned (HBU) for subsequent sequencing. was inferred by Park et al. (2004) to have occurred 61–26 Ma based Extensive field sampling has shown that populations of Cryptocer- on the COII gene. Maekawa et al. (2005) inferred the divergence cus exist in most of the high-forested regions from the North (Gaol- times between Asian and American lineages at between the late ingzi, Heilongjiang Province) to the South (Yunshanping, Cretaceous and early Tertiary (77.8–58.7 Ma). Yulongxueshan, Yunnan Province) of China. They can live either No study to date has included all known Cryptocercus spp. at a relatively low altitude (702 m, Gaolingzi, Heilongjiang Pro- worldwide to infer the phylogenetic relationships between host vince) or a high altitude (3756 m, Shikaxueshan, Yunnan Province) and Blattabacterium, the timescale of their evolution, and the evo- (Table S1, Fig. 1). During our survey from 2006 to 2014, we failed to lution of their karyotypes. In this study, we sequenced three mito- find Cryptocercus in a number of localities; these are listed in chondrial (12S rRNA, 16S rRNA and COII) genes and one nuclear Table S2. All type specimens and samples of Cryptocercus used in (28S rRNA) gene of all Cryptocercus spp. from China, and two genes this study were deposited in the Institute of Entomology, College of their associated Blattabacterium symbionts (16S rRNA and 23S of Plant Protection, Southwest University (SWU) except some indi- rRNA), including eight new species inferred on the basis of their cated as ‘Museum of Hebei University’ (HBU). karyotypes. Combining these sequences with previously published sequences, we performed phylogenetic and divergence date analy- 2.2. Karyotype analysis ses, and inferred the biogeographic history and evolution of chro- mosome number across all known species within the genus. To For karyotype analysis, a total of 13 species were collected from calibrate evolutionary rates, we used fossils from termites, the 18 locations shown in Table S1. For each of the species described closest relatives of Cryptocercus. Calibrating evolution on the basis (see Supplementary Material), 3 individuals were used. Mitotic of fossils closely related to the taxa under investigation is thought chromosomes from the testes of males (thirteen species except C.
Fig. 1. Distribution and collection localities of analyzed specimens of Cryptocercus. Numbers for sampling localities are as indicated in Table S1 (blue circle meaning samples from central and southwestern China, purple circle meaning samples from Manchuria and Korean Peninsula) and Table S2 (empty circle). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.) Y. Che et al. / Molecular Phylogenetics and Evolution 98 (2016) 201–209 203 hirtus) and from follicle cells in the ovaries of females (C. hirtus models were chosen as follows: GTR + I + G: 12S, 16S; TrN + I + G: (YPG)) were examined following Luykx (1983). Karyotypes are COII_pos1, COII_pos2, 28S; TrN + G: COII_pos3 using PartitionFin- reported as the diploid complement. der V1.1.1. Two independent sets of Markov chains were run, each with one cold and three heated chains for 1 Â 107 generations, and 2.3. DNA extraction, amplification, purification and sequencing every 1000th generation was sampled. Convergence was inferred when a standard deviation of split frequencies <0.01 was com- We sampled 4 genes in this study: mitochondrial 12S rRNA, 16S pleted. Sump and sumt burninfrac was set to 25% and contype rRNA and COII, and nuclear 28S rRNA. Total DNA was extracted was set to allcompat. from hindleg tissues of samples preserved in 100% ethanol. The Additional analyses of the Cryptocercus dataset (COII, 12S, 16S extraction procedure was according to the TIANamp Genomic and 28S) plus M. darwiniensis and seven other cockroaches DNA Kit (Tiangen Biotech, Beijing). Fragments of each gene were (Table S5) were performed as described above to compare the amplified using PCR; primers for these amplifications are given topologies with those based on a Blattabacterium dataset. in Table S3. For Blattabacterium analyses, 16S and 23S sequences from For PCR amplification, a 25 ul cocktail of 1 ll DNA template, strains present in Cryptocercus, M. darwiniensis, and seven other
15.25 ul double-distilled H2O (ddH2O), 2 ul MgCl2 (25 mM), 2.5 ul cockroach hosts (Table S5) were obtained from GenBank and com- 10 ⁄ PCR Loading Buffer, 0.25 ul Taq DNA polymerase (Takara; bined with those from the 21 Cryptocercus taxa examined in this 100 mM Tris–HCl, pH8.3, 500 mM KCl), 2 ul dNTP mixture (1 mM study (note that Blattabacterium is absent from all termites other concentration of each dNTP) and 1 ul of each primer was used. than M. darwiniensis). Analyses were performed as described The PCR conditions included an initial denaturing step (95 °C for above. The best-fit model for both 16S and 23S was HKY + I + G. 3 min), followed by 34 cycles of denaturation (30 s at 95 °C), annealing (1 min at 49–54 °C) and extension (1 min at 72 °C), 2.5. Divergence dating analysis and finally an extension period of 72 °C for 5 min. The amplified products were electrophoresed by a 1% agarose gel. PCR products We performed divergence date analyses based the combined were purified following the ColumnMate Gel Extraction Mini Kit mitochondrial/nuclear dataset of Cryptocercus and 44 termite out- instructions. In the case where sequencing was not successful, groups (see Table S4). The molecular clock was calibrated using purified PCR fragments were cloned and sequenced. seven minimum age constraints based on termite fossils as shown We sequenced 2 Blattabacterium genes: 16S rRNA and 23S rRNA in Table 1. Analyses were performed using a relaxed molecular- from 38 Cryptocercus samples and Pycnoscelus surinamensis.To clock model using the Bayesian phylogenetic software BEAST 1.8.0 obtain bacterial DNA, the posterior-most abdominal sternites of (Drummond and Rambaut, 2007). Rate variation was modeled the host were removed and a portion of the fat body was dissected. among branches using uncorrelated lognormal relaxed clocks Endosymbiont DNA was extracted from the fat body and the (Drummond and Rambaut, 2007), with a single model for all genes. extraction procedure was carried out according to the TIANamp A Yule speciation process was used for the tree prior (Gernard, 2008) Genomic DNA Kit (Tiangen Biotech, Beijing). Endosymbiont 16S and posterior distributions of parameters, including the tree, were rRNA and 23S rRNA genes were amplified using the four primers estimated using MCMC sampling. We performed two replicate (33f, 660r, 660f, and 1294r) and two sets of primers (23Saf, MCMC runs, with the tree and parameter values sampled every 23Sar, 23Sbf and 23Sbr) respectively, as described by Clark et al. 5000 steps over a total of 50 million generations. A maximum clade (2001), and the temperature profile for PCR also followed that of credibility tree was obtained using Tree Annotator within the BEAST Clark et al. (2001). PCR products were purified using the TaKaRa software package with a burn-in of 1000 trees. Acceptable sample Extractor Kit Ver.4.0. Extracted products were cloned into a T- sizes and convergence to the stationary distribution were checked vector (promega) and five randomly chosen clones were using Tracer 1.5 (Drummond and Rambaut, 2007). sequenced. All sequences of Chinese Cryptocercus and Pycnoscelus surinamensis endosymbionts were submitted to GenBank. 2.6. Likelihood ratio tests
2.4. Sequence alignment and phylogenetic analysis To examine congruency between the topologies inferred for the Cryptocercus and Blattabacterium datasets, we compared tree In this study, the lengths of 12S rRNA, 16S rRNA, COII and 28S topologies using the approximately unbiased (AU) and Shi- excluding the primer sequenced were approximately 369 bp, modaira–Hasegawa (SH) test in Consel (Shimodaira and 398 bp, 404 bp and 557 bp, respectively. Sequences were aligned Hasegawa, 2001). These topology-constrained analyses were con- using the software MUSCLE 3.8 and were manually adjusted. COII, ducted using the same sequence data and settings as those used 12S, 16S, and 28S sequences of Cryptocercus species present on in the unconstrained analyses. Seven maximum-likelihood analy- GenBank plus those from 44 termite outgroups (Table S4) were ses, each involving a different topological constraint (Table 2) combined with those from the 21 Cryptocercus taxa examined in determined on the basis of the maximum likelihood tree for the this study. Analyses were subsequently performed on the align- host dataset, were performed using RAxML (Stamatakis, 2006) ment that had been manually adjusted, and another alignment that independently on the Blattabacterium dataset. The score of each had ambiguous characters removed via GBlocks, using the default constraint topology was compared with that of the maximum- parameters on the online GBlocks Web Tool (Castresana, 2000). To likelihood unconstrained tree inferred from the Blattabacterium test the congruence of the mitochondrial genes (COII, 12S, 16S) and dataset. We used the GTR + G model and performed 100 individual the nuclear gene (28S), we performed the incongruence length dif- RAxML runs to infer the tree. The log-likelihood of each tree was ference (ILD) test, implemented in Paup4b10 (Swofford, 2002) with estimated in RAxML and these values were used for the AU and 1000 replicates. SH test in Consel. Subsequent analyses were performed on the combined dataset using Maximum likelihood (ML) and Bayesian inference (BI). BI 2.7. Biogeographic analyses was performed using MrBayes 3.2 (Ronquist et al., 2012) and ML was performed using RAxML 7.7.1 (Stamatakis et al., 2008). For Biogeographical reconstructions were performed using the phy- ML, the GTRGAMMA model was selected for the concatenated logenetic tree generated by BEAST with outgroups removed. We datasets, with 1000 bootstrap replicates. For BI, the best-fitting used S-DIVA analysis (Statistical Dispersal-Vicariance Analysis) in 204 Y. Che et al. / Molecular Phylogenetics and Evolution 98 (2016) 201–209
Table 1 Fossils used for estimation of divergence time of major clades in the analysis of Cryptocercus with 44 termite outgroups.
Species Age (Ma)/Minimum Calibration group Soft maximum Reference Age constraint bound (97.5% for group probability) Baissatermes lapideus 137 Cryptocercus + Isoptera 250 Engel et al. (2007) Glyptotermes grimaldii 18 Glyptotermes + Neoterrmes + Rugitermes + Cryptotermes 150 Engel and Krishna (2007a) Dolichorhinotermes dominicanus 18 Dolichorhinotermes + Schedorhinotermes 100 Schlemmermeyer and Cancello (2000) Reticulitermes antiquus 44 Reticulitermes + Coptotermes + Heteroterrmes 100 Engel and Krishna (2007b) Coptotermes priscus 26 Coptotermes + Heterotermes 70 Emerson (1971) Anoplotermes sensu lato 18 Anoplotermes++Ateuchotermes + Astalotermes 70 Krishna and Grimaldi (2009) Microcerotermes insularis 18 Microcerotermes + Embiratermes + Syntermes + Labiotermes 70 Krishna and Grimaldi (2009)
Table 2 Results of topology test among Blattabacterium dataset alternative topologies.
Tree constraint Blattabacterium OBSa AU-testb SH-testb C. relictus and C. kyebangensis monophyletic 0.2 0.552 0.855 C. shangmengensis sp. nov., C. zagunaoensis sp. nov., C. pingwuensis sp. nov., C. primarius, C. convexus and C. matilei monophyletic 1 0.478 0.864 C. shangmengensis sp. nov. and C. primarius monophyletic 1.2 0.538 0.806 C. shangmengensis sp. nov., C. zagunaoensis sp. nov., C. pingwuensis sp. nov., C. habaensis sp. nov., C. primarius, C. convexus, 2.5 0.377 0.918 C. arcuatus and C. matilei monophyletic C. garciai, C. darwini and C. punctulatus monophyletic 4 0.08 0.671 C. meridianus, C. relictus and C. kyebangensis monophyletic 15.1 0.045 0.363 All constraints listed above, except C. garciai, C. darwini and C. punctulatus 16.7 0.090 0.311
a OBS indicates the difference in the likelihood between datasets being allowed to have different topologies and the highest likelihood obtained when all datasets are constrained to have the same topology. b Significance levels were determined in the tests. The values indicate the probability that the value of the ML tree for a given dataset is significantly higher than that for the alternative topologies. P < 0.05 significantly different trees.
RASP 3.02 (Yu et al., 2014) to reconstruct the possible ancestral sp. nov., C. habaensis sp. nov., C. wuxiensis sp. nov., C. shennongjiaen- ranges of Cryptocercus. Four geographical regions were divided, sis sp. nov., C. ningshanensis sp. nov. and C. neixiangensis sp. nov.); i.e., A – Central and Southwestern China (including Hengduan each is named after their type locality, and species descriptions Mountains, Qin-Daba Mountains), B – Manchuria and Korea Penis- are provided in the Supplementary material. Chromosome num- sula, C – Western America, and D – Eastern America. We loaded bers in Asian species of Cryptocercus were found to vary between 10,000 MCMC trees from BEAST analysis, and discarded the first 2n =15to2n = 41. 500 trees prior to running S-DIVA with default parameters. The number of maximum areas was kept as 4. 3.3. Phylogenetic analysis among Cryptocercus
3. Results The results of an incongruence length difference (ILD) test indi- cated that it was appropriate to combine the mitochondrial genes 3.1. Sequence variation (COII, 12S, 16S) with the nuclear gene (28S) (p = 0.99). Following concatenation of these four genes we performed analyses on an Approximately 1963 bp of the combined of 12S rRNA (461 bp), alignment that had been trimmed using GBlocks, as well as an 16S rRNA (454 bp), COII (404 bp) and 28S rRNA (644 bp) host unmodified alignment. The results from each of the analyses were sequences, were analyzed for 35 ingroup samples (Table S1) and essentially identical, and we present only the results based on the 44 outgroup taxa (Table S4). 1032 bp were variable, 896 bp were former. The topology derived from ML analyses show that Nearctic conserved and 813 bp were parsimony-informative. The average and Palaearctic Cryptocercus groups each form monophyletic nucleotide composition proportions for the combined data parti- groups (Fig. 3). The Asian lineages examined formed two major tion were: T, 24.7%; C, 21.3%; A, 39.0%; and G, 15.1%. and well supported clades, a and b. Within Clade a, taxa from 16S rRNA and 23S rRNA Blattabacterium sequences (76 in all) of the Hengduan Mountains (numbers 4–12 in Figs. 1 and 3) formed 16 Cryptocercus species from 22 locations were determined in this a monophyletic group, with the exception of C. meridianus (number study. The lengths of 16S rRNA and 23S rRNA excluding the primer 1inFigs. 1 and 3), which was most closely related to taxa from sequenced were approximately 1259 bp and 1116 bp respectively. Northeastern China (Manchuria) and Korea (numbers 20–24 in Numbers of variable and parsimony-informative positions were Figs. 1 and 3) (Clade c). Cryptocercus spp. from the Qin-Daba moun- 181 and 114 for 16S rRNA and 165 and 98 for 23S rRNA genes tains (numbers 13–19 in Figs. 1 and 3) formed a monophyletic respectively. The nucleotide composition of both genes showed group (Clade d). an A–T bias (63.2% and 54.8%, respectively). 3.4. Phylogenetic relationships among Blattabacterium spp. 3.2. Taxonomy and chromosome numbers of Chinese Cryptocercus BI and ML analyses of the 16S and 23S rRNA Blattabacterium Chromosome numbers were found to be highly variable among datasets produced similar trees. Nearctic and Palaearctic Cryptocer- Chinese taxa (Fig. 2). On the basis of divergent chromosome cus Blattabacterium each formed monophyletic groups (Asian numbers and genetic divergence we identified 8 new species (C. group: BPP = 100, MLB = 100; American group: BPP = 100, shangmengensis sp. nov., C. zagunaoensis sp. nov., C. pingwuensis MLB = 100) (Fig. 4). The Asian endosymbiont lineages examined Y. Che et al. / Molecular Phylogenetics and Evolution 98 (2016) 201–209 205
Fig. 2. Mitotic chromosomes of Cryptocercus species in males except C. hirtus (YPG) (a), C. primarius (DCP) (b), C. matilei (BEX) (c), C. shangmengensis sp. nov. (LJG) (d), C. convexus (BRG) (e), C. zagunaosis (DZMG) (f), C. pingwuensis sp. nov. (BSG) (g), C. habaensis sp. nov. (HBXS) (h), C. relictus (SFLC) (i), C. relictus (MDF) (j), C. hirtus (YPG) (k), C. wuxiensis sp. nov. (YTL) (l), C. shennongjiaensis sp. nov. (SBT) (m), C. ningshanensis sp. nov. (LBYG) (n), C. neixiangensis sp.nov. (BTM).
Fig. 3. Maximum likelihood (ML) tree derived from analysis of combined data 12S rRNA, 16S rRNA, COII and 28S rRNA genes. d near nodes indicating both BPP and MLB > 95, j representing only MLB > 95. Outgroups are not shown. See Table S1 for abbreviations of each location and number of location. Numbers after the species name are chromosome numbers and number of location. The topology shown was very similar to that derived from BI analysis, with some minor differences (see Fig. S2). formed two major clades, one of which was from the Hengduan Fig. 4 shows a comparison of the phylogeny of a subset of Mountains (numbers 1, 4–12 in Figs. 1 and 4), the other from endosymbionts with that of their hosts. Some differences in the Northeastern China (Manchuria), Korea, Russia and Qin-Daba relationships between these two groups of taxa were found. For Mountains (numbers 13–22 in Figs. 1 and 4). example, C. meridianus clusters with the Northeastern Asia group 206 Y. Che et al. / Molecular Phylogenetics and Evolution 98 (2016) 201–209
(C. relictus + C. kyebangensis); C. punctulatus is the sister to the was not found to be significant based on the SH test. Overall, group of C. darwini and C. garciai. However AU and SH tests found the results from both the SH and AU tests indicate that host that these differences were not significantly different in almost all and bacterial topologies are not significantly different. We there- cases (Table 2). The constraint ‘‘C. meridianus + C. relictus + C. kye- fore could not reject the null hypothesis that co-cladogenesis has bangensis” resulted in a likelihood score that was significantly occurred throughout the evolutionary history of hosts and their lower than the best tree from the Blattabacterium dataset based symbionts, despite the apparent topological incongruities shown on the AU test (P = 0.045), although the difference in these trees in Fig. 4.
Fig. 4. Comparison of Cryptocercus and Blattabacterium Bayesian (BI) and Maximum likelihood (ML) tree derived from analysis of the concatenated 12S rRNA, 16S rRNA, COII and 28S rRNA datasets and the combined data 16S rRNA and 23S rRNA, respectively. d near nodes indicating both BPP and MLB > 95, j representing only MLB > 95. Numbers above nodes indicate the percentage of posterior probabilities in BI analyses and bootstrap support in ML analyses (1000 replicates). Asterisks indicates nodes that were supported in 100%. See Table S1 for abbreviations of each location and number of location. Numbers below or after the species name are the number of location. Y. Che et al. / Molecular Phylogenetics and Evolution 98 (2016) 201–209 207
3.5. Divergence date analyses from 8 new and 2 recently discovered (Wang et al., 2015) Asian species, along with those previously reported from Asia and North The timescale for evolution of Cryptocercus species diversifica- America. Our analyses of Cryptocercus and Blattabacterium datasets tion based on 12S, 16S, COII and 28S and calibrated using 7 termite recovered similar topologies, in agreement with previous studies fossils is shown in Fig. 5 (further details are shown in Fig. S3). The that indicated strict vertical inheritance of these symbionts divergence of the lineages leading to termites and Cryptocercus was (Clark et al., 2001; 2003; Maekawa et al., 2005). Differences in estimated to have occurred 145.84 Ma (137–170.63 Ma 95% CI). the phylogenies inferred from host and endosymbiont were not American and Asian Cryptocercus were estimated to have diverged found to be significantly different based on almost all AU and SH 55.09 Ma (41.55–72.28 Ma 95% CI). tests, although the position of C. meridianus is somewhat anoma- lous (see below). 3.6. Biogeographic analyses The estimated divergence time of termites and Cryptocercus (145.84 Ma (137–170.63 Ma 95% CI) shown in Fig. S3) is largely A phylogenetic tree showing the reconstruction of ancestral dis- consistent with some recent estimates (Misof et al., 2014: tribution ranges is shown in Fig. 5. The ancestor of extant Asian and 145 Ma; Tong et al., 2015: 140 Ma), although slightly younger than American Cryptocercus was inferred to have been distributed others (Djernæs et al., 2015: 185 ± 19 Ma; Legendre et al., 2015: across a larger region incorporating areas in both Asia and North 195 Ma). The relative lack of branches from 150 Ma to 60 Ma in America. Ancestral Asian Cryptocercus were inferred to have origi- the lineage leading to Cryptocercus compared with the termite nated in Central China; ancestral North American taxa were clade (Fig. S3) indicates either very low levels of diversification of inferred to have been present in a wider area encompassing both the genus through this period, or extinction of a number of lineages Western and Eastern North American areas. Vicariance was that emerged during this period. The Cryptocercus examined fall inferred between the western and eastern American Cryptocercus into two distinct clades (American and Asian) with high support group. In Asia, the Cryptocercus group from Manchuria and Korean values, which is consistent with the results of previous studies Peninsula was inferred to have been of Central Chinese origin. (Grandcolas et al., 2001; Clark et al., 2001; Maekawa et al., 2005; Lo et al., 2006). Our results do not support the notion that Asian 4. Discussion C. relictus is the sister group to the North American Cryptocercus species (Kim et al., 2013). The 55.09 Ma (41.55–72.28 Ma 95% CI; 4.1. Global phylogeny and biogeography of the Cryptocercidae Figs. 5 and S3) divergence time of the Asian and American Crypto- cercus lineages overlaps with the range of divergence times (58.7– The present study represents the most comprehensive phyloge- 77.8 Ma) estimated by Maekawa et al. (2005), but is somewhat netic investigation of Cryptocercus to date, including sequences younger than the estimate of 115–70 Ma estimated by Clark
Fig. 5. Reconstruction of Cryptocercus ancestral distribution using Statistical Dispersal-Vicariance Analysis. The map shows the biogeographic areas, namely: A – Central and Southwestern China (including Hengduan Mountains, Qin-Daba Mountains), B – Manchuria and Korean Peninsula, C – Western America, and D – Eastern America. The phylogenetic tree of Cryptocercus was based on 12S rRNA, 16S rRNA, COII, and 28S rRNA using BEAST; a timescale for the evolution of the genus is shown at the bottom of the tree. 208 Y. Che et al. / Molecular Phylogenetics and Evolution 98 (2016) 201–209 et al. (2001). The different minimum rates of endosymbiont 16S as well as those from Northeastern Asia are required to investigate evolution by Maekawa et al. (2005) (0.0084 substitutions/ this result further. site/50 Ma) and by Clark et al. (2001) (0.006 substitutions/ site/50 Ma) may account for the different estimated divergence 4.3. Chromosome evolution of Cryptocercus times. Within American Cryptocercus, phylogenetic relationships are similar to those inferred previously (Clark et al., 2001; The generally low variability of morphological characters Maekawa et al., 2005) with the exception of the phylogenetic posi- among species of Cryptocercus has led some researchers to use tion of C. wrighti and C. punctulatus. Clark et al. (2001) hypothe- chromosome number and DNA sequence divergence to distinguish sized that the speciation events in the eastern USA took place species (Burnside et al., 1999). We have used a similar approach in during the past 25 Ma, which is in broad agreement with our combination with molecular phylogenetic analysis to propose 8 results. new species of Asian Cryptocercus. There are 10 different male Based on our reconstruction of the ancestral distribution ranges karyotypes among Chinese Cryptocercus, i.e., 2n = 15, 17, 19, 21, and divergence times of Cryptocercus worldwide (Fig. 5), we offer 25, 29, 31, 35, 37 and 41;chromosomal numbers found in North the following tentative scenario for the current global distribution American Cryptocercus species vary from 2n =37 to 2n =47 of Cryptocercus. The ancestral population that would give rise to (Nalepa et al., 2002; Burnside et al., 1999; Everaerts et al., 2008). Cryptocercus and termites diverged into two lineages during the The overall range in the genus (2n = 15–47) is comparable to that late Jurassic/early Cretaceous. During the Cretaceous (62.1– found in chromosomally diverse butterfly genera, such as Lysandra 150.24 Ma) the ancestor of extant Cryptocercus inhabited the tem- (n = 24–93; Talavera et al., 2013) and Agathymus (n = 5–38; perate deciduous forests of the Arcto-Tertiary complex in the Lukhtanov, 2014). Chromosomal fusion–fission events may northern regions of the globe. A general cooling trend began during account for the variable chromosome numbers of Cryptocercus spe- the Cretaceous, which forced the Arcto-Tertiary community (pre- cies (Lo et al., 2006; Everaerts et al., 2008). sumably including ancestral Cryptocercus lineages) to move south In our study, phylogenetically divergent Cryptocercus species (MacGinitie, 1958) into Asia and North America. The boreotropical had the same chromosome numbers, for example, C. pingwuensis flora is thought to have spread between Eurasia and America dur- sp. nov. (BSG) and C. shennongjiaensis sp. nov. (SBT) (2n = 29), C. ing the early Eocene (55 Ma) via early connections through Ber- habaensis sp. nov. (HBXS) and C. meridianus (YSP) (2n = 19), C. wux- ingia, and possibly through the North Atlantic Land Bridge iensis sp. nov. (JTL) and Cryptocercus darwini from America (Tiffney, 1985). The estimate for the separation of Asian and Amer- (2n = 37). The results support the proposal by Everaerts et al. ican Cryptocercus lineages (i.e. 55.09 Ma (41.55–72.28 Ma 95% CI)) (2008) that distinct lineages with the same chromosome numbers combined with our S-DIVA analyses suggest that gene flow can represent different species. between these two lineages ceased by the early to mid Eocene. The possibility that Cryptocercus evolved in one region and then migrated to the other via land bridges cannot be ruled out, 5. Conclusion although the branching order of taxa in Figs. 3 and 5 does not sup- port this scenario. Our study has shed light on the factors leading to the global dis- tribution pattern of Cryptocercus. Compared with the termite lin- 4.2. Phylogeny and biogeography of Asian Cryptocercidae eage, the Cryptocercus lineage has undergone far less diversification and/or higher levels of extinction since these lin- By the Eocene (56–34 Ma), a number of previous topographic eages diverged 146 Ma. This may be due to the failure of the lat- barriers are thought to have receded, and most areas of China are ter to adapt to warmer climes. We conclude that the ancestral believed to have belonged to one zoogeographic province (Guo, Cryptocercus population existed in a continuous region that encom- 1983). Thus Cryptocercus may have been widespread in areas passed parts of northern Asia and North America until the late Cre- including the Qin-Daba Mountains, the Hengduan Mountains and taceous/early Tertiary period. This ancestral species the diversified Manchuria (including Northeastern China and Russia) during this into two main lineages as it was forced toward the south due to period. Toward the end of the Eocene, large extinctions of gym- cooling conditions, with gene flow between Asian and American nosperms, including pine, fir, larch and spruce – the hosts of Cryp- populations ceasing by the early to mid Eocene. The last common tocercus – are thought to have occurred when global temperatures ancestor of Asian Cryptocercus was inferred to have existed in Cen- declined sharply (Crisp and Cook, 2009; Wang and Ran, 2014). This tral China 20.97 Ma. The phylogenetic position of C. meridianus may explain our finding that that last common ancestor of Asian based on mitochondrial and nuclear DNA data was unexpected Cryptocercus existed as recently as 20.97 Ma (15.78–27.21 Ma due to its geographic position, and differed from its position based 95% CI). One interpretation of our RASP analysis is that when the on Blattabacterium sequences. Further investigations on this spe- climate of southern China became warmer and wetter in the Mio- cies, and the group as a whole using larger numbers of genes are cene (Jacques et al., 2011), the survivors present in the Hengduan warranted. Our study paves the way for a better understanding and Qin-Daba Mountain regions moved north-eastwards toward of this important group of cockroaches. Manchuria and Korea, forming the extant distribution pattern. C. kyebangensis from Korea formed a sister taxon relationship with the group of C. relictus from Russia and Northeast China, Acknowledgments and C. meridianus from Yunnan on the basis of mitochondrial and nuclear DNA analyses (Figs. 3 and 5). This relationship is somewhat We are sincerely grateful to Prof. Guodong Ren (Hebei Univer- unexpected, given the large geographical distance between C. sity, China) for his kindness in loaning specimens to us and also meridianus and the species from Northeast China, Russia, and thank Prof. Jinjun Wang (Southwest University, China) for his kind Korea, and the close geographic proximity of C. meridianus to other permission for the research to take place in the laboratory of species in the Hengduan Mountains. Importantly, the sequences of molecular ecology. This study is supported by the National Natural C. meridianus obtained in our study were highly similar to those Sciences Foundation of China (Nos. 31093430 and 31472026), and obtained in previous studies (Lo et al., 2006), confirming that the also partly by the Project sponsored by the Scientific Research C. meridianus sequences we obtained are not artifactual. Further Foundation for the Returned Overseas Chinese Scholars, State Edu- analyses of taxa from the region in which C. meridianus is found cation Ministry. Y. Che et al. / Molecular Phylogenetics and Evolution 98 (2016) 201–209 209
Appendix A. Supplementary material Klass, K., Nalepa, C., Lo, N., 2008. Wood-feeding cockroaches as models for termite evolution (Insecta: Dictyoptera): Cryptocercus vs. Parasphaeria boleiriana. Mol. Phylogenet. Evol. 46, 809–817. Supplementary data associated with this article can be found, in Krishna, K., Grimaldi, D., 2009. Diverse Rhinotermitidae and Termitidae (Isoptera) in the online version, at http://dx.doi.org/10.1016/j.ympev.2016.02. Dominican amber. Am. Mus. Novit. 2640, 1–48. 005. Legendre, F., Nel, A., Svenson, G.J., Robillard, T., Pellens, R., Grandcolas, P., 2015. Phylogeny of Dictyoptera: dating the origin of cockroaches, praying mantises and termites with molecular data and controlled fossil evidence. PLoS ONE 10 (7), e0130127. http://dx.doi.org/10.1371/journal.pone.0130127. Lo, N., Bandi, C., Watanabe, H., Nalepa, C., Beninati, T., 2003. Evidence for cocladogenesis between diverse dictyopteran lineages and their intracellular References endosymbionts. Mol. Biol. Evol. 20, 907–913. http://dx.doi.org/10.1093/molbev/ msg097. Beccaloni, G.W., 2014. Cockroach Species File Online. Version 5.0/5.0. Lo, N., Tokuda, G., Watanabe, H., Rose, H., Slaytor, M., Maekawa, K., Bandi, C., Noda, World Wide Web electronic publication.