Interspecific Hybridization and Complete Mitochondrial Genome Analysis of Two Ghost Moth Species

Simple Summary The Chinese cordyceps is a valuable parasitic Ophiocordyceps sinensis fungus–Thitarodes/Hepialus larva complex. In view of culturing this complex, a method for the artificial rearing of the Thitarodes/Hepialus ghost moth hosts was established. Deterioration of the host insect population and low mummification rates in infected larvae constrain effective cultivation. Hybridization of Thitarodes/Hepialus populations may overcome this problem. Thitarodes shambalaensis and Thitarodes sp. were inbred or hybridized, and the biological parameters, larval sensitivity to the fungal infection and mitochondrial genomes of the resulting populations were investigated. Hybridization of T. shambalaensis and Thitarodes sp. allowed producing a new generation. One hybrid population (T. shambalaensis females mated with Thitarodes sp. males) showed increased population growth as compared with the parental Thitarodes sp. population. The sensitivity of the inbred larval populations to four fungal isolates of O. sinensis differed. The complete mitochondrial genomes of T. shambalaensis, Thitarodes sp. and the hybrid population were 15,612 bp, 15,389 bp and 15,496 bp in length, respectively. A + T-rich regions were variable in sizes and repetitive sequences. The hybrid population was located in the same clade with T. shambalaensis, implying the maternal inheritance of mitochondrial DNA. Abstract The Chinese cordyceps, a parasitic Ophiocordyceps sinensis fungus–Thitarodes/Hepialus larva complex, is a valuable biological resource endemic to the Tibetan Plateau. Protection of the Plateau environment and huge market demand make it necessary to culture this complex in an artificial system. A method for the large-scale artificial rearing of the Thitarodes/Hepialus insect host has been established. However, the deterioration of the insect rearing population and low mummification of the infected larvae by the fungus constrain effective commercial cultivation. Hybridization of Thitarodes/Hepialus populations may be needed to overcome this problem. The species T. shambalaensis (GG♂ × GG♀) and an undescribed Thitarodes species (SD♂ × SD♀) were inbred or hybridized to evaluate the biological parameters, larval sensitivity to the fungal infection and mitochondrial genomes of the resulting populations. The two parental Thitarodes species exhibited significant differences in adult fresh weights and body lengths but not in pupal emergence rates. Hybridization of T. shambalaensis and Thitarodes sp. allowed producing a new generation. The SD♂ × GG♀ population showed a higher population trend index than the SD♂ × SD♀ population, implying increased population growth compared with the male parent. The sensitivity of the inbred larval populations to four fungal isolates of O. sinensis also differed. This provides possibilities to create Thitarodes/Hepialus populations with increased growth potential for the improved artificial production of the insect hosts. The mitochondrial genomes of GG♂ × GG♀, SD♂ × SD♀ and SD♂ × GG♀ were 15,612 bp, 15,389 bp and 15,496 bp in length, with an A + T content of 80.92%, 82.35% and 80.87%, respectively. The A + T-rich region contains 787 bp with two 114 bp repetitive sequences, 554 bp without repetitive sequences and 673 bp without repetitive sequences in GG♂ × GG♀, SD♂ × SD♀ and SD♂ × GG♀, respectively. The hybrid population (SD♂ × GG♀) was located in the same clade with GG♂ × GG♀, based on the phylogenetic tree constructed by 13 PCGs, implying the maternal inheritance of mitochondrial DNA.


Introduction
The Chinese cordyceps (Ophiocordyceps sinensis fungus-Thitarodes larva complex) is a valued biological resource endemic to the Tibetan Plateau and widely used in medicinal treatments including fatigue, asthma, respiratory and kidney diseases and as health foods, especially in many Asian countries [1][2][3][4]. The annual yield of the natural Chinese cordyceps has decreased sharply in recent years due to over-exploitation and habitat degradation [5,6]. Due to the extremely limited wild resource and huge market demand, this wild fungusinsect complex costs USD 60,000-75,000 per kilogram [7][8][9][10][11]. Artificial cultivation of the Chinese cordyceps is necessary to protect this valuable bio-resource and to supply commercial trade [4,11].
The insect host species of O. sinensis belong to the family Hepialidae (Lepidoptera). Thitarodes armoricanus Oberthür was the first host insect of O. sinensis in China to be reported and biologically characterized [12]. Research on the diversity and taxonomy of Hepialidae expanded rapidly during the 1980s. Abundant diversity and phylogeographical structures for the host insects of O. sinensis have been reported [13,14]. Comparative phylogenetic analyses have suggested coevolutionary relationships between O. sinensis and its host insects [14,15]. It appears that most host insect species of the O. sinensis fungus have a very narrow distribution on the Tibetan Plateau, and host insect species might vary among different mountain ranges and even from different sides and habitats of the same mountain [16]. Different insect host species or strains show significant differences in morphology, biology and ecology [16][17][18][19][20]. Long-distance dispersal of host insects seems very limited due to the short lifespan of the adult and the heavy abdomens of females that carry eggs; thus most shared genotypes are between strains from adjacent geographic regions [14]. Given the complex and harsh ecological environment in the Tibetan Plateau, most Hepialidae species have a narrow-area distribution type [21]. Of ninety-one named Hepialidae insects spanning thirteen genera reported to be related to host insects of the O. sinensis fungus, fifty-seven are considered potential host species of the fungus and are distributed throughout the Tibetan Plateau [17]. However, the described insect host species or strains of O. sinensis have never been confirmed by a hybridization technique.
Artificial cultivation of the Chinese cordyceps in low-altitude areas is successful and contains three important milestones: mass production of the host insects, cultivation of effective O. sinensis fungus and formation of a fruiting body from the infected insect cadaver. The method for the large-scale artificial rearing of the Thitarodes/Hepialus insect hosts is established [11,19,20]. Culture parameters such as food, temperature, humidity and culturing time influence the efficiency of artificial cultivation of different insect species and stages [19,20]. The natural foods of Thitarodes/Hepialus larvae are tender roots and buds of plants of the Polygonaceae family such as Polygonum viviparum, P. sphaerostachyum and P. capitatum, as well as other plants in nine different families (e.g., Ranunculaceae, Juncaceae and Cyperaceae) [22]. The roots of Potentilla anserina and carrots (Daucus carota) are favorite larval foods for indoor cultivation [4,23]. Thitarodes species have long and unusual life cycles; it takes 263 to 494 days for Thitarodes jianchuanensis and 443 to 780 days for T. armoricanus to complete the life cycle, including egg, larval instars L1-L9, pupa and adult; the larvae can develop into pupae from the L7, L8 or L9 instar [20]. During the long life cycle and several successive generations in the culture rooms, the insects are usually prone to deterioration [19,24]. How to overcome this common phenomenon during insect host rearing is a key issue for the commercial production of the Chinese cordyceps.
Various O. sinensis strains from different locations in the Tibetan Plateau have been isolated [2,14,25] and cultured in solid media and liquid media to obtain conidia and blastospores [26]. Stable fruiting body production of O. sinensis with mature ascospores Insects 2021, 12, 1046 3 of 18 by artificial media without living insects has also been successfully realized [25]. Two efficient methods are used to infect the host larvae with the O. sinensis fungus: the larvae are routinely infected by the mature ascospores collected from the wild Chinese cordyceps or from the artificial fruiting bodies, and the larvae are injected into the hemocoel with the blastospores from the liquid culture containing maltose as a carbon source [26]. In the hemolymph of the larvae, the spindle blastospores exhibit a dimorphic developmental process [26] and may produce exponentially by budding growth and/or grow into elongate hyphal bodies (pseudohyphae) and hyphae by apical growth under the induction of unknown factors [26,27], as reported in the dimorphic fungi Candida albicans and Ustilago maydis [28] and the entomopathogenic fungus Metarhizium rileyi [29]. The living infected host larvae might harbor the spindle blastospores in the hemolymph for several months, contrary to other entomopathogenic fungi such as M. anisopliae and Beauveria bassiana which cause the death of their host larvae within a few days [30,31]. The slow mummification of the larvae post infection is an obstacle for cost-efficient production of Chinese cordyceps [4,8,11]. Selection of an O. sinensis fungus isolate with high mummifying potential and a Thitarodes insect host species or strain with high sensitivity to the fungal infection should contribute to overcoming this obstacle in the artificial production of the Chinese cordyceps.
Insight into the biological and molecular characters of the inbred and hybrid populations is elementary for the effective artificial cultivation and evolutionary analysis of these Thitarodes insects. In this study, the hybridization between T. shambalaensis and an undescribed Thitarodes species from two different locations in the Tibetan Plateau was demonstrated. The fitness parameters (such as the number of eggs per female, egg hatching rates, larval fresh weights, larval survival rates, female and male pupal ratios, population trend indexes), larval sensitivity to the fungal infection and mitochondrial genomes of the resulting inbred and hybrid populations were determined to evaluate the hybridization effects.

Morphological and Molecular Characteristics of Thitarodes Insect Populations
The pupae of two Thitarodes insect populations were, respectively, from the mountains in Gongga (referred to as GG♂× GG♀) ( The valve pattern of the male genitalia is an important characteristic for the morphological identification of Hepialidae insects [42,43]. The female and male Thitarodes pupae were differentiated by their genitalia. Briefly, in the last abdominal segment, females exhibit a long longitudinal suture linked to the previous abdominal segment without papillary structures, whereas males exhibit a short longitudinal suture between two papillary structures that is not linked to the previous abdominal segment [44]. The males of GG♂× GG♀ and SD♂× SD♀ populations were dissected to show the valve patterns in the laboratory. For the molecular identification of these Thitarodes populations, Cytochrome b and cox1 sequences were amplified with the primers CB1 (TATGTACTACCATGAGGACAAATATC) and CB2 (ATTACACCTCCTAATTTATTAGGAAT) [42,45] and LCO1490 (GGTCAACAAATCATAAA-GATATTGG) and HCO2198 (TAAACTTCAGGGTGACCAAAAAATCA) [46], respectively.

Inbred and Hybrid Thitarodes Populations
Four inbred and hybrid combinations (GG♂× GG♀, SD♂× SD♀, SD♂× GG♀, GG♂× SD♀) were created with 50 female and 75 male adults for each combination, but the population GG♂× SD♀ could not be established due to technical issues related to climatization of the culture room. Three replicates were set up for each combination. The male and female pupae were housed in cartons (L = 104 cm; W = 50 cm; H = 50 cm) with moist moss at 9-17 • C and 50-80% relative humidity. When the adults emerged, they were housed in small cylindric nets (D = 28 cm; H = 32 cm) to allow mating for 3-5 days. The collected eggs from the mated females were transferred to a culture room and maintained at 9-13 • C to establish the experimental populations in the Institute of Zoology, Dongguan (43 m above sea level), Guangdong Province, China.
To evaluate the development, survival, fertility and sensitivity to the fungal infection in the resulting Thitarodes populations, 600 eggs from each inbred or hybrid combination were surface-sterilized for 3 min with a solution containing 2.5 mL of 4 M NaOH, 0.5 mL of 12% NaOCl and 21.5 mL of distilled water [47], rinsed 3 times with sterile distilled water and placed in a sterile plastic container (L = 48 cm; W = 35 cm; H = 17 cm) containing 2 kg coconut peat (65% of water content) and 1 kg Potentilla anserina roots as food at 9-13 • C. For each hybridization combination, 30 containers were established. When the larvae reached the third instar, they were individualized into a plastic cup (D = 3.5 cm; H = 6.5 cm) with the same peat and food as above (15 g coconut peat and 15 g food for each cup) to avoid larval cannibalism [20]. Fresh food was added to each cup every 2 months to obtain 6th instar larvae (average fresh weight = 0.52 ± 0.03 g) for fungal infection by the injection method. The larval number in each container was recorded, and the average hatch rate was calculated. At the sample date (every 30 days), the survival rates, longevity, fresh weight, body length and sex proportion of pupae and adults and fecundity were recorded.

O. sinensis Fungal Isolates
KD, YN, XZ and QH fungal isolates of O. sinensis isolated from the fruiting bodies of wild Chinese cordyceps, respectively, from Sichuan, Yunnan, Tibet and Qinghai, China, were cultured on PPDA medium (liquid PPDA medium: 200 g potato extract, 20 g glucose, 10 g peptone, 1.5 g KH 2 PO 4 , 0.5 g MgSO 4 , 20 mg vitamin B 1 and 1000 mL distilled water; solid PPDA medium: 15% agar in liquid PPDA medium) at 13 • C. The fungal isolates were identified by using the amplified sequence from the internal transcribed spacer (ITS; ITS1-5.8S-ITS2) of the nuclear ribosomal DNA as described by [48]. The identified O. sinensis isolates were preserved at −80 • C in the Institute of Zoology, Guangdong Academy of Science, Guangzhou, China.
The fungal colonies cultured on the PPDA plates at 13 • C for 60 days were transferred to 250 mL flasks containing 150 mL liquid PM medium (200 g potato extract, 20 g maltose, 10 g peptone, 1.5 g KH 2 PO 4 , 0.5 g MgSO 4 , 20 mg vitamin B 1 and 1000 mL distilled water) [26]. The flasks were incubated on a 120 rpm shaker at 13 • C, the blastospores from the flasks were harvested after 50 days by using three layers of sterile lens papers to remove hyphae and large particles, and the filtered solution was centrifuged at 8000 rpm for 15 min at 10 • C. The harvested blastospores were re-suspended in sterile phosphate-buffered Insects 2021, 12, 1046 5 of 18 saline (PBS; pH 7.0) at a concentration of 3.0 × 10 6 blastospores per mL and kept at 4 • C for less than 3 days before use for larval infection.

Larval Infection of Inbred Populations by O. sinensis Isolates
The larvae from 2 inbred populations (GG♂× GG♀, SD♂× SD♀) were injected with KD, YN, XZ or QH fungal isolates of O. sinensis. An aliquot of 4 µL blastospore suspension containing 1.2 × 10 4 blastospores was injected into each 6th instar larva by a microinjection system (IM-31; Narishige, Tokyo, Japan). One hundred and eighty larvae were used for each replicate, and three replicates were set for each injection. Larvae injected with PBS buffer or without any injection were set as controls. The injected larvae were reared at 4 • C for one week and then transferred to a culture room at 13 • C. After 90 days, about 10 µL of hemolymph of each injected larva (6th instar) was sampled to confirm the presence of the growing blastospores stained by Calcofluor White (Sigma, Kanagawa, Japan) and observed by a fluorescence microscope (IX73; Olympus, Tokyo, Japan). The injected larvae were reared at 13 • C until the larvae became stiff and were coated with growing mycelia. The mummified larvae with head upward were then planted into soil of 55-60% humidity to induce the formation of stroma at 4 • C. The survival and mummification of the injected larvae were monthly checked. Data on larval infection of the hybrid populations could not be gathered due to an insufficient number of larvae available for fungal injection.

Analysis of the Mitochondrial Genomes
Three male adults from GG♂and SD♂and three larvae from SD♂× GG♀ in dry ice were used for mitochondrial genome sequencing by Shanghai BIOZERON Co., Ltd., with the routine method [41]. The nucleotide sequences of protein-coding genes (PCGs) from the annotated mitochondrial genomes were translated to protein sequences using the invertebrate mitochondrial code. For the base composition of the nucleotide sequences, the composition skewness was calculated as follows: [49]. Thirteen PCGs and two rRNA genes were inferred based on comparison with mitochondrial genomes of 10 previously sequenced Hepialidae species (T. damxungensis, T. gonggaensis, T. pui, T. renzhiensis, T. sejilaensis, Thitarodes sp., H. xiaojinensis, T. yunnanensis, Napialus hunanensis, Endoclita signifer). The location and secondary structures of the 22 tRNA were predicted by tRNAscan-SE (http://lowelab.ucsc.edu/ tRNAscan-SE/) (accessed on 13 November 2020). After the removal of the termination codon, the codon usage frequency and the first, second and third base use frequency of the codon were calculated using MEGA 7.0. The overlapping regions and intergenic spacers between genes were manually counted. The entire A + T-rich region was subjected to a search for the tandem repeats using the Tandem Repeats Finder program [50].
To construct the phylogenetic relationships within Hepialidae in Lepidoptera, 10 complete mitochondrial genomes of the above hepialid species were downloaded from GenBank. Drosophila melanogaster was used as an outgroup. A maximum likelihood (ML) tree was built in MEGA 7.0 using the nucleotide sequence of 13 PCGs, based on the "find best DNA/protein models (ML)". The "GTR + G" model was chosen for phylogenetic analysis because it produced the lowest values for both the BIC (Bayesian information criterion) and the AICc (corrected Akaike information criterion). The confidence values of the ML tree were evaluated via a bootstrap test with 1000 iterations.

Data Analysis
The data are expressed as means ± SE. The average survival rates, fresh weight, body length and sex proportions of pupae and adults at the sample time points were determined. In addition, a population trend index (I) was calculated with I = P II /P I , where P I = numbers of pupae in the parental generation, and P II = numbers of pupae in the next generation [20]. The data were analyzed with SPSS 21.0 (SPSS Inc., Chicago, IL, USA) to compare the differences among the treatments. Differences among means by Tukey's multiple-range test were considered significant at p < 0.05.

Morphological and Molecular Identification of Two Thitarodes Species
The SD♂× SD♀ population from Sichuan Province, China, was considered to be an undescribed Thitarodes species, by the phylogenetic analysis, although the valve pattern of the male genitalia of SD♂× SD♀ resembled that of Thitarodes kangdingensis ( Figure 1) (Prof. Zhiwen Zou, personal communication). The GG♂× GG♀ population also from Sichuan Province, China, was confirmed to be Thitarodes shambalaensis [43], based on the valve pattern of the male genitalia ( Figure 1) and the sequences of the cox1 fragment. The complete cox1 sequences of the two species were 1531 bp and were submitted to GenBank (accession numbers OK104111 and OK047724, respectively).
The data are expressed as means ± SE. The average survival rates, fresh weight, body length and sex proportions of pupae and adults at the sample time points were determined. In addition, a population trend index (I) was calculated with I = PII/PI, where PI = numbers of pupae in the parental generation, and PII = numbers of pupae in the next generation [20]. The data were analyzed with SPSS 21.0 (SPSS Inc., Chicago, IL, USA) to compare the differences among the treatments. Differences among means by Tukey's multiple-range test were considered significant at p < 0.05.

Morphological and Molecular Identification of Two Thitarodes Species
The SD♂ × SD♀ population from Sichuan Province, China, was considered to be an undescribed Thitarodes species, by the phylogenetic analysis, although the valve pattern of the male genitalia of SD♂ × SD♀ resembled that of Thitarodes kangdingensis ( Figure 1) (Prof. Zhiwen Zou, personal communication). The GG♂ × GG♀ population also from Sichuan Province, China, was confirmed to be Thitarodes shambalaensis [43], based on the valve pattern of the male genitalia ( Figure 1) and the sequences of the cox1 fragment. The complete cox1 sequences of the two species were 1531 bp and were submitted to GenBank (accession numbers OK104111 and OK047724, respectively).

Development from the Pupae to Next-Generation Pupae in Inbred and Hybrid Populations
The data on the fresh weight and body length of the pupae from the parental species were collected. The fresh weights were 0.88 ± 0.04 g and 0.67 ± 0.03 g for SD♂ × SD♀ and GG♂ × GG♀ female pupae and 0.59 ± 0.05 g and 0.45 ± 0.03 g for SD♂ × SD♀ and GG♂ × GG♀ male pupae, respectively. The body lengths were 2.73 ± 0.21 cm and 2.50 ± 0.05 cm for SD♂ × SD♀ and GG♂ × GG♀ female pupae and 2.46 ± 0.08 cm and 2.25 ± 0.04 cm for SD♂ × SD♀ and GG♂ × GG♀ male pupae, respectively ( Table 1). The pupae developed into adults in 30-45 days at 9-17 °C. No significant differences were found for the fresh weights and body lengths of the pupae between the two populations, except for the fresh weight of female pupae, which differed significantly between the populations ( Table 1). The ratios of females and males in the pupae (SD♂ × SD♀: 1.27 ± 0.16; GG♂ × GG♀: 1.04 ± 0.06) and adults (SD♂ × SD♀: 0.89 ± 0.06; GG♂ × GG♀: 0.87 ± 0.03) were also not significant.

Development from the Pupae to Next-Generation Pupae in Inbred and Hybrid Populations
The data on the fresh weight and body length of the pupae from the parental species were collected. The fresh weights were 0.88 ± 0.04 g and 0.67 ± 0.03 g for SD♂× SD♀ and GG♂× GG♀ female pupae and 0.59 ± 0.05 g and 0.45 ± 0.03 g for SD♂× SD♀ and GG♂× GG♀ male pupae, respectively. The body lengths were 2.73 ± 0.21 cm and 2.50 ± 0.05 cm for SD♂× SD♀ and GG♂× GG♀ female pupae and 2.46 ± 0.08 cm and 2.25 ± 0.04 cm for SD♂× SD♀ and GG♂× GG♀ male pupae, respectively ( Table 1). The pupae developed into adults in 30-45 days at 9-17 • C. No significant differences were found for the fresh weights and body lengths of the pupae between the two populations, except for the fresh weight of female pupae, which differed significantly between the populations ( Table 1). The ratios of females and males in the pupae (SD♂× SD♀: 1.27 ± 0.16; GG♂× GG♀: 1.04 ± 0.06) and adults (SD♂× SD♀: 0.89 ± 0.06; GG♂× GG♀: 0.87 ± 0.03) were also not significant. The emergence rates were 38.38% and 48.96% for SD♂× SD♀ and GG♂× GG♀ females and 43.57% and 57.58% for SD♂× SD♀ and GG♂× GG♀ males, respectively, showing no significant differences between the two parental insect populations ( Table 1).
The females and males of the GG♂× GG♀ population were mating all day and night, like those of SD♂× GG♀. However, those of SD♂× SD♀ usually mated in the evening and at night. The adults did not feed, and their life span usually lasted 5-7 days at 9-17 • C. Contrary to the clean eggs from GG♂× GG♀, the eggs from SD♂× SD♀ were coated with a sticky secretion. The average number of eggs per female for SD♂× SD♀ (512 ± 3) was significantly higher than that for the other populations, whereas the number of eggs in one milliliter (3976 ± 109) was higher and the weight of one thousand eggs (0.18 ± 0.01 g) was lower for SD♂× SD♀ compared with other populations, indicating the smaller egg size of SD♂× SD♀ (Table S1). The hatching rate for SD♂× GG♀ (12.62 ± 2.80%) was significantly lower than that for the other populations (Table S1). Thus, these two insect populations exhibited characteristic differences in mating behavior, egg size and the presence of a sticky Insects 2021, 12, 1046 7 of 18 secretion coating the eggs but showed no significant differences in the ratio of female and male pupae or in the pupal emergence rates. The emergence rates were 38.38% and 48.96% for SD♂ × SD♀ and GG♂ × GG♀ females and 43.57% and 57.58% for SD♂ × SD♀ and GG♂ × GG♀ males, respectively, showing no significant differences between the two parental insect populations (Table 1). The females and males of the GG♂ × GG♀ population were mating all day and night, like those of SD♂ × GG♀. However, those of SD♂ × SD♀ usually mated in the evening and at night. The adults did not feed, and their life span usually lasted 5-7 days at 9-17 °C.
The emergence rates were 38.38% and 48.96% for SD♂ × SD♀ and GG♂ × GG♀ females and 43.57% and 57.58% for SD♂ × SD♀ and GG♂ × GG♀ males, respectively, showing no significant differences between the two parental insect populations (Table 1). The females and males of the GG♂ × GG♀ population were mating all day and night, like those of SD♂ × GG♀. However, those of SD♂ × SD♀ usually mated in the evening and at night. The adults did not feed, and their life span usually lasted 5-7 days at 9-17 °C.

Males
The emergence rates were 38.38% and 48.96% for SD♂ × SD♀ and GG♂ × GG♀ females and 43.57% and 57.58% for SD♂ × SD♀ and GG♂ × GG♀ males, respectively, showing no significant differences between the two parental insect populations (Table 1). The females and males of the GG♂ × GG♀ population were mating all day and night, like those of SD♂ × GG♀. However, those of SD♂ × SD♀ usually mated in the evening and at night. The adults did not feed, and their life span usually lasted 5-7 days at 9-17 °C.
The emergence rates were 38.38% and 48.96% for SD♂ × SD♀ and GG♂ × GG♀ females and 43.57% and 57.58% for SD♂ × SD♀ and GG♂ × GG♀ males, respectively, showing no significant differences between the two parental insect populations (Table 1).

Strains
Thitarodes sp. The females and males of the GG♂ × GG♀ population were mating all day and night, like those of SD♂ × GG♀. However, those of SD♂ × SD♀ usually mated in the evening and at night. The adults did not feed, and their life span usually lasted 5-7 days at 9-17 °C.

T. shambalaensis p-Values
Note: the rows with the same letters indicated no significant differences (p > 0.05).
The GG♂× GG♀ and SD♂× SD♀ populations were hybridized in the culture room, and the larvae successfully became pupae in the resulting hybrid populations. The fresh weights of the larvae from the inbred and hybrid populations did not vary significantly with culture times in 12 months ( Figure S1). The survival rates of the larvae were stable at 80-100% in the first 10 months but decreased sharply after 11 months, especially in the SD♂× SD♀ and SD♂× GG♀ populations ( Figure S2), due to the increasing larval mortality before pupation. The larvae became pupae after 22 months. The ratios of the resulting female and male pupae were 0.28 ± 0.15 for SD♂× SD♀, 0.61 ± 0.05 for SD♂× GG♀ and 0.78 ± 0.03 for GG♂× GG♀, which were significantly different. Population trend index values were quite variable, with I = 0.01 for SD♂× SD♀, 0.32 for SD♂× GG♀ and 2.25 for GG♂× GG♀, indicating different proportions of pupal numbers in the previous generation over the next generation in the inbred and hybrid populations. The adults emerged from the pupae after 23 months. Unfortunately, the experiments were discontinued due to the emergence of too few adults caused by the high mortality (>70%) of the pupae. Nonetheless, the above results indicate that hybridization of Thitarodes sp. and T. shambalaensis allowed harvesting a next generation of adults at least from the SD♂× GG♀ population.

Larval Infection of Inbred Populations by O. sinensis Isolates
As shown in Table S2, the percentages of the larvae carrying the blastospores varied at 120 days post infection, from 20.37 ± 5.38% for the larvae of GG♂× GG♀ injected with fungal isolate XZ to 62.96 ± 1.96% for the larvae of GG♂× GG♀ injected with fungal isolate QH; for SD♂× SD♀, the mummification rate of the larvae containing fungal isolate XZ was significantly lower than that of those containing fungal isolates KD, QH and YN; for GG♂× GG♀, the mummification rate of the larvae containing fungal isolate XZ was significantly lower than that of those containing fungal isolates KD, QH and YN. For fungal isolates KD, XZ and YN, no significant differences in mummification rate were observed among the two infected larval populations (Table S2). Sixty and 90 days after infection, no significant differences in the percentages of the larvae carrying blastospores were found among both larval populations. Although the larval hemocoel was filled with growing blastospores after 90 days, the fresh weights of the larvae in both populations did not differ markedly (Table S3).

Mitochondrial Genome Analysis
Organization and base composition. The complete mitochondrial genomes of two inbred and one hybrid Thitarodes populations (SD♂× SD♀, GG♂× GG♀ and SD♂× GG♀) were a circular DNA molecule of 15,389 bp, 15,612 bp and 15,496 bp in length, respectively (accession number: MZ675586, MZ675587 and MZ675588) (Figure 2). Like most other metazoan mitochondrial genomes, each of three Thitarodes mitochondrial genomes contained 13 PCGs, 22 tRNAs, 2 rRNAs and a large non-coding control region. Among the 37 genes in each mitogenome, there were 9 PCGs and 14 tRNAs encoded in the heavy strand, while 4 PCGs, 8 tRNAs and 2 rRNAs were encoded in the light strand. The mitochondrial genome structure was compact. The gene order of the Thitarodes mitochondrial genomes was uniform (Figure 2).
The mitochondrial genome content of these three Thitarodes populations was A + T-biased, ranging from 80.87% (SD♂× GG♀) to 82.35% (SD♂× SD♀) ( Table 2). The A + T content of the SD♂× SD♀ mitochondrial genome was 82.35%, which was larger than that of the other two mitochondrial genomes. The AT skew in the forward strand of the SD♂× SD♀ mitochondrial genome was slightly positive (0.008), which was different from the other two mitochondrial genomes (0.020) ( Table 2). Likewise, The GC skew of the SD♂× SD♀ mitogenome (−0.182) was also obviously different from those of the other two (−0.231 to −0.234).
Protein-coding genes. The 13 PCGs in these mitochondrial genomes included 7 NADH dehydrogenase subunits (nad1-6, nad4L), 3 cytochrome c oxidase subunits (cox1-3), 2 ATPase subunits (atp6, atp8) and one cytochrome b gene (cytb). The lengths of the 13 PCGs in the mitochondrial genomes of SD♂× SD♀, GG♂× GG♀ and SD♂× GG♀ were 11,073, 11,067 and 11,067, respectively (Table 2). When the termination codons were excluded, the 13 PCGs in SD♂× SD♀, GG♂× GG♀ and SD♂× GG♀ were composed of 3680, 3678 and 3678 codons, respectively. These findings indicate a high degree of similarity in the PCG code number among the three mitochondrial genomes (Figure 3). The mitochondrial genome content of these three Thitarodes populations was A + Tbiased, ranging from 80.87% (SD♂ × GG♀) to 82.35% (SD♂ × SD♀) ( Table 2). The A + T content of the SD♂ × SD♀ mitochondrial genome was 82.35%, which was larger than that of the other two mitochondrial genomes. The AT skew in the forward strand of the SD♂ × SD♀ mitochondrial genome was slightly positive (0.008), which was different from the other two mitochondrial genomes (0.020) ( Table 2). Likewise, The GC skew of the SD♂ × SD♀ mitogenome (−0.182) was also obviously different from those of the other two (−0.231 to −0.234).   Protein-coding genes. The 13 PCGs in these mitochondrial genomes included 7 NADH dehydrogenase subunits (nad1-6, nad4L), 3 cytochrome c oxidase subunits (cox1-3), 2 ATPase subunits (atp6, atp8) and one cytochrome b gene (cytb). The lengths of the 13 PCGs in the mitochondrial genomes of SD♂ × SD♀, GG♂ × GG♀ and SD♂ × GG♀ were 11,073, 11,067 and 11,067, respectively (Table 2). When the termination codons were excluded, the 13 PCGs in SD♂ × SD♀, GG♂ × GG♀ and SD♂ × GG♀ were composed of 3680, 3678 and 3678 codons, respectively. These findings indicate a high degree of similarity in the PCG code number among the three mitochondrial genomes (Figure 3). The codon frequency analysis of the SD♂ × SD♀, GG♂ × GG♀ and SD♂ × GG♀ mitochondrial genomes showed that a total of 61 codons were used for transcription, with the   The codon frequency analysis of the SD♂× SD♀, GG♂× GG♀ and SD♂× GG♀ mitochondrial genomes showed that a total of 61 codons were used for transcription, with the absence of UAG (Figure 3). The GG♂× GG♀ mitochondrial genomes had 10 more codons than SD♂× SD♀. However, the most frequently used codon in the three mitochondrial genomes was UUA for Leu, followed by AUU for Ile (Figure 3).
The fraction of codons encoding the hydrophobic amino acids (Met, Trp, Phe, Val, Leu, Ile, Pro, Ala) in the mitochondrial genomes of SD♂× SD♀, GG♂× GG♀ and SD♂× GG♀ were 56.91%, 56.60% and 56.62%, respectively (Figure 3), reflecting the biased usage of A/T nucleotides and the hydrophobic nature of respiratory membrane complexes. The codon distribution patterns of the three compared mitochondrial genomes ( Figure 3) were consistent with the finding that Ile, Leu, Phe, Ser, Asn and Tyr are the six amino acids most frequently used, whereas Lys is rare in Hepialidae [51].
Transfer RNA and ribosomal RNA genes. The predicted cloverleaf structures for 22 tRNA genes are presented in Figure S3. For three mitochondrial genomes of SD♂× SD♀, GG♂× GG♀ and SD♂× GG♀, the rrnL (16S rRNA) and rrnS (12S rRNA) genes were identified, being 1355 bp, 1336 bp and 1336 bp and 777 bp, 778 bp and 778 bp in size, respectively, falling into the reported range for the Hepialidae (1324-1375 bp, 740-781 bp) [41] ( Table 2). The rrnL gene was located between trnL1 (TAG) and trnV (TAC), while rrnS was located between trnV (TAC) and the A + T-rich region ( Table 2 and Figure 2). The A + T percentages of rRNA in three mitochondrial genomes were 85.10% to 85.60%. These rRNA characteristics are consistent with those observed in other Lepidoptera [41]. Twenty-two tRNAs were encoded in two mitochondrial genomes of the GG♂× GG♀ and SD♂× GG♀ populations, ranging from 60 bp to 73 bp in size and spread across the entire genome. SD♂× SD♀ ranged from 61 bp to 71 bp. All tRNAs were shown to be folded into the expected clover-leaf secondary structure except for trnS1 (UCU), which lacks the dihydrouridine (DHU) loop ( Figure S3). This feature is common to most of the available lepidopteran mitochondrial genomes [52].
Non-coding and overlapping genes. The complete mitochondrial genomes of SD♂× SD♀, GG♂× GG♀ and SD♂× GG♀ were very compact with a total of 198, 190 and 188 non-coding bp dispersed among 20, 17 and 17 pairs of neighboring genes ranging from 1 to 42 bp, 1 to 45 bp and 1 to 45 bp, respectively ( Table 2).
The longest spacer sequence was located between nad5 and trnH. A 15 bp intergenic spacer located between the trnS2 and nad1 contained the "ATACTAA" motif, which is a common feature across lepidopteran insects [53,54], but in Hepialidae species, the noncoding region contained an "ATACTA" sequence followed by T or C (Figure 4). The results are consistent with the report from [40]. In addition, the complete mitochondrial genomes of SD♂× SD♀, GG♂× GG♀ and SD♂× GG♀ were 42, 24 and 24 bp overlapping nucleotides located in 7, 10 and 10 pairs of neighboring genes with a length from 1 to 25 bp. The longest overlapping nucleotides (25 bp) existed between trnL1 and rrnL. The atp8 and atp6 had 4 bp overlapping nucleotides ( Figure S4). Those seven nucleotides "ATGATAA" are a common feature across the lepidopteran mitochondrial genomes [50].  A + T-rich region. The length and A + T content of the A + T-rich regions were 554 bp and 91.70% in SD♂ × SD♀, 787 bp and 89.83% in GG♂ × GG♀ and 673 bp and 90.64% in SD♂ × GG♀ ( Table 2). The A + T-rich region was located between the rrnS and trnI genes. These repeat sequences accounted for some of the variations in mt genome length ( Table  2). The A + T-rich region of T. renzhiensis is the longest of all the sequenced Lepidoptera mitochondrial genomes; the shortest is 319 bp in O. lunifer [55]. This variation in length could be related to the number and lengths of tandem repeats in the control region [56]. SD♂ × SD♀ had no repeating sequences ( Table 2). Two repeated sequences (114 bp) in GG♂ × GG♀ were detected, each with one more base A. Compared with GG♂ × GG♀, SD♂ × GG♀ lacked a set of repeated sequences, and two repeated sequences were 113 bp and 114 bp, respectively, which were only one A base apart (Table 2). This sequence repetition is not unique. There are four repetitive sequences of 118 bp in length in T. sejilaensis, four 107 bp repeat sequences in A. yunnanensis, eight 113 bp repeats in T. renzhiensis [36], six 112 bp repeats in T. gonggaensis [39] and five 119 bp repeats in T. pui [37].

Phylogenetic Relationships and Taxonomic Relation
To confirm the evolutionary position of the host insects of O. sinensis, a phylogenetic tree of 11 species using published mitochondrial genomes (10 Hepialidae, 1 outgroup) and three genomes from the present study was constructed based on the concatenated nucleotides' alignment of 13 PCGs or each PCG by the ML method. SD♂ × SD♀ was classified into a separate clade and GG♂ × GG♀ and SD♂ × GG♀ into another separate clade ( Figure 5 and Figure S5). Thus, SD♂ × SD♀ and GG♂ × GG♀ were confirmed to be different species. SD♂ × SD♀ was considered to be an undescribed Thitarodes species according to the present database, and GG♂ × GG♀ was identified as T. shambalenensis by the cox1 fragment. The genetical characteristics of SD♂ × GG♀ were close to those of GG♀, GG♂. A + T-rich region. The length and A + T content of the A + T-rich regions were 554 bp and 91.70% in SD♂× SD♀, 787 bp and 89.83% in GG♂× GG♀ and 673 bp and 90.64% in SD♂× GG♀ ( Table 2). The A + T-rich region was located between the rrnS and trnI genes. These repeat sequences accounted for some of the variations in mt genome length ( Table 2). The A + T-rich region of T. renzhiensis is the longest of all the sequenced Lepidoptera mitochondrial genomes; the shortest is 319 bp in O. lunifer [55]. This variation in length could be related to the number and lengths of tandem repeats in the control region [56]. SD♂× SD♀ had no repeating sequences ( Table 2). Two repeated sequences (114 bp) in GG♂× GG♀ were detected, each with one more base A. Compared with GG♂× GG♀, SD♂× GG♀ lacked a set of repeated sequences, and two repeated sequences were 113 bp and 114 bp, respectively, which were only one A base apart ( Table 2). This sequence repetition is not unique. There are four repetitive sequences of 118 bp in length in T. sejilaensis, four 107 bp repeat sequences in A. yunnanensis, eight 113 bp repeats in T. renzhiensis [36], six 112 bp repeats in T. gonggaensis [39] and five 119 bp repeats in T. pui [37].

Phylogenetic Relationships and Taxonomic Relation
To confirm the evolutionary position of the host insects of O. sinensis, a phylogenetic tree of 11 species using published mitochondrial genomes (10 Hepialidae, 1 outgroup) and three genomes from the present study was constructed based on the concatenated nucleotides' alignment of 13 PCGs or each PCG by the ML method. SD♂× SD♀ was classified into a separate clade and GG♂× GG♀ and SD♂× GG♀ into another separate clade ( Figure 5 and Figure S5). Thus, SD♂× SD♀ and GG♂× GG♀ were confirmed to be different species. SD♂× SD♀ was considered to be an undescribed Thitarodes species according to the present database, and GG♂× GG♀ was identified as T. shambalenensis by the cox1 fragment. The genetical characteristics of SD♂× GG♀ were close to those of GG♀, GG♂. A + T-rich region. The length and A + T content of the A + T-rich regions were 554 bp and 91.70% in SD♂ × SD♀, 787 bp and 89.83% in GG♂ × GG♀ and 673 bp and 90.64% in SD♂ × GG♀ ( Table 2). The A + T-rich region was located between the rrnS and trnI genes These repeat sequences accounted for some of the variations in mt genome length (Tabl 2). The A + T-rich region of T. renzhiensis is the longest of all the sequenced Lepidopter mitochondrial genomes; the shortest is 319 bp in O. lunifer [55]. This variation in length could be related to the number and lengths of tandem repeats in the control region [56] SD♂ × SD♀ had no repeating sequences ( Table 2). Two repeated sequences (114 bp) in GG♂ × GG♀ were detected, each with one more base A. Compared with GG♂ × GG♀, SD♂ × GG♀ lacked a set of repeated sequences, and two repeated sequences were 113 bp and 114 bp, respectively, which were only one A base apart ( Table 2). This sequence repetition is not unique. There are four repetitive sequences of 118 bp in length in T. sejilaensis, fou 107 bp repeat sequences in A. yunnanensis, eight 113 bp repeats in T. renzhiensis [36], six 112 bp repeats in T. gonggaensis [39] and five 119 bp repeats in T. pui [37].

Phylogenetic Relationships and Taxonomic Relation
To confirm the evolutionary position of the host insects of O. sinensis, a phylogeneti tree of 11 species using published mitochondrial genomes (10 Hepialidae, 1 outgroup and three genomes from the present study was constructed based on the concatenated nucleotides' alignment of 13 PCGs or each PCG by the ML method. SD♂ × SD♀ was clas sified into a separate clade and GG♂ × GG♀ and SD♂ × GG♀ into another separate clad ( Figure 5 and Figure S5). Thus, SD♂ × SD♀ and GG♂ × GG♀ were confirmed to be differen species. SD♂ × SD♀ was considered to be an undescribed Thitarodes species according to the present database, and GG♂ × GG♀ was identified as T. shambalenensis by the cox1 frag ment. The genetical characteristics of SD♂ × GG♀ were close to those of GG♀, GG♂.