Three New Species of the Wood Roach, Cryptocercus (Blattodea: Cryptocercidae), From the Eastern United States

Craig A. Burnside 1, Paul T. Smith and Srini Kambhampati

Department of Entomology
Kansas State University
Manhattan, KS 66506 USA

1 For correspondence contact
Dr. Craig A. Burnside
Tel: (785) 532-6154; Fax: (785) 532-6232

Submitted for publication: December 3, 1998
Erratum added February 4, 2000

Keywords:  Cockroaches, mitochondrial DNA,  species description



Three new species of the wood roach, Cryptocercus (Blattodea: Cryptocercidae), are described from the Appalachian mountains in the eastern United States. The description is based on consistent differences in DNA sequence of the two mitochondrial rRNA genes, chromosome number, geographic range, and morphology. Until recently, it was assumed that all Cryptocercus in the eastern United States were conspecific.  However, data on variation in DNA sequence and chromosome number from samples collected over much of the distribution of Cryptocercus in the eastern United States suggested the existence of four evolutionary lineages and forms the basis for the new species described here.   In addition to describing the new species, we provide diagnostic characters, notes on the geographical distribution, and DNA sequence for the 12S rRNA and 16S rRNA gene fragments of holotype specimens.


The cockroach genus, Cryptocercus (Blattodea: Cryptocercidae) at present consists of four recognized species worldwide: Cryptocercus relictus in eastern Russia, Cryptocercus primarius in eastern China, Cryptocercus clevelandi in the western United States and Cryptocercus punctulatus in the eastern United States. Prior to 1996, all Cryptocercus in the United States were considered a single species (C. punctulatus).   However, Kambhampati et al. (1), based on variation in DNA sequence of the mitochondrial rRNA genes and chromosome number, and mating studies, suggested that there may be five or more species of Cryptocercus in North America.  Kambhampati et al. (1), provided the quantitative evidence and framework for the designation of individuals in Oregon as a new species (C. clevelandi) by Nalepa et al. (2). In this paper, we describe three new species of Cryptocercus from the eastern United States, based on variation in DNA sequence of mitochondrial rRNA genes, chromosome number, geographic distribution, and morphology.


Insects:  Insects were collected from decaying pine and oak logs throughout the Appalachian and Allegheny Mountains in the eastern United States during August 1998.  The collection sites were located in the states of Georgia, Alabama, North Carolina, Tennessee, Kentucky, Virginia and West Virginia. Wooded areas were searched for decomposing logs, which were split open to access the Cryptocercus galleries.  Nymphs and adults were placed immediately in 80% ethanol. When possible, specimens were collected from at least two different logs at each location, and the individuals collected at any one location generally represented several families.

A list of locations and other collection details are given in Table 1. A total of 24 locations are represented in this study. Of these, 16 sites were for which Kambhampati et al. (1) reported data; however, their sample size (number of individuals per location) was relatively small, with sites in the eastern United States being represented by 1-2 individuals. We resampled eight of the same locations as Kambhampati et al. (1) to increase the sample size  by 2-5 fold. Since within site nucleotide variation tended to be small or nonexistent, the eight sites sampled by Kambhampati et al. (1), which were not resampled and for which only a single individual was available, were included in the present analysis. Eight new sites were sampled and included in this analysis.

DNA Extraction, PCR, and DNA sequencing: DNA was extracted from a small portion of the muscle from the third coxa as described by (1). Polymerase chain reaction was carried out as described by (3). Portions of the mitochondrial 16S rRNA and 12S rRNA genes were amplified using the primers in (4). The fragments were purified as described by (1).  Sequencing reactions using the dRhodamine dye terminator kit (ABI, Inc.) were set up according to the manufacturer's instructions. The DNA pellet was dried and sent to the University of Florida's DNA Sequencing Core Laboratory for automated sequencing on an ABI 377 sequencer.

Phylogenetic, karyotype, and morphological analyses: Both strands of the amplified fragment were sequenced.  The sequences and chromatograms were examined using ABI's Sequence Navigator software and any ambiguities between the complementary strands were corrected. The sequence of the amplified fragment was obtained from several individuals from each of 24 locations (see Table 1 for sample sizes) to estimate the extent of within site variation.  Sequences were initially aligned using CLUSTAL X and manually adjusted as needed.  Unweighted and weighted maximum parsimony analyses were carried out using PAUP* ver. 4d64 (written and provided by D.L. Swofford).  In PAUP*, the multiple equally parsimonious heuristic or branch and bound search options were used with tree bisection-reconnection branch swapping. An initial parsimony analysis in which all characters were weighted equally and gaps were treated as a fifth base was undertaken. This was followed by a successive weighting of characters (5, 6, 7) based on rescaled consistency index (= product of consistency and retention indices (8)) and the best fit of the characters over all equally parsimonious trees. The successive weighting procedure was iterated until no change in tree length occurred. For both weighted and unweighted parsimony, 100 replications of random taxon addition sequences was used. Bootstrap analysis was performed on both unweighted and weighted data sets using the "fast" stepwise addition search rather than the full heuristic search due to computational limitations. However, we bootstrapped the dataset for 10,000 iterations. Decay index (9) was calculated using the computer program AutoDecay 3.03 (written by T. Eriksson and N. Wikstorm). Because the DNA sequence of  the rRNA genes among individuals from a given location was nearly invariant, we included the DNA sequence of one individual chosen at random per location in the phylogenetic analysis. The choice of the individual did not have an effect of the topology of the tree. The sequences reported in this study have been deposited in GenBank database (accession numbers pending). The aligned dataset is available from the authors upon request.

Kambhampati et al. (1)  reported chromosome number for about 50% of the sites included in this study.  We refer the reader to (1)  for details of methods.  Kambhampati et al. (1)  showed that there is a  close correspondence between DNA sequence of the two mitochondrial rRNA genes and chromosome numbers. That is, individuals representing each chromosome number sort out as monophyletic lineages with strong quantitative support when the DNA sequence of the the rRNA genes are used in a parsimony analysis. This enables one to infer the chromosome number of an individual reliably based on the DNA sequence of the  mitochondrial rRNA genes.

To identify differences in morphology, adult males and females were examined under a dissecting microscope. Initially we examined the characters mentioned in (2) as being diagnostic for C. clevelandi relative to C. punctulatus.   Many of the characters mentioned in  (2) were invariant among the eastern U.S. populations.  Ten to fifteen individuals of  each chromosome number type from different locations were examined for morphological  differences.


The morphological features of Cryptocercus throughout the United States are essentially as described by (10) and (2).  The latter authors reported a few minor differences among samples from Oregon and Virginia.  We examined the morphology of specimens from several locations (with known chromosome number) and could identify only two consistent differences in morphology (see below for details). Given the paucity of morphological differences among samples collected in the eastern United States, a general morphological description is perhaps of little or no utility. That is, the new species we describe here are morphologically highly similar to C. punctulatus. However, the levels of divergence in DNA sequence and chromosome number are consistent with those expected among congeneric species (see (1) for detailed comparisons to other cockroaches and insects). Therefore, we base species designations, within this cryptic species complex, on three criteria: consistent and fixed differences in DNA sequence of the two mitochondrial rRNA genes, differences in chromosome number previously reported by (1), and geographic range.

Variation in DNA sequence of mitochondrial genes: DNA sequence was obtained from a total of 83 individuals for the 12S rRNA fragment and 78 individuals for the 16S rRNA fragment (including those reported by (1)). The alignment was relatively straightforward and required the insertion of few gaps. We used the homologous DNA sequence of the migratory locust, Locusta migratoria, as the outgroup. In addition we included in the analysis, the DNA sequence of several individuals from Oregon (=C. clevelandi) because previous analysis (1) suggested that C. clevelandi is basal among North American Cryptocercus. The use of either L. migratoria, C. clevelandi, or both as outgroups, resulted in trees which were identical in topology (not shown).

The alignment of the sequences of the two mitochondrial genes resulted in a total of 880 characters. Of these, 370 (42%) were variable and 193 (22%) were parsimony informative. There was little variation in the DNA sequence among individuals collected within a location, varying from 0% to 0.3% for 12S rRNA and 0% to 1% for the 16S rRNA (Table 2) (The variation in the 16S rRNA fragment among individuals in Asheville, at 3%, was higher than other sites due to one individual which had a few unique characters). Variation among individuals with the same chromosome number ranged from 0-3% (Table 3).

Unweighted parsimony, in which all characters were given equal weight and gaps were treated as a fifth base, resulted in the identification of 8 equally parsimonious trees, each of 638 steps. Within the samples from the eastern United States, three monophyletic lineages were identifiable (Figure 1): (1) Asheville, NC; Bear Trap Gap, NC; Little Switzerland, NC; Panther Creek State Park, TN; Doggett Mountain, NC; and Roan State Park, TN (2) Powell Gap, VA; Mountain Lake, VA; Dripping Rock, VA; and Monongahela National Forest, WV; (3) Vogel State Park, GA; Cherry Log, GA; and Black Rock Mountain, GA; Highlands, NC;  Buck Creek Trail, NC; Lake Guntersville, AL; Desoto State Park, AL; Cumberland Mountain State Park, TN; Price Mountain, TN; Cheaha State Park, AL; Daniel Boone National Forest, KY; Big South Fork, TN; Swain County, NC; and Little River Canyon, AL. Each of the evolutionary lineages derived from the DNA sequence of the mitochondrial rRNA genes corresponds to a different chromosome number. The lineages 1 and 2 correspond to chromosome numbers 2n = 45 and 43, respectively. Lineage 3 includes chromosome numbers 2n = 37 and 39; however, the three samples representing the chromosome number 2n = 39 formed a monophyletic lineage with moderate bootstrap and decay index support (Figure 1). Individuals with 2n = 37 chromosome number were paraphyletic. That is, within these samples, there were two lineages, interrupted by a monophyletic lineage consisting of samples from Georgia (2n = 39). One consisted of 8 samples and another of 3 samples. However, the individuals with 2n = 37 chromosome number formed a monophyletic lineage in both neighbor joining and maximum likelihood analyses (trees not shown). In the neighbor-joining analysis, which was undertaken using Kimura's 2-parameter distance in PAUP*, individuals with 2n = 37 chromosome number formed a monophyletic lineage with reasonable bootstrap support (60%) whose sister group were individuals with 2n = 39 chromosome number. In addition, preliminary results from unpublished data from the ITS2 region also supports the monophyly of these groups (pers. comm. Shaon Hossain). Given the differing results from the parsimony and neighbor joining analyses and the fact that the samples share the same chromosome number, we will designate each lineage as a single species (see below).

The weighted parsimony analysis resulted in a tree (Figure 2) that was similar to the one based on unweighted parsimony ( Figure 1). Any differences were due to a greater resolution of relationships within individual clades. There was greater bootstrap and decay index support for each of the evolutionary lineages mentioned above . The inclusion or the exclusion of the gaps did not have an effect on the topology of the trees.

Chromosome number variation: As mentioned above, Kambhampati et al. (1) identified the presence of four different chromosome number types in the eastern United States. Kambhampati et al. (1) discussed the fact that the observed chromosomal differences (with male complements of 23II+X, 22II+X, 21II+X, 19II+X and 18II+X) suggest an evolutionary series rather than random changes as indicated by the distribution of chromosome numbers on the phylogenetic tree (Figure 1 and Figure 2) and their geographic distribution (see below). The data as a whole suggest the splitting of lineages was accompanied by a reduction in chromosome number as shown in Figure 1 where the nodes of successively interior branches are associated with successively lower chromosome numbers. It is worth noting that the missing chromosome number in the above series, 20II+X, has not yet been detected. Although it is possible that this chromosome number does not exist, it also possible that the region(s) in which it occurs have not been sampled.

Morphological variation: The species of Cryptocercus have traditionally been distinguished based on variation in the armature of the legs and the sculpturing of the posterior margin of the terminal abdominal segments (10, 11, 12). Nalepa et al. (2) utilized minor variation in these characters for assigning the Cryptocercus in the northwestern United States to a new species (C. clevelandi). To identify differences in these and other morphological characters among the four chromosome number types from the eastern United States, we examined 10-20 adults of each sex from several localities under a dissecting microscope. The following characters were examined for differences: (1) number of antennal segments + flagellomeres, (2) number of spines associated with the posteroventral and posterodorsal edges of the femora, (3) tibial spines, (4) structure of the posterior margin of 8th abdominal sternum, and (5) structure of the terminal abdominal segments (9th and 10th) and their associated structures (i.e., cerci, styli and phallomeres). The only notable differences observed among the four chromosome types were the structure of the 10th abdominal tergum (epiproct) and 9th abdominal sternum (subgenital plate). The epiproct among individuals with a chromosome number of 2n = 45 is only approximately triangular, with concave sides, whereas in individuals that are 2n =37, 2n = 39 and 2n = 43 (2n = 43 = C. punctulatus) it is more triangular with straight sides converging to the apex ( Figure 3). Likewise, the sub-genital plate of the 2n = 45 individuals is more broadly rounded between the bases of the styli than in the 2n = 37, 2n = 39 and C. punctulatus individuals, where it is more evenly rounded (Figure 3). With the exception of these differences, individuals with the chromosome numbers 2n = 37, 2n = 39, and 2n = 45 are morphologically similar to C. punctulatus.  We refer the reader to Hebard (11) for a general description of  Cryptocercus of all four chromosome types except for differences noted above. The lack of significant morphological variation among the chromosome number types is not surprising given the constraints imposed by certain restrictive life history characteristics of the genus (e.g., entire life cycle within galleries in rotting logs). Furthermore, substantive, adaptive morphological divergence can be expected when speciation is driven by directional selection and/or accompanied by significant habitat or niche shifts, neither of which is evident in Cryptocercus. If the life history remains essentially the same and speciation was driven by vicariance, significant morphological variation among closely related taxa is not a necessary consequence (14). Moreover, although we do not yet have an estimate of the divergence time among the various chromosome numbers, it appears that the divergence may have occurred within the last 50 million years based on the degree of sequence divergence in the two mitochondrial rRNA genes (see 1 , 2, and 13 for details).

Diagnostic characters: Since we detected only two obvious morphological differences between chromosome number 2n =45 and others (Figure 3), several diagnostic bases that characterize the evolutionary lineages/chromosome number types are provided in Table 4.  Although there are not any characters that individually can differentiate among the four species in the eastern United States, a combination of two or more sites enables the discrimination among the eastern lineages as well as among these lineages and C. clevelandi. The fact that these variable bases are based on multiple individuals per location and chromosome numbers, suggests that many of the characters are consistent enough to be diagnostic for each lineage. Furthermore, these characters could potentially provide a basis for distinguishing among the species in the eastern United States using RFLP analysis.


Cryptocercus darwini n. sp. This species is characterized by the following chromosome number: male diploid complement: 37 (18 autosomal bivalents + X chromosome). Although females have not been examined for this study, they are expected to have a diploid complement of 38 (18 autosomal bivalents + 2 X chromosomes). Kambhampati et al. (1) reported that where examined, the female Cryptocercus contained an additional X chromosome relative to males due to the XX female: XO male sex determination system of cockroaches (15).

The holotype of C. darwini is characterized by the following sequence of the 12S rRNA and 16S rRNA genes. The diagnostic bases, (=characters) that enable the discrimination of C. darwini from one or more of the other Cryptocercus species are given in Table 4.



The geographical distribution of C. darwini is shown in Figure 4 and Figure 5. On a north-south axis, its distribution includes northeastern Alabama, inside western Talledega National Forest, central Tennessee, extending to northcentral Tennessee and southcentral Kentucky. There is some ambiguity with regard to the individuals from Cumberland Mountain State Rustic Park, Big South Fork National River and Recreation Area, and Daniel Boone National Forest. Nonetheless, we will include the above three samples within C. darwini with the caveat that further sampling and study is needed to determine with greater certainty the status of the above three samples. On an east-west axis, the distribution of C. darwini extends east to Highlands, NC, in the Nantahala and Pisgah National Forests (Figure 4 and Figure 5). It is not clear how far west in Alabama, Tennessee and Kentucky, the distribution of Cryptocercus in general extends. However, the montane, forested habitat that Cryptocercus apparently prefers does not occur much west of Nashville, Tennessee.

Etymology: This species is named in honor of the late Charles Robert Darwin, who proposed the theory of natural selection.

Holotype:  Desoto State Park, AL.  Deposited in Kansas State University's Museum of Entomological and Prairie Arthropod Research.

Cryptocercus garciai n.sp. This species is characterized by the following chromosome number: male diploid complement: 39 (19 autosomal bivalents + X chromosome). Although females have not been examined for this study, they are expected to have a diploid complement of 40 (18 autosomal bivalents + 2 X chromosomes) (1).

The holotype of C. garciai is characterized by the following sequence of the 12S rRNA and 16S rRNA genes. The diagnostic bases (=characters) that enable the discrimination of C. garciai from one or more of the other Cryptocercus species are given in Table 4.

12S rRNA:

16S rRNA:

The geographical distribution of C. garciai is shown in Figure 4 and Figure 5. Based on sampling to date, this species has a relatively limited distribution in the montane forest region of northern Georgia. All samples were collected in the Chattahoochee National Forest. The westernmost sample comes from a location ~5 km east of Cherry Log, GA. The easternmost sample was obtained from Black Rock Mountain State Park, GA (1), which is only approximately 35 KM from Highlands, NC, a location from which C. darwini was collected. It is not clear how far south the distribution of C. garciai extends.  Apparently the habitat of Cryptocercus does not extend further south than Gainesville, GA, which is approximately 45 KM from the southern edge of the Chattahoochee National Forest.  A survey of insects in logs conducted south of the Chattahoochee National Forest near Athens, GA did not reveal the presence of Cryptocercus (M. R. Whiles, pers. comm.).

Etymology: This species is named in memory of the late Jerry Garcia, a member of the musical group, The Grateful Dead. What a long strange trip its been.

Holotype: Cherry Log, GA. Deposited in Kansas State University's Museum of Entomological and Prairie Arthropod Research.

Cryptocercus wrighti n. sp. This species is characterized by the following chromosome number: male diploid complement: 45 (22 autosomal bivalents + X chromosome). Although females have not been examined for this study, they are expected to have a diploid complement of 46 (22 autosomal bivalents + 2 X chromosomes) (1).

This holotype of C. wrighti is characterized by the following sequence of the 12S rRNA and 16S rRNA genes. Diagnostic bases (=characters) that enable the discrimination of C. wrighti from one or more of the other Cryptocercus species are given in Table 4.

12S rRNA:

16S rRNA:

The geographical distribution of C. wrighti is shown in Figure 4 and Figure 5.  Based on sampling to date, the southern edge of its distribution begins southwest of Asheville, NC, in the southwestern corner of the state. Samples of this species have been collected  near the North Carolina - Tennessee border, southwestern North Carolina, and extreme southwestern Virginia . The northern most point of collection is Roan Mountain State Park in the northeastern corner of TN in the Cherokee National Forest. It remains to be determined whether any Cryptocercus occur in the area between the inferred geographical distributions of C. darwini and C. wrighti.

Etymology: This species is named in honor of the late Sewall Wright, the eminent population geneticist.

Holotype: Little Switzerland, NC.  Deposited in Kansas State University's Museum of  Entomological and Prairie Arthropod Research.

Based on the sampling to date, it appears that the three new species described here occur within a few kilometers of one another in extreme southwestern North Carolina ( Figure 5).  For example, the samples from Buck Creek Trail, NC (C. darwini) and Bear Trap Gap Overlook, NC (C. wrighti) were collected within approximately 15 kilometers of each other and yet they belong to two different evolutionary lineages. A fine scale sampling is required to determine the exact boundaries of the three new species in this region. Such sampling will also provide evidence for hybridization, if any, which may be occurring among these species.

These species descriptions are intended as part of the permanent, public, scientific record. Additional copies of this paper may be obtained from:


In this paper, we described three new species of Cryptocercus based on variation in DNA sequence, chromosome number and morphology.  These descriptions increases the number of Cryptocercus species worldwide from four to seven. North America now has five species of Cryptocercus, with four in the eastern United States (along the Appalachian and Allegheny Mountains) and one in northwestern United States. The four species that occur in the eastern United States are morphologically very similar to one another.  However, they vary significantly in chromosome number, DNA sequence of the mitochondrial ribosomal RNA genes, and are found within discrete geographic ranges.  It is not yet known whether the extensive variation in DNA sequence and chromosome number that we observed in Cryptocercus from the eastern United States is also characteristic of populations in the northwestern United States.


We thank Brian Forschler (University of Georgia) for valuable help and advice on collecting Cryptocercus. This study was supported by National Science Foundation Grant DEB-9806710 and Hatch Project H-497 to S.K. This is contribution number 99-239-J of the Kansas Agricultural Experiment Station.


  1. Kambhampati, S., Luykx, P., and Nalepa, C.A.  Evidence for sibling species in Cryptocercus punctulatus, the wood roach, from variation in mitochondrial DNA and karyotype. (1996) Heredity. 76, 485-496
  2. Nalepa, C.A., Byers, G.W., Bandi, C. and Sironi, M.  Description of Cryptocercus clevelandi from the Northwestern United States, molecular analysis of bacterial symbionts in its fat body and notes on biology, distribution and biogeography.  (1997) Ann. Entomol. Soc. Amer. 90, 416-424.
  3. Kambhampati, S., Black, W.C., IV and Rai, K.S.  Random amplified polymorphic DNA of mosquitoes: Techniques, applications and statistical analysis. (1992)  J. Med. Entomol. 29, 939-945.
  4. Kambhampati, S., and Smith, P.T.  PCR primers for the amplification of four insect mitochondrial gene fragments. (1995) Insect Molec. Biol. 4, 233-236.
  5. Farris, J.S.  A successive approximations approach to character weighting.  (1969)  Syst. Zool. 18, 374-385.
  6. Carpenter, J.M.  Choosing among multiply equally parsimonious trees. (1988)  Cladistics 4, 291-296.
  7. Carpenter, J.M.  Successive weighting, reliability and evidence. (1994) Cladistics 10, 215-220.
  8. Farris, J. S.  The retention index and the rescaled consistency index. (1989)  Cladistics 5, 417-417.
  9. Bremer, K.  Branch support and tree stability. (1994) Cladistics. 10, 295-304.
  10. Cleveland, L. R., Hall, S. R., Sanders, E. P., and Collier, J. The wood-feeding roach Cryptocercus its protozoa and the symbiosis between protozoa and the roach. (1934) Mem. of the Amer. Acad. of Arts and Sci. 17, 185-342.
  11. Hebard, M.  The Blattidae of North America north of the Mexican boundary. (1917) Mem. Amer. Entomol. Soc. 2,17.
  12. Bie-Bienko, G. Y.  Blattodea. In: Faune de l'URSS vol. 40, pp. 332-336. Institute of Zoology, Academy of Sciences URSS, Moscow.  (1950)
  13. Simon, C., Frati, F., Beckenbach, A., Crespi, B., Liu, H., and Flook, P. Evolution, weighting, and phylogenetic utility of mitochondrial gene sequences and a compilation of conserved polymerase chain reaction primers.  (1994) Ann. Entomol. Soc. Amer. 87, 651-701.
  14. Larson, A.  Speciation and morphological change. In: D. Otte and J.A. Endler (eds.) Speciation and its consequences. pp. 579-598. Sinauer Assoc. Sunderland, MA. (1989)
  15. Roth, L. M. Interspecific mating in Blattaria. (1970) Ann. Entomol. Soc. Amer. 63, 1282-1285.

ERRATUM - February 4, 2000
An uncredited statement was deleted from the text of Results and Discussion Section. Table I had two corrections made for the source of specimens from sample nos. 5 and 17.
© 1999 Epress Inc.