Mites and host plants
To establish laboratory populations, the collected mites were reared on leaves of Yoshino cherry, Prunus × yedoensis Matsum., which were placed on a water-soaked sponge in Petri dishes (9 cm in diameter) under constant climatic conditions (25±1℃, 60-70% relative humidity and16:8h light:dark photoperiod). We placed the leaves underside up and the perimeter was covered with water-soaked tissue paper. In winter, these populations were kept as diapause females, which were reared from eggs under 15±1°C and 8:16 light:dark photoperiod. Diapause females were put onto black paper and set into glass vials, which were kept in a refrigerator (ca. 5°C) for five months from December to April under darkness. We collected Tetranychus kanzawai Kishida from Japan (city: Miyakojima, latitude-longitude: 24°45'N-125°23'E, host plant: Benincasa hispida (Thunb.) Cogn., date: January 31, 2008), and used it as the outgroup of phylogenetic analysis of A. viennensis.
DNA preparation and sequencing
In measures of genetic distance among populations and the phylogenetic analyses, we used the cytochrome c oxidase subunit I gene (COI) of mitochondrial DNA (mtDNA). DNA was extracted from a single female mite from each population by using PrepMan Ultra Sample Preparation Reagent (Thermo Fisher Scientific Inc.). To amplify the fragment of mtCOI region, PCR was carried out using primers given in Table S2 [49, 50] in a 36 µl reaction mixture containing 0.5 ul of DNA sample, 3.6 µl of 10×Ex Taq buffer (20mM mg2+ plus, Takara Bio Inc.), 0.14 µl of TaKaRa Ex Taq (5U/µl, Takara Bio Inc.), 2.88 µl of dNTP mix (2.5mM each, Takara Bio Inc.), 0.72 µl of each primer (10pmol/ul each) and 27.44 µl of ddH2O. PCR cycling conditions were 3 min at 94°C, followed by 35 cycles of 1 min at 94°C, 1 min at 51°C and 1.5 min at 72°C, and a final extension at 72°C for 10 min. In some samples, the fragment was not amplified. Therefore, in these samples, we carried out PCR by decreasing the annealing temperature or increasing the number of cycles. PCR products were purified using MinElute PCR Purification Kit (QIAGEN). The purified products were sequenced using ABI BigDye Terminator ver. 3 Cycle Sequencing Kit (Applied Biosystems) and ABI3130xlGenetic Analyzer (Applied Biosystems).
Phylogenetic analyses and genetic distance measurements
Obtained sequences of the COI (618 bp, GenBank accession numbers: LC579422-LC579429) for A. viennensis and T. kanzawai and the COI sequences for A. quercivorus from previously published data (accession numbers: AB981238 and LC435686) [51] were aligned using CLUSTAL W in MEGA X [52]. A maximum likelihood (ML) tree of the aligned COI sequences was constructed with MEGA X [52]. As the substitution model for the ML tree, we used the Tamura 3-parameter model in which non-uniformity of evolutionary rates among sites is modeled by using a discrete Gamma distribution, because the model performed better than other models according to the Bayesian Information Criteria (BICs) in ML fits of 24 different nucleotide substitution models. Reliability of trees was evaluated by the bootstrap test (N=1,000).
Kimura 2-parameter genetic distances [53] were calculated among the populations using MEGA X [52]. We did not correct the data for their phylogenetic independence in the genetic distances as other papers did [28, 29, 38], since we also focus on asymmetries in reproductive incompatibility among the reciprocal combinations.
Endosymbiont infections
To detect the presence of Wolbachia, Cardinium, Spiroplasma and Rickettsia in the mites, we carried out PCR assay for these endosymbionts using primers given in Table S2 [54–58]. DNA was extracted from five female mites from each population, by homogenizing them in a 1.5 ml microtube with 18 µl of STE buffer (100mM NaCl, 10mM Tris-HCl, 1mH EDTA, pH 8.0) and 2 µl proteinase K, and incubating them at 55℃ for 30 min and 95℃ for 3 min. PCR was carried out in a 20 µl reaction mixture containing 1.0 µl of DNA sample, 2 µl of 10×NH4 Reaction buffer (Nippon Genetics Co., Ltd), 1 µl of 50mM MgCl2 Solution, 0.2 µl of BIOTAQ DNA Polymerase (5U/µl, Nippon Genetics Co., Ltd), 0.4 µl of dNTP mix (10mM each), 1 µl of each primer (10pmol/µl each) and 13.4 µl of ddH2O. PCR cycling conditions were 3 min at 95°C, followed by 36 cycles of 30 sec at 95°C, 30 sec at 52°C and 30 sec at 72°C, and a final extension at 72°C for 5 min. To be sure, we carried out this check for endosymbiont infections twice by using the same DNA template from each population (i.e., two technical replicates per DNA sample). We used DNA of Wolbachia-infected Panonychus mori Yokoyama (Toyama, voucher specimen no. 665), Cardinium-infected Tetranychus urticae (Koch) (red form, Nagano, no. 171), Spiroplasma-infected Tetranychus truncatus Ehara (Inner Mongolia, no. 199) and Rickettsia-infected Nephotettrix cincticeps (Uhler) [59] as positive controls of Wolbachia, Cardinium, Spiroplasma and Rickettsia infection, respectively. Distilled water was used as negative control.
Cross experiments
We carried out cross experiment among the seven populations of A. viennensis in all combinations and both directions (42 combinations). As controls, intra-population crosses were carried out in each population (7 controls). The number of replicates in each cross combination ranged between 12 and 30 (Table S1). We also carried out 16 different backcrosses by using F1 hybrids obtained from the following cross experiments: F (female) × T (male), F (female) × I (male), F (female) × J (male), T (female) × F (male), T (female) × I (male), I (female) × F (male), I (female) × T (male), CIM (female) × CN (male), CIM (female) × K (male), CN (female) × CIM (male), CN (female) × K (male), K (female) × CIM (male), K (female) × CN (male), J (female) × F (male), J (female) × CIM (male) and J (female) × CN (male). In the other 26 crosses we obtained few hybrid females, therefore we did not carry out backcrosses with those. The number of replicates in each backcross combination was 10 to 32 (Table S1).
The cross experiments were carried out under the same conditions as the mite rearings. A leaf of Yoshino cherry was placed onto wet sponge in a Petri dish (9 cm in diameter) in the mite rearing. To make the area unified, a square (4 × 4 cm) was created by using strings of water-soaked tissue paper. A female in the last molt before adulthood (teleiochrysalis stage) and an adult male were collected from the mite culture, and placed on the leaf arena. The female and male were allowed to mate and oviposit for five days, then they were removed from the leaf arena. The number of eggs, the offspring survival and the gender of offspring were checked and recorded.
Statistical analyses
Analyses were carried out with the statistical package R version 4.6.2 [60]. We analyzed the ratio of male offspring to egg (#sons / #eggs), offspring mortality among diploid offspring [(#unhatched-eggs + #dead juveniles) / (#eggs - #sons)] and the sum of these two values by values obtained by subtracting viable diploid offspring ratio from [1 - (#daughter / #eggs)] by using generalized linear models (GLMs) with genetic distance, female population and their interaction on these three variables. We applied a quasibinomial distribution as the error distribution to account for overdispersion. For models where the interaction did not have a significant effect, we reanalyzed the effects of genetic distance and female population by removing the interaction term. For models where the interaction did have a significant effect, we reanalyzed the effect of genetic distance in each female population separately. We analyzed the relationship between the fraction of dead offspring in the backcrosses and genetic distance between parent populations by using a GLM. We applied a quasibinomial distribution as the error distribution to account for overdispersion. In the analysis, we included intra-population crosses as the controls. To estimate genetic distance for which reproductive barrier is nearly complete, we reconstructed the quasibinomial GLMs only with genetic distance, and calculated the genetic distances for 99.0% and 99.9% reproductive barriers complete by using the models. In these analyses, we used the R packages stats and MASS [60, 61].