All the cultivars used in this study could be classified into five taxonomic groups [3] , hereafter referred to as Populations (Table 1). And they were taken from the Anshan nursery at the Wuhan Institute of Landscape Architecture, 3 - 5 young leaves of each cultivar were collected at random from three individuals on May 2021. All samples were enclosed in plastic bags and stored at –80˚C in the laboratory prior to DNA extraction.

Genomic DNA were extracted using the Rapid Plant Genome DNA Isolation Kit (B518231, Sangon Biotech, Shanghai, China) according to the production instruction. The DNA concentration was measured with a NanoDrop 2000 and then was stored at –20˚C. We collected 38 SSR primers from some studies on Subgenes Cerasus. These primers were shown in Supplementary Table S1. These primers were labeled with fluorescent dyes (FAM, HEX, or TAM). The amplified products were analyzed by polyacrylamide gel electrophoresis, and after gel imaging, selecting polymorphic primers according to the bands quality.

PCR was performed using 25 µL of reaction solution, the detail as follow: 0.5 µL 10 mM dNTP, 2.5 μL Taq buffer, 0.5 μL of each primer at 20 μmol·L⁻¹, 1 μL genomic DNA, 2.5 μL 25 mM MgCl₂, 1.0 U Taq DNA polymerase, and 17.8 µL double-distilled H₂O. Amplification was carried with the following parameters: an initial denaturation at 95˚C for 3 min, ten cycles of denaturation at 95˚C for 30 s, annealing at 60˚C for 35 s, and extension at 72˚C for 30 s. This was followed by 20 cycles of denaturation at 95˚C for 30 s, annealing at 60˚C for 30 s, and extension at 72˚C for 30 s, with a final extension at 72˚C for 6 min. The PCR products were detected with ABI 3730xl DNA Analyser (Applied Biosystems, USA).

By literatures, 38 pairs of SSR primers of flowering cherry or similar species were collected, and then 8 representative samples from the tested plant materials were used to PCR amplify for polymorphic primers selection. By gel imaging, the stable amplification, clear bands 13 polymorphism were pick out from 38 primers, The detail as Figure 1, these primers information also shown in Supplementary Table S1.

As Figure 2 showing, the genetic similarities range from 0.87 to 0.20, with a mean of 0.494. These cultivars could be divided into four groups at the genetic similarity value of 0.26; the groups were designated A, B, C, and D. Group A was consisted of seven cultivars, with genetic similarities between 0.87 and 0.47. Five of the seven cultivars belonged to P. subhirtella, and the other two belonged to P. yedoensis “Somei-yoshino” and P. pseudocerasus “Introsa”. There were21 cultivars in Group B, including 18 cultivars of P. serrulata and three of P. jamasakura. The genetic similarity valuein Group B ranged from 0.86 - 0.36, and were higher than Group A. This group was characterised by highly diverse flower colours, inflorescences, petal numbers, and flower types. Prunus serrulata “Sekiyama” and P. serrulata “Hongye” were found to be closely related, with the similarity value of 0.86. Group C comprised nine cultivars with genetic similarities ranging from 0.55 - 0.34; this group was mainly composed of cultivars and hybrid progenies of P. campanulata. Group D included P. conradinae, P. cerasoides var. rubea, and P. pseudocerasus “Keio-zakura”.

The stress value in the NMDS analysis was 0.22, indicating a poor fit according to Kruskal J B’s standard of division stress value [21] . In Figure 3(a), The linear fit was poor (R² = 0.775) but the non-metric fit was high (R² = 0.95), suggesting that non-metric analyses were appropriate. In Figure 3(b), the sizes of the sample points with bubbles were representative of the fitness value; larger bubbles indicated poorer fit. Figure 3(c) showed the clustering of the cultivars in the two-dimensional space of nmds1 and nmds2. As the cluster analysis, the NMDS

analysis indicated that the cultivars separated into four groups when K = 4. Group 1 included three accessions: P. conradinaes (10), P. cerasoides var. Rubea (1), and P. pseudocerasus “Keio-zakura” (23). Since these belong to three different species in Subgen. Cerasus, their spatial distributions were scattered in Figure 3(c). Group 2 included nine cultivars of bell flower cherry (P. campanulata). The Groups 3 and 4 cultivars were native to Japan and belonged to P.serrulata and P. subhirtella, respectively, indicating a narrow genetic background.

A total of 167 alleles were obtained from the 40 flowering cherry cultivars using the 13 SSR primers, with an average of 12.8 alleles per locus in Table 2. Among the locuses, the observed heterozygosity (H_o) and expected heterozygosity (H_e) ranged from 0.161 - 0.825 and 0.338 - 0.853, with means of 0.458 and 0.670, respectively. H_e was higher than H_o. The mean Shannon’s information index (I) and number of effective alleles (N_e) were 1.91 and 5.74, respectively. The number of migrants (Nm) ranged from 0.016 to 0.763, with a mean of 1.576. The fixation index (F) ranged from 0.712 to 2.631, with a mean of 0.318. Random breeding would produce F values close to zero, whereas larger positive values were indicative of inbreeding. The inbreeding coefficients within the sub-populations (F_ST) ranged from 0.087 to 0.252, with a mean of 0.157.

The population-level genetic diversity parameters were shown in Table 3. Among the populations, N_a and N_e were 3.923 - 11.232 and 2.700 - 4.814, with means of 6.062 and 4.326, respectively. H_o ranged from 0.410 to 0.453, and H_efrom 0.564 to 0.770, with means 0.458 and 0.670, respectively. H_o was higher than He in all populations. I value ranged from 1.146 to 1.924, with a mean of 1.417. Population 3, which originated from P. serrulata, had higher values for H_o, H_e, and I compared to other populations, indicating that this population had higher levels of genetic diversity and differentiation. The F values ranged from 0.228 to 0.397, with a mean of 0.321. F was expected to be close to zero in situations of random mating, or under Hardy-Weinberg equilibrium.

The F_ST value among the five populations was 0.036, suggesting that inter-population variability only accounted for 3.6% of the genetic variation in the samples, whereas the other 96.4% of the variation occurred within the populations. Thus, intra-population variation was the major variation in the samples. The AMOVA assessed variation both populations and individuals (Table 4); the results indicated that variation between populations accounted for 4%, the39% variation came from individuals and 57% (p < 0.05) came from intra-individuals. The results showed that the genetic variation in the 40 cultivars were mainly attributable to variation within individuals; In contrast, the genetic differentiation between the populations was very low.

A genetic structure analysis provided detailed information on the level of genome admixture between the populations. The results from Structure and Structure Harvester indicated that the ΔK was optimal at K = 4 (Figure 4). It followed that the optimal number of subpopulations was 4; that was to say, all of the cultivars could be divided into 4 subgroups, we referred to as Population 1-4. As Figure 5, the proportions of the 4 subpopulations were 0.148, 0.214, 0.452, and 0.186, respectively. The H_e values were 0.6635, 0.8553, 0.7589, and 0.6172,

and the F_ST values were 0.1660, 0.0015, 0.1543, and 0.2898, respectively. The results essentially reflected the phylogenetic relationships. Q value ≥ 0.6 indicated relatively pure lines, and Q values < 0.6 indicated admixture. In this study, the Q values for 38 samples were ≥0.6, showing that these lines were relatively pure. Only two individuals were Q values < 0.6, indicating that the two samples were admixed, had a highly diverse genetic composition or were of mixed origin (Figure 5).

The size of the fragments amplified using the 13 SSR primers ranged from 111 to 298 bp. These SSR primers PBBCT34, EMPAS02B, EMPA026, PSCHGMS1, and EMPA027 were found to be highly polymorphic and suitable for classifying ten cultivars at least; the PBBCT34, EMPAS02B, and EMPA026primers could classify 21, 26, and 30 cultivars, respectively. All samples could be classified adequately using a combination of these three primers. To allow for the addition of cultivars in the future, we integrated one additional primerEMPA027, into molecular ID code system (Supplementary Table S2). As an example, the Japanese flowering cherrycultivar P. × yedoensis “Somei-yoshino”, the sizes of the fragments obtained from the PBBCT34, EMPAS02B, EMPA026, and EMPA027 primer sets were 208/218 bp, 166/18 2bp, 206/212 bp, and 162/162 bp, respectively. According to Table 5, the corresponding fragment codes were 54/59, 33/41, 53/56, and 31/31, respectively. These were combined sequentially into a 16-bit string, which comprised the cultivar’s molecular ID code. In the above example, the final molecular ID code was 5459334153563131. The molecular ID code of 40 flowering cherry cultivars were shown in Table 5.

Cluster analysis indicated that our results were generally consistent with morphological classifications [3] . Genetic variation between cultivars within the same lines was low, indicating that these lines had relatively narrow genetic backgrounds. This was attributable to seed production and cross breeding. Conversely, the genetic differences between lines were substantial, with the high diversity due to their wide distribution and numerous cultivars. These results were consistent with the previous findings [10] [13] [22] .

NMDS analysis was used to assess the similarity of data between objects and illustrate the spatial relationships between them [3] . This method could sort, cluster and reduce dimensions. In this study, the close relationships of 40 samples were clearly and vividly illustrated in two-dimensional space. At K = 4, the cluster results were similar to UPGMA.

The results of the population structure analysis were consistent with those of the cluster analysis and the NMDS, with the 40 tested samples dividing into 4 groups. Each method had its own unique advantages; for instance, NMDS could reflect relationships between genetics and geographical distribution in two-dimensional space. However, for individuals with complex genetic backgrounds, the results of the different methods showed differences in clustering and population structure divisions, as demonstrated by the results of Haiwen Zhang [23] .

The genetic diversity among cultivars was found to be very high. Compared to similar species, the mean of Shannon’s information index (I = 1.417) was higher than those of P. mahaleb (I = 0.1720) [24] and P. serrulata (I = 0.939) [25] , but lower than that of P. pseudocerasus (I = 1.525) [26] . The high genetic diversity may be associated with out-crossing mating systems, wide geographic distributions, and large difference in climate and habitat conditions among the cultivars. The diversity of the cultivars originating from Prunus subg. Cerasus was also an important factor.

Population genetic structure analysis was an important method to insight into genetic relationships among cultivars, and SSR molecular marker technology could be used to determine the degree of genetic variation and differentiation at the population level [26] [27] [28] . Genetic differentiation was an important parameter of population genetic structure. According to Wright [29] , if F_ST < 0.25, genetic differentiation was low; in our study it was very low (F_ST = 0.036). The AMOVA indicated that only 4% of the genetic variation occurred among populations, whereas 39% occurred among individuals and 57% within individuals. Thus, genetic variation within the populations was the major variation. This was consistent with the results of Cai Yu Liang and Chen Jiao [26] [30] . Seed dispersal mechanisms had a significant effect on genetic differentiation among populations; because flowering cherries mainly relied on animal dispersal, the range of animal activities restricted dispersal [30] .

Nm affected the genetic differentiation of populations. Generally speaking, when Nm > 1, genetic drift within populations was impeded, preventing differentiation [31] . The Nm value in this study was 1.577. So, genetic variation among the populations was restricted, and most genetic variation was found within populations.

Three methods were typically used to construct plant molecular ID code systems. The first was to use RFAP, RAPD, or AFLP molecular marker technologies, and labeled them 1 or 0 based on the presence or absence of the DNA bands and then the molecular ID code was added to the binary string [32] [33] [34] in decimal or by converting it to decimal, generating a decimal molecular ID code [35] . The second method, developed in recent years, used a combination of SSR technology and capillary electrophoresis to measure the sizes of amplified fragments. The amplified fragments were then sorted and encoded, and several primers’ amplified fragment codes were combined to construct a molecular ID code [36] [37] . The third method was similar to the second method but produces a different type of ID code: the amplified fragments are encoded, and the resulting codes are combined in a sequence to form a molecular ID code [38] .

We used the third method to construct a molecular ID code. To improve the distinguishability of SSR primers, we used both amplified fragments to encode in each locus. Given the increasing number of cherry cultivars and breeds, it was important to allow for increasing in future. Thus, we added one additional primer in the construction of our molecular ID code system. In this study, we encoded 99 fragments of 102 to 298 bp. If this was not sufficient for the number of codes required, two English letters (aa~zz) could also be appended, allowing for an additional 676 fragments. Therefore, this system could meet future needs and also ensure the ongoing uniformity of the 16-bit code.