Assessment of the Breeding Potential of a Set of Genotypes Selected from a Natural Population of Akebia trifoliata (Three–Leaf Akebia)

: Akebia trifoliata (three-leaf akebia) has long been used as a medicinal herb and has the potential to be used in diverse ways, especially as a fruit crop. However, efforts to domesticate and cultivate new varieties for commercial use are only in their infancy. Here, we evaluated the genetic diversity of 29 genotypes, which were previously selected from a natural population consisting of 1447 genotypes and exhibiting high resistance to fungal diseases and a smooth peel of A. trifoliata using 85 genome-speciﬁc single sequence repeat (SSR) markers. We also characterized variation in 19 phenotypic traits and nutritional components. Large variation in phenotypic traits and nutritional components was observed, especially in vitamin C, seed/pulp, and fruit color. Correlation analyses revealed that many phenotypic traits and nutritional components were signiﬁcantly correlated. A principal component analysis identiﬁed ﬁve principal components, which explained 83.2% of the total variation in the data. The results of the SSR analysis revealed that 80 of the 85 SSR markers were polymorphic; the total number of alleles ampliﬁed was 532. The expected heterozygosity was 0.672, and Shannon’s information index was 1.328. A Ward dendrogram and unweighted pair group method with arithmetic mean dendrogram revealed high diversity among the 29 genotypes and suggested that the measured morphological and nutritional traits were genetically independent of disease resistance and texture traits, as well as SSR marker loci. Finally, our results suggest that additional rounds of selection from the selected population, despite its small size, could be effective for the development of new A. trifoliata fruit cultivars.


Introduction
Akebia trifoliata (three-leaf akebia) is a woody perennial climbing vine in the family Lardizabalaceae [1] that is used in traditional Chinese medicine [2,3]. It also has the potential to be bred as a new edible fruit [4], a future oil crop [5], and ornamental plant [6]. A. trifoliata is a popular medicinal plant that has been widely used in East Asia for at least 2000 years, and it was recently listed in the national pharmacopeias of China [7], Japan [8], and Korea [9]. A. trifoliata has also received much attention from researchers for the high content and quality of oil in its seeds, and some studies have suggested that A. trifoliata could be used as an oil crop in the future [10,11]. A. trifoliata has been cultivated as an ornamental plant for its bright flowers, diverse shapes, colorful fruit, elegant vines, and evergreen leaves in Japan [12], the United States [13], and China [6]. In addition, the fruit of A. trifoliata has been used as an ingredient in skin care products as well as a source of pectin, vinegar, and industrial raw materials [5,14].
in 2018 [31], only 239 genotypes showed intermediate to strong resistance to powdery mildew, leaf spot, and anthracnose. Only 29 of these 239 genotypes were saved following the second round of selection on texture characters during fruit ripening from early August to October in 2018. All cloned lines of the 29 genotypes were planted in a randomized complete block design (2 m × 3 m spacing) with three replicates, each of which had 10 individuals of each genotype.

Phenotypic Evaluation
Three ripe fruits from each individual and three individuals from each replicate were randomly selected. The mean of nine fruits represented the observed or measured value of a given genotype in a replicate, and the mean of three replicates was used to represent the observed or measured value of the phenotypic traits of a given genotype. Eleven phenotypic traits including ripening date (GP), fruit color (FC), fruit shape (FS), fruit weight (FW), peel weight (PW), seed weight (SW), pericarp thickness (PT), longitudinal diameter (LD), transverse diameter (TD), proportion of seeds (Pos), and edible proportion (EP) were assessed. FW, PW, and SW were weighed using an electronic balance with 0.01 g accuracy, and PT, LD, and TD were measured using digital calipers with 0.1 mm accuracy [32]. Formulas for Pos and EP were as follows: Pos = SW/FW × 100% and EP = (FW − PW − SW)/FW × 100%. The fruit developmental stage of each genotype was also recorded.

Analysis of Nutritional Components
The main nutritional components, including reducing sugar (RS), protein (Pro), vitamin C (VC), total flavonoids (TF), titratable acid (TA), total free amino acids (TFAAs), and total hydrolyzed amino acids (THAAs), in the fruit of 29 genotypes were measured to determine the nutritional value of A. trifoliata as a fruit crop. The RS and TA content in 2 g of A. trifoliata pulp was measured following previously described methods [33]. The classical protocol recommended by Lee et al. [34] and the distillation micro-Kjeldahl

Phenotypic Evaluation
Three ripe fruits from each individual and three individuals from each replicate were randomly selected. The mean of nine fruits represented the observed or measured value of a given genotype in a replicate, and the mean of three replicates was used to represent the observed or measured value of the phenotypic traits of a given genotype. Eleven phenotypic traits including ripening date (GP), fruit color (FC), fruit shape (FS), fruit weight (FW), peel weight (PW), seed weight (SW), pericarp thickness (PT), longitudinal diameter (LD), transverse diameter (TD), proportion of seeds (Pos), and edible proportion (EP) were assessed. FW, PW, and SW were weighed using an electronic balance with 0.01 g accuracy, and PT, LD, and TD were measured using digital calipers with 0.1 mm accuracy [32]. Formulas for Pos and EP were as follows: Pos = SW/FW × 100% and EP = (FW − PW − SW)/FW × 100%. The fruit developmental stage of each genotype was also recorded.

Analysis of Nutritional Components
The main nutritional components, including reducing sugar (RS), protein (Pro), vitamin C (VC), total flavonoids (TF), titratable acid (TA), total free amino acids (TFAAs), and total hydrolyzed amino acids (THAAs), in the fruit of 29 genotypes were measured to determine the nutritional value of A. trifoliata as a fruit crop. The RS and TA content in 2 g of A. trifoliata pulp was measured following previously described methods [33]. The classical protocol recommended by Lee et al. [34] and the distillation micro-Kjeldahl method described by Oyeleke [35] were used to determine the content of VC and Pro, respectively.

TF Content
A mixture of 2 g of pulp with 10 mL of 70% aqueous ethanol solution was sonicated for 30 min (20 kHz, KQ-300DE, ShuMei, Shanghai, China). The liquid was then placed into a 50 mL volumetric flask, and the process was repeated three times; ethanol solution was added until the total volume was 50 mL. Next, 10 mL of supernatant liquid was transferred into a 25 mL volumetric flask, and 2 mL of 5% NaNO 2 , 2 mL of 10% AlCl 3 , and 6 mL of 1 M NaOH were sequentially added at an interval of 6 min; 70% aqueous ethanol solution was added until the total volume was 25 mL. After 10 min, 2 mL of solution from the 25 mL volumetric flask was used to detect the absorbance at 510 nm using a spectrophotometer (UV-2401, Shimadzu Co., Tokyo, Japan); rutin was used as the standard to make the calibration curve.

TFAA Content
To determine the TFAA content, approximately 2 g of homogenized pulp was mixed with 1 mL of 15% salicylsulfonic acid and then left to stand for 1 h at room temperature. The TFAAs were separated using an amino acid analyzer (L-8900; Hitachi Co., Tokyo, Japan) equipped with a chromatographic column (855-4507; Hitachi Co., Japan) and operated with a pH-dependent lithium buffer gradient elution procedure at wavelengths of 570 and 440 nm for 148 min. To determine the THAA content, approximately 2 g of homogenized pulp was mixed with 10 mL of 6.0 mol L −1 HCl in a glass tube at 110 • C for 24 h for hydrolysis. After being cooled, the mixture was filtered, and deionized water was added until the total volume was 25 mL. Next, 1 mL of filtrate was evaporated to dryness, and then the dried filtrate was moistened with 1 mL of deionized water and filtered with a 0.45 µm Millipore filter with a pH-dependent sodium buffer gradient elution procedure at a wavelength of 550 nm for 100 min. Seventeen amino acids (MembraPure, Berlin, Germany) were used as standard samples.

Evaluation of SSR Markers
Genomic DNA was extracted from 0.2 g leaf samples from the parent plant using the modified cetyltrimethylammonium bromide method [36]. A total of 85 pairs of genomespecific SSR markers were developed based on the 150 bp sequences of the regions flanking the SSR loci using two PERL scripts 'p3_in.pl' and 'p3_out.pl' provided by the MISA package; these markers were then used for genotyping. Polymerase chain reaction (PCR) was carried out in a PTC-200 thermal cycler (MJ Research, Waltham, MA, USA) with the following program: 3 min at 95 • C; 34 cycles of 15 s at 95 • C, 15 s at the annealing temperature, and 2 min 30 s at 72 • C; and a final step of 5 min at 72 • C [37]. The PCRamplified products were run on an 8% denaturing polyacrylamide gel with the Sequi-Gen GT System (Bio-Rad, Hercules, CA, USA) at constant power (110 V) for 1.5 h. Silver staining was conducted following a previously published protocol [38], and the 50 bp DNA ladder was used as the size standard (Promega, Madison, WI, USA).

Data Analysis
The means of the traits were calculated and used for correlation analyses and principal component analysis (PCA), which were conducted in SPSS software [39]. The minimum value, maximum value, standard deviation (SD), and coefficient of variation (CV; SD/mean × 100) were calculated for the measured traits. Shannon's information index (I) was used as a variability index. A cluster analysis using Ward's method was conducted in PAST statistics software [40] to characterize patterns of variation among genotypes; a Ward dendrogram based on all the variables measured indicated morphological similarities among genotypes. A scatter plot of principal component 1 (PC1) vs. PC2 was created using PAST statistics software. The polymorphic information content (PIC), number of observed alleles (Na), number of effective alleles (Ne), Shannon's information index (I), observed heterozygosity (Ho), expected heterozygosity (He), and Nei's genetic distance (NGD) were calculated using PopGene32 and GenALEx6.41 software [41,42].  Table 1. There were pronounced differences in many traits among the 29 genotypes; I for 10 out of 12 phenotypic traits was greater than 2, although half of the corresponding CVs of these 10 traits were less than 20%. The other two traits (FC and FS) had CVs greater than 40%, but I of FC and FS was less than 2 ( Table 1). The variation in FW, PW, SW, and EP was high because their values of I were greater than 2, and their CVs were also greater than 20%.

Images of
using PAST statistics software. The polymorphic information content (PIC), number of observed alleles (Na), number of effective alleles (Ne), Shannon's information index (I), observed heterozygosity (Ho), expected heterozygosity (He), and Nei's genetic distance (NGD) were calculated using PopGene32 and GenALEx6.41 software [41,42].  Table 1. There were pronounced differences in many traits among the 29 genotypes; I for 10 out of 12 phenotypic traits was greater than 2, although half of the corresponding CVs of these 10 traits were less than 20%. The other two traits (FC and FS) had CVs greater than 40%, but I of FC and FS was less than 2 ( Table 1). The variation in FW, PW, SW, and EP was high because their values of I were greater than 2, and their CVs were also greater than 20%. Analysis of the distributions of FW, PW, SW, and EP revealed that only the distribution of EP was normal (p = 0.202) ( Figure 2b). Among the 29 genotypes, the mean FW was 237.7 g (CV = 23%, I = 2.54), and it ranged from 135.2 g to 446.9 g; the mean EP was 21.4 (CV = 22%, I = 2.73), and it ranged from 10.7% to 29.5%.

Diversity of Nutritional Components
All seven indexes indicated that there was large variation in nutritional components among the 29 genotypes (Table 1). I was greater than 2 for all components except TF. Phenotypic diversity for three important nutritional components-VC, RS, and TFAAs-of fruit was high based on the CVs greater than 20% and I values greater than 2. Specifically, the mean VC was 18.4 mg/100 g (CV = 60.3%, I = 2.13), and it ranged from 6.72 to 54.4 mg/100 g; the mean RS was 13 g/100 g (CV = 21.5%, I = 2.82), and it ranged from 8.71 to 18.2 g/100 g; the mean TFAA content was 706.4 ng/mg (CV = 40.7%, I = 2.57), and it ranged from 312.4 to 1443.1 ng/mg. Although the I for TF was less than 1, TF had the highest CV (151.6%), and it ranged from 1.0 to 29.1 mg/g, with a mean of 3.55 mg/g. The TA content was generally low in A. trifoliata fruits but varied greatly among the 29 genotypes (Table  1). Among these four important nutritional characters, only the distribution of the RS data was normal (p = 0.146) (Supplemental Figure S1). Analysis of the distributions of FW, PW, SW, and EP revealed that only the distribution of EP was normal (p = 0.202) ( Figure 2b). Among the 29 genotypes, the mean FW was 237.7 g (CV = 23%, I = 2.54), and it ranged from 135.2 g to 446.9 g; the mean EP was 21.4 (CV = 22%, I = 2.73), and it ranged from 10.7% to 29.5%.

Diversity of Nutritional Components
All seven indexes indicated that there was large variation in nutritional components among the 29 genotypes (Table 1). I was greater than 2 for all components except TF. Phenotypic diversity for three important nutritional components-VC, RS, and TFAAs-of fruit was high based on the CVs greater than 20% and I values greater than 2. Specifically, the mean VC was 18.4 mg/100 g (CV = 60.3%, I = 2.13), and it ranged from 6.72 to 54.4 mg/100 g; the mean RS was 13 g/100 g (CV = 21.5%, I = 2.82), and it ranged from 8.71 to 18.2 g/100 g; the mean TFAA content was 706.4 ng/mg (CV = 40.7%, I = 2.57), and it ranged from 312.4 to 1443.1 ng/mg. Although the I for TF was less than 1, TF had the highest CV (151.6%), and it ranged from 1.0 to 29.1 mg/g, with a mean of 3.55 mg/g. The TA content was generally low in A. trifoliata fruits but varied greatly among the 29 genotypes (Table 1). Among these four important nutritional characters, only the distribution of the RS data was normal (p = 0.146) (Supplementary Figure S1).

Relationships among Morphological and Nutritional Traits
Out of a total of 120 correlations among 16 morphological and nutritional traits assessed, 16 correlations were significant (p < 0.001), and 14 of these correlations were positive (Table 2). There were 11, 1, and 4 significant coefficients for relationships between two morphological traits (36 total correlations), between two nutritional traits (21 total correlations), and between one morphological and one nutritional trait (63 total correlations), respectively. There were 11 significant coefficients among the nine morphological traits, which involved FW, PW, LD, TD, and SW. For example, FW was significantly positively correlated with PW (r = 0.96), LD (r = 0.92), TD (r = 0.76), and SW (r = 0.69). There was only one significant coefficient among the seven nutritional traits. There were four significant coefficients among the phenotypic and nutritional traits, which involved TD, LD, EP, VC, and TF. EP was significantly negatively correlated with both S/P (r = −0.79) and TFAAs (r = −0.57). VC were significantly positively correlated with LD (r = 0.64); TF were significantly positively correlated with TD (r = 0.61) and VC (r = 0.63); TFAAs were significantly positively correlated with S/p (r = 0.56). There were no significant correlations between TA and other traits as well as THAAs and other traits, except Pos.

Principal Component Analysis
According to the scree plot ( Figure S2), there were five principal components (PCs) generated by the PCA, which explained 83.2% of the total variance (Table 3). PC1 explained 35.9% of the total variance, and FW, PW, LD, TD, and SD were the traits that loaded most heavily on PC1. PC2 explained 19.2% of the total variance, and EP, VC, RS, and TFAAs loaded most heavily on PC2. PC3 accounted for 11.9% of the total variance, and traits loading most heavily on PC3 included PT, Pro, S/P, and THAAs. PC4 and PC5 explained 8.70% and 7.43% of the total variance and correlated most strongly with TF and TA, respectively. The characteristics that loaded most heavily on the first three PCs showed the greatest variation among the 29 genotypes.
A two-dimensional scatter plot of PC1 vs. PC2 is shown in Figure 3. PC1 represented a size and shape axis, and PC2 revealed a contrast between nutritional quality and palatability (Table 3).

SSR Analysis
A total of 85 pairs of newly developed SSR primers based on the published A. trifoliata genome [43] (NCBI database under BioProject ID PRJNA671772) were used to determine the genetic diversity of the 29 genotypes. All 85 pairs of SSR primers produced clear bands in all samples; 82 of the 85 pairs were polymorphic and produced a total of 532 alleles among the 29 selected genotypes. AK18 had high Na and high PIC, whereas AK34 had low Na and low PIC (Figure 4, Table S1). The percentage of polymorphic loci was 96.5%. The mean Na was 6.488, and it ranged from 1 to 15 alleles. The mean Ne was 3.527, and it ranged from 1 to 8.327. The mean Ho was 0.121, and it ranged from 0 to 0.962. The mean He was 0.672, and it ranged from 0.043 to 0.901. The mean I was 1.328, and it ranged from 0.150 to 2.365.

SSR Analysis
A total of 85 pairs of newly developed SSR primers based on the published A. trifoliata genome [43] (NCBI database under BioProject ID PRJNA671772) were used to determine the genetic diversity of the 29 genotypes. All 85 pairs of SSR primers produced clear bands in all samples; 82 of the 85 pairs were polymorphic and produced a total of 532 alleles among the 29 selected genotypes. AK18 had high Na and high PIC, whereas AK34 had low Na and low PIC (Figure 4, Table S1). The percentage of polymorphic loci was 96.5%. The mean Na was 6.488, and it ranged from 1 to 15 alleles. The mean Ne was 3.527, and it ranged from 1 to 8.327. The mean Ho was 0.121, and it ranged from 0 to 0.962. The mean He was 0.672, and it ranged from 0.043 to 0.901. The mean I was 1.328, and it ranged from 0.150 to 2.365.

Cluster Analysis Based on Phenotypic Traits and SSR Data
A Ward dendrogram based on the similarities between the phenotypic traits and nutritional components clustered the 29 genotypes into four groups (Figure 5a). The first group consisted of two genotypes (1023 and 1411), which were characterized by a thin peel, large fruit size, high EP, VC, and TF; and late-maturing (growth period = 170-180 d).
The second group consisted of 11 genotypes, which were characterized by small fruit size, high RS, and mid-late maturing (growth period = 169-177 d). The third group consisted

Cluster Analysis Based on Phenotypic Traits and SSR Data
A Ward dendrogram based on the similarities between the phenotypic traits and nutritional components clustered the 29 genotypes into four groups (Figure 5a). The first group consisted of two genotypes (1023 and 1411), which were characterized by a thin peel, large fruit size, high EP, VC, and TF; and late-maturing (growth period = 170-180 d).
The second group consisted of 11 genotypes, which were characterized by small fruit size, high RS, and mid-late maturing (growth period = 169-177 d). The third group consisted of 15 genotypes, which were characterized by high TFAAs, medium fruit size, and precocity (growth period = 157-165 d). The fourth group only included one genotype (710), which had the largest fruit size and took the shortest time to mature (growth period = 150 d).

Identification of Promising Genotypes
Based on previous studies of commercial fruits, the screening standards of the main phenotypic traits and nutritional components were FW ≥ 200 g, EP ≥ 21%, RS ≥ 13 g·100 g −1 , VC ≥ 20 mg·100 g −1 , and TFAAs ≥ 450 ng·mg −1 . We identified the following four excel-  (Table S2 and Figure 6). An unweighted pair group method with arithmetic mean (UPGMA) dendrogram based on NGD was also produced (Figure 5b); the branch lengths indicate genetic distance. The 29 genotypes were divided into four groups with a genetic distance of 0.68; however, the topologies of the UPGMA dendrogram and Ward dendrogram differed. For example, 1023 was located in the most basal group in the two dendrograms, whereas 13 genotypes, including 78 and 516, were located in the same group.

Importance of Genetic Diversity in A. trifoliata Breeding
Genetic diversity has a large influence on both the effectiveness and the reliability of selection in plant breeding programs. Characterizing the genetic divergence and the degree of relatedness among various traits is thus essential for ensuring that selection leads to crop improvement [44]. However, the domestication process consists of a series of stages and several rounds of selection. Evaluating the genetic diversity of the population after the first round of selection is particularly important for ensuring the efficacy of subsequent rounds of selection.
Although efforts to domesticate A. trifoliata were only recently initiated, priority traits during pre-breeding have included disease resistance, especially resistance to major diseases as well as physical features such as fruit texture. A total of 29 resistant genotypes with desirable textures were selected from a collection of 1447 wild-collected plants, and these genotypes were subjected to another round of selection to enhance phenotypic traits, such as fruit color (FC), fruit shapes (FS), and fruit size (FW ≥ 200 g), and increase the content of nutritional components, such as vitamin C (VC), reducing sugar (RS), and total free amino acids (TFAAs). Amounts of VC, RS, and TFAAs greater than 20 mg·100 g −1 , 13 g·100 g −1 , and 450 ng·mg −1 , respectively, might be beneficial to human health and affect the purchasing decisions of consumers [45][46][47][48]. Large variation was observed in several phenotypic traits (Table 2); the CVs and I of FW, PW, SW, S/P, and EP were greater than 20% and 2.0, respectively. Large variation was also observed in several nutritional components, especially VC, RS, and TFAAs (Table 1). These findings suggest that the genetic diversity underlying phenotypic traits and nutritional characters was high and that these traits were possibly genetically independent of the traits targeted in the first round of selection, such as disease resistance and fruit texture traits. The Ward dendrogram also indicated that genetic diversity was high for several morphological traits (Figure 5a).
Previous studies have shown that the early selection of highly polymorphic species, especially natural populations of commercially unexploited plants, does not typically result in significant reductions in genetic variation [49,50]. SSR markers have been widely used for genetic mapping in crop species [51,52] as well as studies of genetic diversity [30] for their hypervariability, co-dominant inheritance, high reproducibility, high polymorphism with multiple alleles per locus, and high abundance and transferability between species. The Na and PIC values of the microsatellite markers for the 29 genotypes of A. trifoliata (Table S2) and the UPGMA dendrogram based on the NGD of the SSR data indicated high diversity among the 29 genotypes (Figure 5b). Although there were slight differences in the topology of the Ward dendrogram and the UPGMA dendrogram (Figure 5a,b), both dendrograms suggested that the genetic diversity of the population of 29 genotypes was high. Thus, further rounds of selection and evaluation of genotypes from this population would be effective for improving phenotypic traits and nutritional components. Such future studies will enhance our understanding of the dynamics of genetic diversity in crops under artificial selection.

Genetic Improvement of A. trifoliata
Plant breeding line selection must be based on multiple breeding objectives; however, accomplishing multiple breeding objectives can be difficult through a single selection strategy [53]. Knowledge of the associations among multiple objectives is important for enhancing the efficacy of selection because improvement in one trait can often lead to improvements in other traits through indirect selection [54]. Although correlations between fruit weight and fruit shape have been previously detected in A. trifoliata [18,19], correlations between morphological and nutritional traits have not been explored. In this study, LD was significantly positively correlated with VC (Table 2), which indicates that the improvement of LD by direct selection could lead to improvements in VC through indirect selection. Likewise, direct selection for the content of TF and S/P could lead to improvements in TD and TFAA, respectively, via indirect selection because of the significant correlations. However, direct selection for EP could lead to decreases in the content of TFAAs because of the significant, negative relationships of EP with TFAAs (Table 2). Overall, the new information provided in our study regarding the correlations between morphological and nutritional traits and the high genetic diversity will facilitate future efforts to genetically improve A. trifoliata.
PCA was used to reduce the complexity of our dataset, identify the traits underlying differences between genotypes, and determine the traits and trait axes explaining the most variation in the data [55]. The 15 traits assessed in this study were reduced to five PCs, which explained 83.2% of the total variance (Table 3), indicating that input variables were highly correlated ( Table 2). All morphological indices except PT loaded on PC1, and all major nutritional indices, such as VC, RS, and TFAAs, loaded on PC2 (Table 3).
Cultivar development in plant breeding programs is a complex process involving a cyclical procedure over a long period. Thus, the selection of an appropriate breeding strategy is essential for the success of plant breeding programs, and ideal strategies should be simple, economically feasible, and capable of efficiently achieving breeding objectives [56]. The results of the correlation analyses and PCA indicate that FW and TFAAs would be important targets of selection for obtaining superior genotypes for the development of a new A. trifoliata fruit crop. PT and TF would be important targets of selection for the development of a new medicinal crop, and both SW and S/P would be important targets of selection for the development of A. trifoliata as an oil crop.

Progressive Improvement of A. trifoliata as a New Commercial Fruit through Successive Selection
Adaptation to cultivated environments and resistance to diseases in such environments are typically the first problems needing to be resolved in the domestication of wild plants so that they can be cultivated at commercial scales [57,58]. Phenotypic traits and nutritional components become the targets of selection only after adaptation and disease resistance have been achieved. Although it has long been used as a medicinal herb in East Asia, efforts to cultivate and domesticate A. trifoliata as a fruit crop are only in their infancy [13]. Ongoing breeding efforts are aimed at developing commercial cultivars with improved fruit quality such as larger fruit size, higher EP, and larger amounts of nutrients. VC has been reported to be 21.58 mg·100 g −1 in strawberry [59] and 2.6 mg·100 g −1 in plum [60], and RS has been reported to be 7.15 g·100 g −1 in pineapple [61] and 13 g·100 g −1 in litchi [62].
The TFAAs of fruit crops have generally received less attention in previous studies [63]. Compared with these fruits, four excellent genotypes were identified in this study: 422, 548, 761, and 1023; the values of FW, EP, RS, VC, and TFAAs of the fruit of these genotypes were greater than 200 g, 21%, 13 g·100 g −1 , 20 mg·100 g −1 , and 450 ng·mg −1 , respectively. Although the EP of the fruits of these excellent genotypes was low in our study, which was similar to the EP of the fruits of genotypes in previous studies [18,19], the EP of the fruit of the genotypes from this study and previous studies is much higher compared with the ER of the fruit of wild plants (16.5%) [13].
Step-by-step breeding strategies according to the desirable order of objective traits are needed to further improve EP.

Conclusions
High genetic diversity was observed for the 29 genotypes based on the phenotypic traits, nutritional components and SSR data; thus, further selection of some of these traits from this population could prove effective. A stepwise selection strategy could be particularly useful for improving A. trifoliata as a fruit crop. Four genotypes (422, 548, 761, and 1023) with excellent properties (FW, RS, VC, and TFAAs) were identified with high potential to be developed as new commercial cultivars. The results of our study indicate that new fruit cultivars of A. trifoliata with excellent commercial characteristics and nutritional components could be developed with sufficient breeding efforts.
Supplementary Materials: The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/horticulturae8020116/s1, Table S1: Statistical values of microsatellite markers on 29 genotypes of A. trifoliata; Table S2: The characterized of 29 genotypes in 19 phenotypic traits and nutritional components of A. trifoliata; Figure S1: Distributions of VC (vitamin C), TFAA (total free amino acid), RS (reducing sugar), and TF (total flavonoids); Figure S2: The scree plot of principal component analysis.