In the past 5 years, we have investigated the distribution and conservation status of wild C. oleifera in the wide regions of southwest China (Sichuan province). Some individuals with excellent traits and distinct morphological variation were selected and marked in the field. A total of 143 wild accessions were finally clustered and used in this study, including Ya’an (YA, 60), Zigong (ZG, 16), Yibin (YB, 18), Luzhou (LZ, 7), Dazhou (DZ, 9), and Neijiang (NJ, 33) (Figure 1). All the germplasms were more than 10 years old growing naturally in the wild without fertilization and management. We performed a field survey of these resources for 2 consecutive years (2020 and 2021) and collected leaves, flowers, fruits, and seeds for the measurement of morphological traits.

At the stage of fruit physiological ripening, 30 healthy mature leaves, blooming flowers, and disease-free mature fruits per tree were randomly collected around the outer part of the canopy. For the determination of fatty acids and chemical composition, fresh fruits were picked from trees, saved in ice boxes, and rapidly transferred to the laboratory, where they were stored at 4°C.

All experiments were performed in triplicate. The mean of each trait based on a 2-year investigation and measurement was used for statistical analysis. For qualitative and pseudo-qualitative characters, traits were classified into 10 grades, 1 grade< X − 2σ, 10 grades > X + 2σ; each grade interval is 0.5σ between 1 and 10 grades; X and σ are the mean and standard deviation, respectively. The morphological diversity was evaluated by the frequency of trait dispersion and Shannon’s diversity index (H′). The statistics of quantitative parameters were measured, including minimum (Min), maximum (Max), mean, median, standard deviation (SD), coefficient of variation (CV, %), and H′. The H′ for each trait was calculated by using the following formula: H "  =   −   P i   ×  ln  ( P i ) (Pi is the proportion of the individual number of this trait in total individual number) (Lei et al., 2018). The CV for all quantitative traits was calculated as CV = S /x , where S is the standard deviation and x ¯ is the mean (Das and Divakara, 2011). The IBM SPSS Statistics version 20.0 (SPSS Inc., Chicago, IL, USA) was performed to estimate correlation among all quantitative traits with the Pearson correlation coefficient. Principal component analysis (PCA) was also applied to determine the relationship among the individuals. OriginPro 9.1 (OriginLab, Northampton, Massachusetts, USA) was used to perform cluster analysis. Correlation and bivariate correlation analyses were calculated by omicshare tools. (https://www.omicshare.com/tools/Home/Soft/getsoft). M-TOPSIS was achieved by MatLab 16.0 (MathWorks Inc., Natick, Massachusetts, USA).

According to the description of the oil-tea Camellia DUS test guidelines (LY/T 2742-2016), the leaf shape was classified into four ratings (subcircular, elliptical, long elliptical, and lanceolate). Nevertheless, based on the findings of our study (Figure 2A; Table 1), the leaf shape of the detected 143 accessions should be classified into five categories: subcircular (16.08%), oval (11.89%), lanceolate (6.29%), long elliptical (11.19%), and elliptical (54.55%). In this study, the shape of the leaf apex was classified into four categories: taper (53.85%), blunt (15.38%), round (6.99%), and sharp (23.78%) (Figure 2B; Table 1). As shown in Table 1, the H′ values for the leaf shape and apex shape were 1.31 and 1.15, respectively. For the quantitative characteristics of leaves, the CV values ranged from 11.16% to 24.86%, and the leaf area ranged from 669.34 to 2,320.00 mm² with the highest CV. The H′ of leaf quantitative traits varied between 1.99 and 2.11 and also showed high phenotypic diversity (Table 2).

The relative height between pistil and stamen was categorized into three categories among all 143 individuals: equal (36.36%), pistil higher (20.98%), and stamen higher (42.66%) (Table 1). The H′ for relative pistil and stamen height was 1.06 (Table 1). Additionally, white with crimson spots (1.4%) was observed in 143 wild C. oleifera resources (Figure 3A). The predominant petal color of 143 germplasm resources was white (98.60%), and very low phenotypic diversity was detected in petal color with an H′ value of 0.07 (Table 1). In addition, the stamen petalody, it should be noted, was a phenomenon that occurred with high frequency, representing the unstable flowering phenotype in wild C.oleifera resources (Figure 3B). Relatively high levels of diversity were also detected in flower quantitative traits; the CV for flower quantitative characteristics such as number of petals, number of sepals, number of stigmas, petal length, and flower crown ranged from 11.81% to 21.49%. The CV for the number of sepals was the greatest and ranged from 3.00 to 9.20. The H′ of flower quantitative traits varied between 1.40 and 2.08 (Table 2).

According to the morphologic variation in Figure 4, fruit shape was assigned eight ratings among the 143 accessions, the majority of which were spherical (52.45%), followed by peach (12.59%), oblate (10.49%), ovoid (9.09%), ellipsoid (8.39%), olive (4.90%), obovoid (1.40%), and gourd (0.70%) (Figure 4A; Table 1). The peel color was divided into six grades among the 143 individuals; the highest distribution frequency was green peel (47.55%), and the lowest was purple-red peel (0.70%). The red-green (22.38%) and yellow-green peel (24.48%) also covered a relatively high frequency (Figure 4B; Table 1). The H′ values for fruit shape and peel color were 1.50 and 1.59, respectively. The seed color was classified into four categories based on Figure 5A’s depiction; the types and distribution frequencies of each category were dark brown (23.08%), black (26.57%), tan (47.55%), and brown (2.8%), respectively. The seed shape was classified into five types, representing hemispherical (14.69%), irregular (20.98%), spherical (4.90%), renal-like (41.96%), and conical (17.48%), respectively (Figure 5B: Table 1). The H′ values of seed color and shape were 1.14 and 1.43, respectively (Table 1). The mean and median values of the fruit quantitative characteristics were consistent, representing the stability and typicality of the detected samples (Figure 6; Supplementary Table S2). The high level of CV values ranged from 10.22% to 35.05%, suggesting the richness of variability for the majority of these detected traits. Among these, the oil rate of fresh fruit had the largest CV value with a mean value of 7.79 and ranged from 1.45 to 14.64 (Supplementary Table S2). The shape index of fruit showed a lower level of diversity, with a CV value of 10.22% and a mean value of 1.02, ranging from 0.71 to 1.41. The results of fruit and economic characters presented similar levels of phenotypic diversity, with H′ values ranging from 1.96 to 2.07 (Supplementary Table S2).

Considering the analysis of fatty acid component content, the eicosenoic acid showed the highest variability (43.49%), ranging from 0.16% to 0.74% with a mean of 0.31%. As the main ingredient of fatty acid in C. oleifera oil, the oleic acid had the lowest variation (5.34%), with a range of 66.07%–83.71%, suggesting the relatively stable content of oleic acid in C. oleifera (Table 3). The H′ values of oil fatty acid characteristics ranged from 1.47 to 2.19, representing a higher level of diversity. For the oil quality index, the peroxide value had the highest CV of 81.31%, and the free acidity also holds a relatively high variation (41.96%). The H′ for free acidity and the peroxide value were 1.78 and 1.63, respectively (Table 3). The variability and genetic diversity were also detected in the lipid nutrient content. The H′ and CV values of α-tocopherol, squalene, and sterol were 1.90, 1.65, and 1.59 and 68.19%, 80.68%, and 72.66%, respectively (Table 3).

In this study, hierarchical clustering was performed to analyze the relationship among the 143 wild C. oleifera germplasm samples. The result showed that all the accessions can be assigned to six distinct groups (Figure 7; Supplementary Tables S3). Group I contained 33 germplasm resources, representing 23.08% of the total accessions. This group was distinguished by the small leaves and high squalene content (Figure 7; Supplementary Table S4). Group II only consists of three germplasm resources (NJ1, YB14, and YB15) with small fruits, low oil content, high fresh seed rate, and tocopherol content. Group III contained 28 resources, which were separated by the high shape index of fruit and moderate other characteristics. Group IV contains nine individuals with large flowers and fruits, a high dry seed rate and oil rate of fresh fruit, and low free acidity and sterol content, all of which could be used as improved breeding materials. Group V contained 55 resources containing 38.46% of the total individuals, primarily from Ya’an. The characteristics of the group included high kernel rate of dry seed, oil rate of the kernel, oil yield, oleic acid and sterol content, low peroxide value, and moderate characteristics that could be used as improved breeding materials. Group VI contained 15 resources representing large leaves and fruits, high free acidity and oleic acid content and a small flower with low palmitic acid, squalene, and α-tocopherol content (Figure 7; Supplementary Table S4).

The Circos methodology was applied to better visualize and intuitively explore relational links among 32 quantitative characteristics with 143 wild C. oleifera germplasm resources (Figure 8A). Different abundances of 32 quantitative characteristics were detected in 143 accessions, among which the abundance of leaf area and sterol content was much higher than other traits. In addition, complex relationships among the 32 quantitative characteristics were estimated based on correlation analysis with the Pearson correlation coefficient in the 143 accessions (Figure 8B). Significant associations among leaves, flowers, fruits, seeds, and oil quality characteristics were shown, especially for some fruit traits. Strong positive correlations exist between fruit weight and fruit height, fruit diameter, peel thickness, and the number of seeds with the coefficient ranging from 0.42 to 0.92. Peel thickness had significantly negative correlations with fresh seed rate, dry seed rate, peroxide value, palmitic acid, stearic acid, squalene, and α-tocopherol contents, with the coefficient ranging from −0.23 to −0.70. Significant positive correlations also exist between oil content (oil rate of kernel, oil rate of dry seed, and oil rate of fresh fruit) and oleic acid (r = 0.18–0.44), and high and significant positive correlations were also shown among the three oil content indices. Strong negative correlations between oil rate of fresh fruit and other traits such as moisture rate of fresh seed, free acidity, peroxide value, stearic acid, eicosenoic acid, squalene, and α-tocopherol were tested, with coefficients ranging from −0.21 to −0.78. A positive correlation (r = 0.28) was also observed between oleic acid and sterol. Among all the 32 quantitative traits, the most positive correlation (r = 0.70) was detected between leaf length and leaf width, while the most negative (r = −0.96) was shown between the content of oleic acid and stearic acid (Figure 8B).

The PCA graph of 143 wild C. oleifera germplasm resources was obtained to demonstrate the distribution of accessions according to differences in quantitative phenotypic characteristics (Figure 9; Supplementary Figure S1). The results of PCA showed a distinctive separation among the 143 wild C. oleifera germplasm resources according to the phenotypic characteristics. Therefore, these phenotypic parameters can be used as an essential criterion for defining wild C. oleifera germplasm resources. Ulteriorly, The PCA was performed to identify the main distinguishing traits of the 32 quantitative characteristics. The dimension implied by the 32 quantitative characteristics was reduced to nine significant components, accounting for 80.79% of the total variance based on eigenvalues greater than 1 (Figure S1). The first factor, which accounted for 24.28% of the total variance comprised the kernel rate of dry seed, the oil rate of the kernel, and the kernel rate of dry seed (Table 4), so it can be referred to as the oil content factor. The fruit weight, diameter, and peel thickness had a higher loading on the second principal component, so it was referred to as the fruit yield factor. The third principal component, also known as the fatty acid factor, primarily represents fatty acids. The fourth component comprises the highest fruit weight load value. The petal length and flower crown had the highest load value, so the fifth factor was designated as the flower factor. In addition, the sixth, seventh, eighth, and ninth factors were referred to as leaf, fruit, free acidity, and fruit shape factors, respectively.

The TOPSIS method incorporating the Mahalanobis distance (M-TOPSIS) is used for comprehensively evaluating germplasm. It is a novel, modified, and more practical synthetic evaluation method than TOPSIS. Here, 11 indexes including leaf width, number of sepals, flower crown, fruit weight, fruit height, fruit diameter, shape index of fruit, oil rate of dry seed, free acidity, stearic acid, and eicosenoic acid contents were selected from 32 quantitative traits to evaluate the 143 wild C. oleifera germplasm resources in all aspects based on correlation and PCA analyses. Following that, a comprehensive evaluation model of C. oleifera resources was constructed by the M-TOPSIS, and the top 10 accessions were screened as YA53, YA13, YA40, YA34, YA57, YA19, YA33, YA41, DZ8, and YA7 (Table 5).

The coefficient of variation can reflect the degree of difference between various phenotypic traits. A strong positive correlation has been reported between the coefficient of variation and the degree of phenotypic difference as well as genetic diversity. It provided a stronger possibility for using phenotypic traits to identify varieties and germplasms (Zhang et al., 2022). Based on the analysis of 41 phenotypic traits of 143 germplasms, significant phenotypic differences were found among various wild C. oleifera germplasms. During the 2 years of investigation, the high degree of phenotypic variation indicated the abundant genetic diversity existing in C. oleifera germplasms. Some workers have also reported similar results in C. oleifera and its relatives, C. meiocarpa (Huang, 2011; He et al., 2020). In the study, the median and mean values of quantitative characteristics from 143 germplasms were nearly identical, indicating that the investigated germplasm resources were representative. The higher value of H′ (1.15–2.11) was detected in leaf traits, reflecting the greater genetic variation in these characters (Tables 1, 2). The leaf trait has been considered an important index in plant science research for it reflects the adaptability of plants to different environments and their ability to self-regulate in response to complex physiological environments (Hu et al., 2022).

Flower variation is the raw material that natural selection can amplify, resulting in plant diversification over time (Herrera, 2005). A significant difference was shown in flower phenotypic traits with H′ from 0.07 to 2.08, implying the foundation of these traits for variety breeding and excellent germplasm choosing (Tables 1, 2). In addition, the stamen petalody was a surprising phenomenon during our investigation (Figure 3B). The stamen of the male sterile mutant shows remarkable petalody, increasing whorls of petals and generating different flower forms, which are extremely important traits for ornamental value and serve as a useful genetic tool because they eliminates the need for artificial emasculation (Li et al., 2021).

The primary reason for cultivating C. oleifera is to extract edible oil. The seeds of C. oleifera are the primary oil storage components. The characteristics of C. oleifera seeds are crucial for oil yield and quality and are essential for DUS testing of Oil-tea Camellia (Zhu et al., 2020). Moreover, fruit traits are the essential phenotypic traits for fruit-producing economic tree species, which directly or indirectly affect the yield of C.oleifera (Zhu et al., 2020). The fruit phenotype is crucial for further elucidating genetic diversity and developing superior varieties, and the fruit descriptors system of C. oleifera should be supplemented and completed with more fruit characteristics, such as the fruit shape should be added with peach, olive, and gourd (Figure 4). In this study, the CV for oil content of fresh fruit (35.05%) from 143 wild C. oleifera resources was the highest and the degree of data dispersion was the greatest (Supplementary Table S2). As an important woody oil crop, the oil content of fresh fruit is directly proportional to oil production and economic value (Hao et al., 2017). Research on fruit traits is critical and closely related to oil yield, an important breeding and variety-promotion reference (Yang et al., 2022).

The oil extracted from C. oleifera is a kind of high-quality edible oil rich in many bioactive substances (Zhu et al., 2019). In this study, the CV values of peroxide value, α-tocopherol, sterol, and squalene were 81.31%, 80.68%, 72.66%, and 68.19%, respectively (Table 3). It indicated that these four traits have an enormous genetic variation (Kumar et al., 2020). On the contrary, the low level of CV for palmitic acid and oleic acid was detected here, which indicated that the variation was slight and the genetic characteristics were stable.

The cluster analysis can gather varieties with similar genetic information into one group, which is conducive to studying the genetic relationship between varieties (Dalmaijer et al., 2022). In this study, the wildest C. oleifera germplasms from YA were clustered into one group (Figure 9), indicating the existence of a certain correlation between phenotypic traits and geographical location among different materials. However, the geographical origin might not play a decisive role in phenotypic traits, some of the accessions from different geographical locations in this study were clustered together (Figure 9). Genetic variation may be occurring frequently when materials from various sources adopted significant habitat differences, which may be caused by the differences between the original environment and the present habitat (Singh et al., 2013). Except for the differences in genetic material, it is also possible that the introduction of germplasm materials to the local long-term planting caused differences in their growth and development (Kumar et al., 2009). In addition, there was obvious phenotypic differentiation among different groups, which can provide optimized germplasm materials for varieties breeding of C.oleifera. The richness of various traits not only improves the efficiency of breeding and speeds up the process of breeding but also provides a fundamental reference for the selection of hybrid parents and the optimization of combinations in breeding (You, 2021).

This study observed significant negative correlations between peel thickness and fresh seed rate, dry seed rate, peroxide value, palmitic acid, squalene, and α-tocopherol (Figure 8), which was consistent with previous findings (He et al., 2020). There were significant positive correlations between oil content and oleic acid, linoleic acid, and sterol, indicating the possibility of excellent resource selection with high oil yield and nutrition content. Significant negative correlations were observed between oleic acid, palmitic acid, and stearic acid. The result agreed with the previous study that unsaturated fatty acid enzymes predominantly catalyze the formation of fatty acids and that oleic acids are formed after the prolongation and desaturation of palmitic acid and stearic acid during the synthesis of fatty acids in plants (Zhao et al., 2015).

PCA is an effective method for reducing the dimensionality of large datasets, which can maximize interpretability, minimize information loss, and determine the most suitable traits that mostly contribute to the variation in the selected materials (Nardo et al., 2005; Jolliffe and Cadima, 2016). In this study, PCA confirmed that the first nine components explained the vast majority of the variation, concentrating on several characteristics, such as the oil content of dry seed, fruit diameter, stearic acid, fruit weight, petal length, leaf width, fruit height, free acidity, and fruit shape indexes (Figure 9; Figure S1). The results suggested that such characters are suitable both for the assessment of genetic diversity and for the phenotypic characterization of wild C. oleifera germplasm.