Origin and culture of rotifers
In this study, we used different clones from three B. asplanchnoidis populations: OHJ, Nakuru, and MNCHU (see Table 1). A clone is composed of the asexual descendants of one female, which has originally hatched from one resting egg. Members of the same clone are genetically identical, since they were produced asexually, while members of different clones (from the same population) differ from each other in much the same way than the individuals of an obligate sexual species. The OHJ population was sampled from Obere Halbjochlacke, a small alkaline pond near Neusiedlersee (Austria). Our Nakuru population derives from a sediment sample of lake Nakuru (Kenya), while the MNCHU population was originally sampled in Chuluutyn Tsagaan Nuur (Mongolia). We used 86 clones from OHJ population, 15 clones from Nakuru population, and two clones from MNCHU population. The smaller number of clones for Nakuru and MNCHU was due to limited amounts of resting eggs from these populations.
Table 1 Clones and genome size range
As mentioned above, clonal cultures were established from the hatchlings of individual resting eggs. Rotifers were cultured in F/2 medium (Guillard, 1983) at 16 ppt salinity and with Tetraselmis suecica algae as food source at ad libitum concentration (500–1000 cells μl−1). Continuous illumination was provided with daylight LED lamps (SunStrip, Econlux) at 30–40 µmol quanta m−2 s−1 for rotifers and 200 µmol quanta m−2 s−1 for algae. Clonal stock cultures were kept at 18 °C and were reinoculated once per week by transferring 20 asexual females to fresh culture medium provided in 20 ml petri dishes. Cross-mating experiments, egg development, and population growth assays were conducted at 23–24 °C.
DNA sequence markers
To obtain rotifer biomass for DNA extraction, clonal cultures with population densities of 10–100 individuals per ml were starved for 16 h, ensuring that rotifers completely emptied their guts of the food algae. Afterward, the washed rotifer biomass was resuspended in 70% ethanol and stored at −20°C. Genomic DNA was isolated using the DNeasy Blood & Tissue kit (Qiagen) according to the manufacturer’s instructions, except that DNA was eluted with 50 μl of 1× TE0.1 buffer (20 mM Tris–HCl, 0.1 mM EDTA, pH 8.0). Concentration and quality of DNA was determined using a NanoDrop spectrophotometer (Thermo Scientific), and the DNA samples were additionally run on a 1% agarose gel.
To confirm correct species assignment, we amplified a 661-bp region of the mitochondrial cytochrome c oxidase subunit I (COI) using primers LCO1490 and HCO2198 from Folmer et al. (1994) and a 543-bp segment of the ribosomal internal transcribed spacer 1 (ITS1) using primers III and VIII from Palumbi (1996). PCR reactions were carried out in 20 μl volumes using HotStarTaq Plus Master Mix Kit (Qiagen). Cycling parameters were as follows: one cycle: 95 °C for 5 min; 5 cycles: 94 °C for 40 s, 48 °C (or 45 °C for clone Nakuru8 from population Nakuru) for 40 s, 72 °C for 1 min; 35 cycles: 94 °C for 40 s, 51 °C for 40 s, 72 °C for 1 min; and 72 °C for 10 min. PCR products were purified using the QIAquick PCR Purification Kit (Qiagen) and were sequenced by a commercial sequencing service (Eurofins MWG Operon). COI and ITS1 sequences were aligned using BioEdit 7.2.5 (Hall, 1999). Sequence identities were calculated in BioEdit, and Blastn (http://blast.ncbi.nlm.nih.gov) searches were conducted to confirm the species identity of our clones. Pairwise genetic distances for ITS1 sequences were calculated in MEGA7 (Kumar et al., 2016) using the Maximum Composite Likelihood model (Tamura et al., 2004). All positions containing gaps and missing data were eliminated. There were a total of 539 positions for ITS1 and 661 for COI in the final datasets. New sequences were deposited in the GenBank database (accession numbers for COI: KU299078-KU299174; accession numbers for ITS1: KU299644-KU299740).
Cross-mating experiments
Cross-mating experiments were used to test for reproductive isolation between clones from the three populations (OHJ, Nakuru, MNCHU). The genome sizes of the parental clones ranged from 211 to 366 Mbp (Table 2). Genome size differences between mating pairs varied between 3 and 155 Mbp, and cross-mating combinations were the following: We crossed females with a rather small genome size from the OHJ population (OHJ22, 211 Mbp) with males with a similar genome size from lake Nakuru (Nakuru4, 214 Mbp), but also with males with larger genome sizes from Mongolia (MNCHU24, 326 Mbp; MNCHU008, 366 Mbp). Further, we crossed females with a comparably large genome size from the MNCHU population (MNCHU24, 326 Mbp) with males with smaller genome sizes from the OHJ population (OHJ22, 211 Mbp; OHJ7, 264 Mbp). We also tried to cross OHJ7 females with MNCHU24 males and MNCHU008 females with OHJ22 males, but these combinations were not successful (see results).
Table 2 Genome size differences of parental clones
In the cross-mating experiments, we used freshly hatched virgin females and males, which were harvested as eggs from dense rotifer cultures that had initiated sexual reproduction. Eggs were detached from the females by vigorously vortexing the rotifer culture in 50-ml Falcon tubes for 10 min. Crossings between clones were accomplished by placing 100 female eggs and 50 male eggs together into the same well of a 24-well plate filled with 750 µl of F/2 medium. After 24 h, when all viable eggs had hatched and animals had time to mate, females were transferred to new wells with fresh food suspension. Before these transfers, we checked briefly for mating behavior. The existence of male mating behavior was recorded, if we observed at least two males circling females within a 5-minute interval. To determine the male/female ratios and male population densities in the cross-mating experiments, we also scored the hatching rates of male and female eggs after the 24-h incubation. Upon fertilization, females were cultured for another 8–15 days to classify them as amictic (=asexual) or mictic (=sexual) and to allow fertilized mictic females to produce resting eggs. If females died without producing any eggs, we could not determine their reproductive mode and counted them as undetermined (category “no eggs” in Table 3). Finally, all resting eggs were stored at 7 °C in the dark for at least 2 weeks. To induce hatching, resting eggs were incubated with food suspension at 23 °C and high light intensities (200 µmol quanta m−2 s−1). Usually after 48 h, the first hatchlings started to emerge, and clonal cultures were initiated. For a maximum of 4 weeks, the remaining resting eggs were checked for hatching. All eggs were transferred to fresh medium at least twice a week.
Table 3 Summary of the cross-mating experiment
Inheritance of genome size
To determine the genome sizes of crossed sexual offspring of different B. asplanchnoidis populations and the genome sizes of their parents, we used the flow cytometric method described in Stelzer et al. (2011), which uses propidium iodide (PI) staining of nuclei. Briefly, clonal rotifer populations were grown from low to high population densities in 1-l flasks, which were aerated with sterile air through a glass tube. One day before biomass preparation, animals were collected from each culture using 60-µm nylon sieves, washed in filtered sea water (salinity: 12 ppt), and starved overnight. On the day of biomass preparation, an aliquot of 350 females was taken from each starved culture and was subjected to the flow cytometry protocol of Stelzer et al. (2011). Two modifications of the original protocol were made. First, for each batch of rotifer biomass, we used one head of a female Drosophila melanogaster (strain ISO-1, C-value: 0.18 pg according to Gregory, 2015) as an internal standard. This 350:1 ratio yields almost identical peak heights for rotifers and Drosophila during flow cytometric analyses. Second, the trypsin digestion step was done at 37 °C in a water bath, rather than at 20 °C as in the original protocol (Stelzer et al., 2011). We found that this modification increased the precision of our flow cytometric measurements, i.e., it resulted in a lower coefficient of variance. The actual measurements were done on an Attune NxT® acoustic focusing cytometer (Thermo Fisher) with an excitation wavelength of 561 nm and a 590–650 nm bandpass filter for detection of PI fluorescence. Flow cytometric data were analyzed using FlowJo software version 10.0.7r2 (FlowJo LLC). At least three replicate measurements were obtained for each rotifer clone, and usually these measurements were done on different days.
Fitness of ‘hybrid’ offspring
To assess the fitness of ‘hybrid’ offspring in comparison with their parents, we measured three parameters: embryonic development time of amictic eggs, population growth rate, and the ability to produce males. The duration of embryonic development was measured using an automated method described in detail in Stelzer (2016). Briefly, this system relies on time-lapse recording of up to 96 amictic eggs in a simple custom-built inverted microscope with a motorized aperture stage that accepts 96-well plates. Amictic eggs were stripped from females 0–30 min after they had been extruded, and hatching of each individual egg was monitored automatically in 30 min intervals. The temperature during preparation of the eggs and hatching incubations was set to exactly 23.5 °C.
Population growth rates were estimated from exponentially growing rotifer cultures (200 ml culture volume). These cultures had been inoculated at a density of 0.25 females ml−1 (=time T
0) and were sampled after three (T
1), four (T
2), and 5 days (T
3) by withdrawing 27-ml culture suspension, followed by fixation with 3 ml Formaline (37%). Population densities at T
1 and T
2 were estimated by counting the complete samples, and at T
3 by counting half of the sample (15 ml) using inverted microscopy. Since our clones differed in their investment into sexual reproduction, we followed the recommendations of Montero-Pau et al. (2014) and calculated the potential intrinsic growth rate, r
pot. Differences in the growth rates between parents and their crossed offspring were statistically analyzed using a general linear model with a hierarchical structure, with “clone” as a random nested variable. All calculations were done in the software package R version 3.2.1 (R Development Core Team, 2015).
The occurrence of sexual reproduction in crossed clones was checked over 3 weeks as part of the serial transfers of our stock cultures. At each weekly transfer, the old culture was closely examined for the presence/absence of males. These assays were intended to detect whether a clone is capable of producing males at all, rather than estimating its sexual propensity.
Amplified fragment length polymorphisms
Amplified fragment length polymorphisms (Vos et al., 1995) were used to characterize the genetic population structure of B. asplanchnoidis populations and to confirm successful cross-mating of clones. In total, we analyzed 127 DNA samples from 103 different B. asplanchnoidis clones (86 clones from OHJ, 15 clones from Nakuru, and two from MNCHU) and 11 ‘hybrid’ clones from lab crosses (six intrapopulation [OHJ × OHJ] and five interpopulation ‘hybrids’ [four OHJ × MNCHU, one OHJ × Nakuru]). We also included 13 biological replicates (from separate biomass preparations of the same clones) and a water sample as negative control.
In the restriction–ligation reactions, 100 ng of genomic DNA were digested and ligated to EcoRI- and MseI-adaptors in 11 μl volumes containing 1× CutSmart buffer, 1 mM ATP, 2.5 U MseI, 5 U EcoRI-HF, 1 WeissU T4 DNA ligase (all: New England Biolabs), 20 pmol MseI adaptor, and 2 pmol EcoRI-adaptor. Restriction–ligation reactions were carried out for 3 h at 37 °C followed by 17 °C overnight in a thermal cycler. Afterwards, 5 μl of the DNA were diluted 20-fold with 1× TE0.1 buffer. The 20 μl preselective amplification reactions contained 4 μl of the diluted DNA prepared by restriction–ligation, 1× Taq PCR Mastermix (Qiagen), 0.25 μM of each of the two preselective primers (MseI + C and EcoRI + 0), 1.25 mM additional MgCl2, and 4.5 μl RNase-free, deionized water (Qiagen). PCR amplification was carried out in a Mastercycler® nexus Gradient (Eppendorf) using the following program: 72 °C for 2 min; 20 cycles: 94 °C for 20 s, 56 °C for 30 s, 72 °C for 2 min; and 60 °C for 30 min. The product was diluted 20-fold with 1× TE0.1 buffer. The 10 μl selective PCR reaction contained 2 μl diluted PCR product, 0.25 μM MseI-primer, 0.05 μM fluorescent-labeled EcoRI-primer, 1.25 mM additional MgCl2, and 1× HotStar PCR Master Mix (Qiagen). The PCR conditions were as follows: 95 °C for 15 min; 10 cycles: 94 °C for 20 s, 66 °C–1°C/cycle for 30 s, 72 °C for 2 min, 20 cycles: 94 °C for 20 s, 56 °C for 30 s, 72 °C for 2 min; 60 °C for 30 min). We used eight different primer combinations for the selective amplification: M47xE24, M50xE25, M48xE13, M59xE14, M47xE12, M47xE25, M49xE13, and M61xE11 (Nomenclature according to KeyGene, http://www.keygene.com). Core sequences of the primers were EcoRI: 5′-Dye-GACTGCGTACCAATTC-NN-3′ and MseI: 5′-GATGAGTCCTGAGTAA-NNN-3′. EcoRI-Primers were labeled with the fluorescent dyes FAM, JOE, Atto550, and Atto565. The most suitable primer combinations were already chosen beforehand using the Selective Amplification Start-Up Kit for Small Plant Genomes (Applied Biosystems). For fragment analysis, 0.5 μl of fluorescent-labeled product was mixed with 10 μl of Hi-Di™ Formamide (Applied Biosystems), 10 μl H2O, and 0.1 μl LIZ-500 Size Standard (Applied Biosystems), denatured, and analyzed on an ABI 3730 capillary sequencer (Applied Biosystems) by a commercial service (Microsynth).
The fsa files generated during fragment analysis were converted to csv format using PeakScanner v1.0 (Applied Biosystems). optiFLP v1.57 (Arthofer et al., 2011) was used to identify optimum scoring parameters using the following settings: maximum peakheight—50, 200, 10 (minimum, maximum, stepwidth); maximum peakwidth—1 (fixed value); minimum peaksize—60, 130, 10; maximum peaksize—250, 400, 10; size tolerance range—0.5; minimum peak distance—off; minimum allele frequency—5, 15, 3; maximum allele frequency—80, 95, 3; Jaccard coefficient; unsupervised mode; lowest group nr.—2; highest group nr.—10; minimum number of profiles per group—2; and paraphyletic groups allowed. optiFLP retains the scoring parameters for the ten runs with highest global R; the run with the absolute maximum of global R was used for further analysis. If two or more runs shared maximum global R considering four decimal places, the run yielding the highest number of loci was selected. The result files of optiFLP were concatenated using tinyCAT v1.2 (Arthofer, 2010).
Dominant marker-based F and θ statistics were computed in Hickory v1.1 (Holsinger et al., 2002). We used a burnin of 50,000 generations, followed by 250,000 sampled generations with a thinning of 5 and the ‘Perform all Analyzes’-option. The deviance information criterion (DIC) was used to identify the best model, but, due to populations containing less than 10 individuals, also the model estimating θ without estimating f was considered.
SplitsTree v4.13.1 (Huson & Bryant, 2006) was used to calculate a Neighbor Network with following settings: character transformation—Jaccard; distances transformation—NeighborNet; splits transformation—equal angle. The resulting network visualization was directly exported from SplitsTree as pdf file.
Two Bayesian clustering approaches were used for detecting the true number of clusters (K) in our population samples: (i) Structure v2.3.3 (Pritchard et al., 2000) was used with default settings, 20,000 MCMC generations burnin, and 180,000 MCMC generations for data acquisition, with 10 repetitions for each K = [1, 7]. Evanno’s delta K algorithm as implemented in Structure Harvester v0.6.94 (Earl & von Holdt, 2011) was used to identify the best K. (ii) Clustering of individuals was performed in BAPS v5.3 (Corander et al., 2008) with default settings and 10 repetitions for each K = [1, 7]. The same software was used for admixture analysis of individuals based on mixture clustering using the default settings and 50 simulations from posterior allele frequencies. As the number of individuals per population differed largely, and Structure is sensitive to unequal population sizes (Kalinowski, 2011), we generated reduced datasets by randomly deleting 49, 73, 86, and 92 profiles of the OHJ population, with 4 replicates for each reduced set. These sets were analyzed by Structure and BAPS as described above, and optimum values for K were recorded. We also performed the unsupervised, iterative, non-Bayesian clustering method implemented in FLOCK v3.1 (Duchesne & Turgeon, 2009). Settings were 50 runs, 50 iterations per run, and random choice of samples for generating the initial partition. K = [2, 7] was searched, and the best value for K was determined as suggested in the manual of the software.