All 59 half-sib families with 6 replications and 6 individual trees of slash pine were planted in a randomized complete block of a 2 m × 3 m spacing in the Yuhang region in Hangzhou, China (30°27′N, 119°49′E), established in March 1994. The specific experimental design has been detailed in Ding (Ding et al., 2022).

Pine resin samples were collected in July 2021 from individuals whose diameter at breast height was close to the mean value of a family. A special plastic pipe with an aperture of 20 mm and volume of 15 ml was fixed at the drilling hole in the trunk to collect resin from the sunny side (Zhang et al., 2017). In total, we collected the resin from 219 representative individuals from 59 unrelated open-pollinated families ( Table S1 ).

Gas chromatography experiments were carried out with a GC 6890 gas chromatograph coupled with a Hewlett Packard GC 5975B mass spectrometer (Agilent 5975B, Santa Clara, CA, USA). The GC-MS operation was performed according to the procedure described by (Chen et al., 2020). To obtain the resin component content, 0.05 g resin was dissolved in 0.5 mL of ethyl alcohol containing 50 µL tetramethylammonium hydroxide. The GC column temperature conditions were as follows: the initial column temperature was 60°C, held for 2 min, increased at 8°C min^-1 to 80°C, and reached a maximum of 280°C at a rate of 2°C per min for 5 min. The helium gas flow was set at 1 ml min^-1. The temperature of the injector was 260°C, and the volume was 1 µL with a 1/50 split ratio. Mass spectra were recorded under electron impact ionization at an electron energy of 70 eV in the range from m/z 30 to 600 along with solvent delay for 3 min.

Resin compositions were identified by matching experimental fragmentation patterns in mass spectra with the NIST08 database through the data processing system of Agilent Chem Station and then comparing with the relevant literature. Monoterpene and sesquiterpene content were determined by isobutylbenzene content, and diterpene content was determined by heptadecanoic acid content. We calculated the resin component content by comparing their peak areas.

Mixed linear model (LMM) with restricted maximum likelihood (REML) analysis was used for the estimation of genetic parameters. The model for a single-trait observation y_i for a tree is shown in Equation (1):

where x_i is a vector linking the fixed effects m to the observation y_i , and b_i , f_i , and e_i are the random block, family, and residual effects, respectively. Stacking those vectors for all trees produces model Equation (2):

where y is a vector of total phenotypic observations, m is the vector of fixed effects (overall mean), and b, f, and e are vectors of bivariate random effects for block, family, and residual effects, respectively. X, Z1, and Z2 are the incidence matrices linking observations to the appropriate effects. We defined the vector of expected values (E) and dispersion matrices (Var) as:

and

where ⊗ and ⊕ are the direct product and direct sum operations, respectively; σ b i 2 , σ f 1 2 , and σ e 1 2 represent the block, family, and residual variances for trait i, respectively; and σ_bibj , σ_fifj , and σ_eiej are the covariances of block, family, and residual between traits i and j.

The variance components from the model were used to calculate the narrow sense heritability (h² ):

The genetic correlations (r _{g ij} ) between trait i and trait j were calculated as:

and phenotypic correlation ( r _{p ij} ):

where σ _{f ij} is the estimated family covariance between trait i and trait j and σ f i 2 and σ f j 2 are the estimated family variances for trait i and trait j.

The breeding value (BV) of each tree is the family random effect value of the LMMs, and the realized genetic gain (ΔG_R ) was computed by subtracting the mean BVs of the selected top ratio trait from its total mean.

The statistics were calculated by the “mmer” and “h2.fun” functions of the “sommer” package (Covarrubias-Pazaran, 2016), and the figures were plotted by the “ggplot2” package (Gómez-Rubio, 2017). All analyses were performed in the R environment (Team, 2013).

In total, we obtained 10 monoterpenes, 3 sesquiterpenes, and 18 diterpenes ( Table S2 ). We then selected four components believed to have broad industrial utility for subsequent analysis, including two monoterpenes (α pinene and β pinene) and two diterpenes (abietic acid and levoprimaric acid). The contents variation of the four compositions have a reasonable range (2.00% to 24.12% in monoterpenes and 2.35% to 28.58% in diterpenes) ( Figure 1 ). Among monoterpenes, α pinene has the highest content (mean value = 14.79%), while levopimaric acid has the highest content in diterpenes (mean value = 21.63%). The β pinene has the largest content variation with a variable coefficient (CV) of 0.42, followed by abietic acid (CV = 0.37), α pinene (CV= 0.29) and levopimaric acid (CV = 0.15).

The estimated narrow-sense heritabilities (h² ) of the four resin components were 0.67, 0.57, 0.63, and 0.51, respectively ( Figure 2 ). Overall, the heritability of the four resin components was high and the monoterpenes had higher heritability than diterpenes. The h² of α pinene was the highest, while the levopimaric acid has the lowest h² . Additionally, the margin of system error for h² is small, between plus or minus 0.13 to 0.17. This indicated that the traits of resin components of slash pine were not greatly affected by environmental factors but were mainly controlled by genetics.

The genetic (r_g ) and phenotypic correlation (r_p ) analyses of the four resin components of slash pine are shown in Figure 3 . The strongest negative genetic correlation (r_g = -0.67) and phenotypic correlation (r_p = -0.89) were found between the two monoterpenes (α pinene and β pinene). The β pinene and abietic acid had the weakest genetic correlation (r_g = 0.62) and phenotypic correlation (r_p = 0.67). Interestingly, the genetic correlation (r_g = 0.67) and phenotypic correlation (r_p = 0.84) between the two diterpenes (abietic acid and levoprimaric acid) were stronger than those between monoterpenes and diterpenes.

We ranked the family performance of slash pine by a breeding value (BV) of α pinene, β pinene, abietic acid and levoprimaric acid in Figure 4 . Apparently, α pinene had a significantly consistent family ranking with β pinene, while a similar result was also found between abietic acid and levoprimaric acid. Unfortunately, β pinene and abietic acid did not show a highly consistent family ranking, indicating a worse correlation between the monoterpenes and the diterpenes, which was consistent with the conclusion of Figure 3 . However, it was still possible to select traits by families according to certain purposes.

The relationship between α pinene, β pinene, abietic acid and levoprimaric acid traits is displayed in Figure 5 . There were 15 families with BVs above the mean value of α pinene and β pinene that were marked in red. In addition, 15 families were identified that involved a BV above the mean of abietic acid and levoprimaric acid, which were labeled as purple. Fortunately, there are 5 families, including 2, 3, 4, 6 and 7, whose BVs are above the mean of all four resin components that deserve to be selected.

The ΔG% (improved percentage of the genetic gain) was calculated with the top 10% to 40% samples from 59 families for each resin component ( Figure 6 ). On the whole, the top 10% of families could produce the highest yield of pine resin among other selection intensities. Except for the top 10%, there was little difference in the performance of the other three intensities. The top 10% yielded nearly 1.5 times as much as the top 40% selection intensity in abietic acid but not significantly in α pinene, β pinene, and levoprimaric acid.

The largest average ΔG% was observed in β pinene, where the top 10% to 40% families yielded 42.7%, 39.9%, 37.1, and 35.1%, respectively, followed by α pinene, abietic acid, and levoprimaric acid, with estimated genetic gains ranging from 37.0% to 31.0%, 40.3% to 28.4%, and 11.8% to 9.3%, respectively.

In this study, we first assessed the variation and heritability of four relatively important resin components in the slash pine population. The results showed that the content of α pinene is approximately 1.4 times higher than that of β pinene, which was similar to the study conducted by Zhang (Zhang et al., 2016) and Lai (Lai et al., 2020). However, the content of α pinene accounted for more than 90% of the total monoterpenes in Masson pine (Liu et al., 2016). The content of levopimaric acid was the highest, accounting for 21.63% of the total content on average, followed by abietic acid, accounting for 8.12%, which were, respectively higher and lower than the results of previous similar studies (Zhang et al., 2016; Lai et al., 2020). These differences may be attributed to differences in environmental and climatic conditions. In addition, turpentine is volatile, and errors in resin collection and measurement must be accounted for (Luan et al., 2022).

Heritability is an important population parameter that can help understand the genetic architecture of complex traits (Lee et al., 2011). In this study, the four resin components were under strong genetic control, which was higher than the results estimated by Lai (Lai et al., 2020). A similar result that heritability of α pinene is higher than β pinene was also found both in slash pine (Zhang et al., 2016) and Masson pine (Liu et al., 2016). An estimate of the heritability of a trait is specific to population, species, and environment, and it may change over time (Li et al., 2020). High heritability will drive the process of genetic improvement of these four important resin components of slash pine. The heritability estimated in this study would be significant for our slash pine population and has some reference value for other studies.

Understanding the genetic and phenotypic correlation among traits contributes to researchers developing breeding strategies more effectively (Lee et al., 2002; Li et al., 2020). The correlation analysis of genetic and phenotypic of the four main resin components found that two monoterpenes (α pinene and β pinene) exhibited a strongly negative correlation, and α pinene and β pinene had negative and positive correlations with the two diterpenes (levopimaric acid and abietic acid), respectively, which agrees with the results of previous reports (Lei et al., 2015; Zhang et al., 2016; Lai et al., 2020). It is believed that α pinene and β pinene may be controlled by the same pair of alleles, and the syntheses of these two components are negatively correlated (Li et al., 2012). Alternatively, they are controlled by different genes, and their synthesis signals suppress each other (Hall et al., 2013; Diao et al., 2022). There was a strongly positive genetic and phenotypic correlation between β pinene and two diterpenes with high content in this study. Previous studies have also shown that β pinene is significantly related with resin yield traits, which highlights the importance of β pinene in high yield of pine resin breeding (Karanikas et al., 2010; De Lima et al., 2016; Neis et al., 2019).

Simultaneous selection of several important monoterpenes and diterpenes in resin is desirable for breeders in a long-term breeding program. Our family breeding value (BV) study of four important slash pine resin components provided the possibility for their joint selection. This method has been successfully applied to the combined selection of slash pine growth and wood traits (Li et al., 2020), as well as the combined selection of leaf color parameters and chlorophyll of Sassafras tzumu (Li et al., 2019). In this study, we selected 5 families, 2, 3, 4, 6, and 7, all of which contained high levels of α pinene, β pinene, abietic acid and levoprimaric acid, providing an important reference for the subsequent establishment of high-resin-yielding slash pine seed orchards.

Genetic gain can characterize the effect of direct or indirect breeding selection. Stronger genetic selection ratios tend to lead to higher genetic gain (Li et al., 2018; Zhang et al., 2018), and the same patterns were also reflected in this study. The β pinene possesses the highest genetic gain, which means that our high resin yield breeding for slash pine is effective.