*2.2. Microspore Culture and Regeneration*

Microspore cultures were conducted according to Na et al. [18] with some modifications. To isolate the microspores, 30 flower buds of 2–3 mm size before flowering were selected and collected in gauze, and then surface sterilized with 1% sodium hypochlorite (NaOCl) for 15 min on a shaker followed by three rounds of 3-minute washing with sterile water. The surface-sterilized buds were then ground in mortar and pestles in 2–3 mL of NLN [19,20] medium supplemented with 13% sucrose, filtered through a 45-μm sieve, and suspended in 30 mL of NLN medium. Centrifugation was performed at 1000 rpm for 3 min, and when the centrifugation was completed, the supernatant was discarded in a clean bench. The same process was repeated three times by adding a new NLN medium. The separated microspores were suspended in 75 mL of NLN medium containing activated carbon at a density of 40,000 microspores to 1 mL and then divided into 2.5 mL petri dishes (60 mm × 15 mm) for incubation and sealed with parafilm and wraps. On average, the concentration and the volume were adjusted to include one flower bud per petri dish [21]. To the induce embryonic development of the microspores, they were treated for 48 hours in a dark condition of 30 ◦C and then maintained at 25 ◦C for about 2 weeks. Afterwards, the embryos were transferred to a light condition of 25 ◦C and shaken at 75–80 rpm to observe embryogenesis. Microspore-derived embryos were maintained in a shaker incubator up to about 7 mm in size and transferred to a half strength of Murashige-Skoog (MS) [22] solid medium containing 3% sucrose for plant regeneration. These counted embryos were defined as embryos from microspores collected at an early stage. The regeneration rate was calculated using the equation below.

Regeneration rate (%) = [The total number of induced embryos from microspores culture/30(mL) × 100,000 (microspores/mL)] × 100

### *2.3. Genotyping-by-Sequencing (GBS) Library and Illumina Sequencing*

A sequencing library was prepared according to the GBS protocol in [23] using two restriction endonucleases, *Nsi*l-HF and *MseI*, which were successfully used for constructing GBS libraries in our previous report [24]. For the construction of the GBS library, a set of 64 barcoded adapters was generated from complementary oligonucleotides with the *Nsi*l-HF (New England Biolabs, Ipswich, MA) overhang sequence and unique barcode of length 5–10 bp and they have been tested using two parental lines to determine the optimum conditions for the GBS analysis. The adapter was treated with 2000 U T4 DNA ligase (New England Biolabs, Ipswich, MA) for 2 h at 22 ◦C and maintained at 65 ◦C for 20 min to eliminate ligase activity. Barcoded DNA samples were collected into a 50 mL falcon tube and purified using a PCI (phenol-chloroform-isoamyl alcohol) extraction method. It was then size-selected with AMPure XP beads (Beckman Coulter, High Wycombe, UK) to exclude small fragments which are preferentially amplified by polymerase chain reaction (PCR). The product from PCR went through size selection again to be the final product for further processes. The concentrations of the GBS libraries were determined using a Qubit fluorimeter (Thermo Fisher Scientific, Grand Island, NY, USA). Subsequent sequencing was performed on an Illumina HiSeq 2500 platform. The sequencing data are deposited in the NCBI's SRA database (PRJNA610719).
