Construction of a DNA Fingerprinting of Improved Varieties of Slash Pines in China Based on 51K Liquid-Phased Probes

1. Introduction

Slash pine(Pinus elliottii) has been cultivated in China for over a hundred years since its introduction [1]. It is characterized by rapid growth, wide adaptability, and high oleoresin yield, making it a significant timber [1,2] and resin-producing [3] species in southern China. Improved varieties are crucial in forest tree genetics and breeding. The large planting area creates a high demand for seedlings of better types of slash pine. Recently, dozens of the improved Slash pine varieties(grafted clones) have been selected and authorized by the variety commission in China. Effective identification of improved varieties is key to tree improvement and breeding. Constructing a DNA fingerprint to highlight the molecular specificity of improved varieties is very necessary.

DNA fingerprinting is a technical tool that enables to efficiently identify plant and animal lineages [4]. Genotyping wild plant species and their cultivated relatives is now often accomplished through the use of plant DNA fingerprinting techniques. Molecular markers are crucial for constructing the DNA fingerprint. They are renowned for their high sensitivity and specificity, reproducibility, high -throughput capability, and co-dominant inheritance [5]. They allow the detection of genetic variation at the DNA sequence level, which makes it possible to identify single nucleotide polymorphisms(SNPs)[6]. DNA-based markers can be differentiated into two types [7], the first is non-PCR-based(RFLP)[8,9] and the second is PCR-based markers(RAPD, AFLP, SSR, SNP, etc.)[10,11,12,13]. Nowadays, using SNP-based probe arrays for genotyping enables the rapid acquisition of a large number of molecular markers. This is an ideal and efficient method that has been widely applied in plants [14]. SNPs offer several advantages, including their abundance, biallelic nature, relatively low mutation rate, even distribution across the genome, and relative ease of detection.[15,16,17]. SNP arrays are flexible because researchers can design specific chips based on their research needs and select particular genetic markers for analysis [14,18,19,20]. This technology of developing a SNP array has been widely applied in forestry, especially in coniferous tree species [18,21,22].

The paper of the development of 51 K liquid-phased probe array for loblolly and slash pines has been published in May 2024[23].It is designed based on the published SNPs and probes of loblolly and slash pines, novel slash pine-specific long-read transcript, and SNPs obtained from a double digest Restriction-Site Associated DNA sequencing(ddRADseq). Furthermore, targeted capture sequencing was performed using 51K liquid-phased probes on three pine species of loblolly, slash, and Caribbean pine. The probe capture had an efficiency of 68.26% on average. Principal component analysis(PCA), phylogenetic analysis, and genetic structure analysis of SNPs genotyped for the three pine species demonstrated the transferability of the SNP array for the three pines. The development of the 51K liquid-phase SNP array represents a significant breakthrough, enabling efficient and cost-effective genotyping of pine species. This technology is of great importance for identifying pine varieties, analyzing genetic diversity, and advancing germplasm research.

As another application for the 51K liquid-phased probes, this study conducted genotyping on slash pine samples, aiming to develop a technical process for identifying improved slash pine varieties. Additionally, the study sought to screen molecular markers that represent improved varieties and construct a DNA fingerprinting for them to efficiently manage the genetic resources(Supplementary Figure S1).

2. Materials and Methods

2.1. Materials and DNA Extraction

All materials were planted in Changle Forest Farm in Zhejiang Province, Hangzhou city.

We selected two set of slash pines(Figure 1): one set for constructing the DNA fingerprinting and the other set for validating the accuracy of the fingerprinting, which we also refer to as the validation population. The first set includes improved slash pine varieties, comprising 29 slash pine varieties officially recognized by the variety committee. These samples, each derived from distinct slash pine clones, were cultivated in a seed orchard. They are distinguished by their superior traits, such as fast growth speed, excellent timber characteristics, and higher turpentine yield. We utilized these 29 samples to develop a DNA fingerprinting for improved slash pine varieties. The second set of samples, referred to as the validation population, is designed to evaluate the accuracy of variety identification using the constructed DNA fingerprinting. For this validation population, we selected a progeny test forest, which includes 369 individuals. Notably, 77 of these individuals come from the 7 improved varieties family used in the first set, indicating that the validation population includes some of the known improved families(Supplementary Table 1,2).

Collect young tender needles from two sets of the slash pines for use as experimental samples. The needle was refrigerated in ice bags before DNA was extracted. M5 HiPer Universal DNA Mini Kit(Mei5 Biotechnology, Beijing) was used to extract the DNA from the sample needles. The genome target regions were captured by 51K liquid-phased probes designed by our lab for subsequently next-generating sequencing. The specific experimental steps are as follows: (1) The DNA was segmented and the end of the DNA fragment was repaired by Pre-PCR to amplify the library; (2) The probe hybridizes with the target region;(3) Hybridization probes were captured by magnetic beads using streptomycin affinity labeling;(4) The enriched target fragment was eluted and amplified by Post-PCR after capture; (5) Second generation sequencing was performed on the target region.

2.2. Data Quality Control and Genotyping

The captured DNA fragments were sequenced by Illumina Xten high-throughput Sequencing platform. The raw data were stored in FASTQ format. The adapters and low-quality data were filtered by FastQC software [24] to obtain clean reads. The steps of data filtering are as follows:(1) Remove reads with adapters;(2) Remove reads with an N content exceeding 10%;(3) Remove reads where more than 50% of bases have a quality score below 10. Subsequently, single nucleotide polymorphism(SNP) detection is implemented using the GATK software toolkit(https://gatk.broadinstitute.org/hc/en-us). According to the localization results of clean Reads in the reference genome, samtools was used to remove duplications(Mark Duplication), and GATK was used to perform local realignment and base recalibration. The reference genome we used is the reference genome of Loblolly pine and long-read transcripts specific to Slash pine. To ensure the accuracy of SNP detection, GATK was used to detect and filter single nucleotide polymorphisms, and the final SNP loci were obtained. The mutation results are displayed and saved in vcf files.

2.3. Core SNP Selection

The raw genotyped SNPs were screened in the following criteria: SNP loci in the Raw data were selected, and the screening criteria were as follows (Table 1): (1) biallelic sites;(2) minor allele frequency(MAF) > 0.3;(3) miss rate=0;(4) conform to the Hardy-Weinberg equilibrium; (5) linkage disequilibrium(LD) value > 0.2;(6) PIC > 0.35.(7) heterozygosity rate < 0.25[7,8,25,26,27].

The core SNPs were obtained after the screening of the above steps. VCFtools [28] and BCFtools [29] were used to perform the filtration. The genetic parameters were calculated by the R script.

2.4. Simplification of SNP Markers Based on Random Forest Model

To improve the efficiency of variety identification, it is crucial to optimize the core SNP loci. The goal is to maximize the number of identifiable individuals while minimizing the number of loci used. We employed the Random Forest algorithm to evaluate the contribution of each SNP to individual identification, using the Gini index to rank SNPs by their importance. The SNPs were ranked in descending order based on the Mean Decrease Gini index, which measures each SNP’s contribution to distinguishing individuals. We selected the 50 SNPs with the highest Gini indices for further validation.(Supplementary Table 3).

To assess the effectiveness of the selection and the genetic diversity of the candidate SNPs, we constructed phylogenetic trees for 29 improved slash pine using both the core SNPs and candidate SNPs. The phylogenetic tree was constructed using the neighbor-joining method and completed in MEGA software.

2.5. Phylogenetic Analysis and Genetic Diversity Analysis of Validation Population

The samples in the validation population underwent the same steps in sections 2.1 and 2.2 to acquire their genotype. The SNPs retained after quality control were initially filtered using the following criteria: minor allele frequency(MAF) < 0.01, missing rate > 0.8. The resulting SNPs were used for validation of population genetic structure and phylogenetic analysis. We calculate the genetic parament by the Shell script.

MEGA 11.0 software was used to construct a Neighbor-Joining phylogenetic tree. Principal component analysis(PCA) was conducted using PLINK software(version 1.07)[26] to assess population stratification and genetic relationships among the individuals. PCA was executed using the command `--noweb` in PLINK, which generates eigenvalues and eigenvectors representing the major axes of genetic variation. The top two principal components were retained for subsequent analysis. The R package ggplot2 is used for visualizing population genetic structure and PCA results.

2.6. Validation of Candidate SNPs Based on Random Forest Model and Construction of DNA Fingerprinting

To verify whether the 50 SNPs we selected have good identification accuracy of improved slash pine varieties, we trained a random forest model using the 50 SNPs from 29 superior varieties. Subsequently, we included the 50 SNP information from 370 samples in the validation population in the model for training. The model will use these 50 SNPs to predict whether the samples in the validation population are improved varieties. In fact, the samples in the validation population can be classified into two distinct categories: certified improved varieties and unimproved varieties (Supplementary Table 2). The predicted classification results from the model will be compared with the actual classifications. We used the “Random Forest” package in R for this analysis and assessed the model’s accuracy by examining the confusion matrix and error rate. Finally, we achieved satisfactory validation results, and we constructed the DNA fingerprinting of improved slash pine varieties using candidate SNPs(Supplementary Table 3).The Materials and Methods should be described with sufficient details to allow others to replicate and build on the published results. Please note that the publication of your manuscript implicates that you must make all materials, data, computer code, and protocols associated with the publication available to readers. Please disclose at the submission stage any restrictions on the availability of materials or information. New methods and protocols should be described in detail while well-established methods can be briefly described and appropriately cited.

3. Results

3.1. Screening and Simplification of Core SNPs

A total of 5,60,567 SNP loci were genotyped for 29 improved varieties of slash pine by target sequencing of 51 K liquid-phased probes. The genotyping rate was 94.61%. A total of 183,849 SNPs with a minor allele frequency(MAF) < 0.01 and a miss rate > 0.8 were retained as the primary dataset for this experiment.

Based on the MAF, miss rate, linked disequilibrium(LD) value, heterozygosity, polymorphism information content(PIC), and other indicators of SNPs, a rigorous screening process (Table 1) was formulated to screen high-quality core SNPs. A total of 3502 SNPs were finally screened. The MAF of 3502 SNPs ranged from 0.241 to 0.276 with an average of 0.258(Figure 2a). The PIC values ranged from 0.299 to 0.320, with an average value of 0.3093(Figure 2b). The heterozygosity values ranged from 0.2992 to 0.3197, with an average value of 0.3093 values ranged from 0.366 to 0.399, with an average value of 0.382(Figure 2c). The results suggest that the 3502 core SNPs have a high degree of polymorphism and are ideal for DNA fingerprinting of improved varieties of slash pine.

We constructed neighbor-joining phylogenetic trees for 29 improved varieties of slash pine using 50 candidate SNPs and 3502 core SNPs, respectively. The results showed that when we calculated phylogenetic trees using 50 SNPs and 3502 SNPs, respectively, the genetic clustering of the 29 improved varieties was largely consistent with the original clustering, with only a few individuals showing different classifications.

3.2. Population Structure of Validation Population

We selected a progeny test population as the validation cohort, consisting of 371 samples. We performed a preliminary analysis of its genetic structure and diversity, dividing it into three subpopulations(Figure 2a,b). The fixation index(Fst) and gene flow index(Nm) indicate that there is only moderate genetic differentiation and low gene flow among the three subpopulations(Table 2). Compared to the first set of samples, this population exhibits lower genetic diversity. The average MAF value of all SNP markers in the validation population is 0.129(Figure 3c), the average PIC value is 0.156(Figure 3d), and the average Het value is 0.187(Figure 3e). These genetic parameters suggest that the population has relatively low genetic diversity. Due to the small genetic differences among subpopulations, this population is more suitable for testing the accuracy of the 50 SNPs in identifying superior varieties.

3.3. Effectiveness of Identifying Improved Varieties with Candidate SNPs

We selected a progeny test population as the validation cohort, consisting of 369 samples. We performed a preliminary analysis of its genetic structure and diversity, dividing it into three subpopulations(Figure 2a,b). The fixation index(Fst) and gene flow index(Nm) indicate that there is only moderate genetic differentiation and low gene flow among the three subpopulations(Table 2). Compared to the first set of samples, this population exhibits lower genetic diversity. The average MAF value of all SNP markers in the validation population is 0.129(Figure 3c), the average PIC value is 0.156(Figure 3d), and the average Het value is 0.187(Figure 3e). These genetic parameters suggest that the population has relatively low genetic diversity. Due to the small genetic differences among subpopulations, this population is more suitable for testing the accuracy of the 50 SNPs in identifying superior varieties.

3.4. Population Structure of Validation Population

To validate whether the 50 SNPs we selected have good identification accuracy, we trained a random forest model using the 50 SNPs from 29 improved varieties. Subsequently, we included the 50 SNP from the validation population in the model for training.The random forest algorithm generated a confusion matrix to display the prediction results(Figure 4a). The results showed that only 5 individuals in the validation population were incorrectly predicted. Sample numbers 117-3, 121-5 and 126-2 which come from improved families and are authorized improved varieties, were predicted as unimproved individuals in our model. Sample numbers 154-1 and 16-3 which were authorized unimproved varieties, were predicted as improved individuals(Supplementary Table 3). The error rate plot(Figure 4b) demonstrates the performance of the Random Forest model as the number of trees increases. The overall OOB error rate (black line) decreases initially and stabilizes as more trees are added, indicating improved model accuracy. The ‘Unimproved variety’ class(green line), shows a lower and more stable error rate compared to the ‘Improved variety’ class(red line). This result indicates that our model performs more consistently when detecting unimproved individuals.

The feature importance plot(Figure 4c) reveals that the top contributing SNPs in classification. SNPs are ranked by their mean decrease in Gini index, with higher values indicating greater importance in reducing classification error. As results, the validation accuracy of our selected candidate SNPs in the validation population reached 98.64% and we identified several SNP markers that made significant contributions to the classification of the samples(Figure 4c).

The result of PCA(Figure 5) illustrates the clustering of all samples based on 50 SNPs, highlighting a clear distinction between improved and unimproved varieties. While most individuals can be successfully distinguished, a small subset of individuals from the validation population were difficult to distinguish, which represented by red triangles (improved varieties) and blue circles (unimproved varieties). The clustering for unimproved varieties is particularly effective, which aligns with the lowest error rate observed for unimproved individuals in the Random Forest model(Figure 4b). These results suggest that while the Random Forest-based approach is useful for identifying improved varieties, it lacks the precision needed for accurate classification within the improved variety group.

Ultimately, we constructed DNA fingerprinting using these 50 SNPs for the 29 recognized improved varieties(Figure 6) and 77 improved individuals in the validation population(Supplementary Figure S3).

4. Discussion

4.1. An Application of the 51K Liquid-Phased Probes for Slash Pine

Slash pine has been widely cultivated in southern China, and has become an important species for afforestation since its introduction [30,31]. Fingerprinting provides core resources for the selection and application of marker-assisted breeding, thereby accelerating the breeding process [32]. The lack of complete genome information on slash pine [9,33,34] prompted us to develop a 51K liquid phase probe for rapid genotyping of individuals. This probe has been successfully applied in the breeding programs of slash, loblolly, and Caribbean pines [23]. In this study, we selected 29 improved varieties of slash pines to construct a DNA fingerprint. A total of 560,567 SNP loci were genotyped for these 29 improved varieties using target sequencing of the probe, achieving a genotyping rate of 94.61%. Ultimately, we constructed a DNA fingerprint for improved slash pine varieties with the candidate SNPs. However, due to the absence of a complete genome, we could only map the SNPs to scaffolds rather than directly to chromosomes. Although these candidate SNPs are already distributed across different scaffolds or genes, we cannot ascertain their even distribution across the 12 chromosomes of pine. This limitation may affect the precision of SNP selection and the accuracy of DNA fingerprinting predictions.

4.2. The Construction Methods of DNA Fingerprinting for the Improved Varieties of Slash Pines

Molecular markers are crucial for constructing DNA fingerprinting profiles. SSR markers are currently the most widely used molecular markers [35]. Numerous studies have examined the genetic diversity of forest tree populations and constructed fingerprint databases for improved varieties using SSR markers, such as Populus spp.[36], Xanthoceras sorbifolia [37], Sophora japonica [38] and Phellodendron amurense [39]. However, SSR markers are disadvantaged by their time-consuming nature and limited quantity, rendering them unsuitable for large-genome species such as pine. In recent years, the advancement of sequencing technology has facilitated the widespread use of SNP-based DNA fingerprinting for genotype identification and population genetic diversity analysis in crops and horticultural plants [4,40,41]. In forestry, economic crops like oil tea(Camellia oleifera)[42] and tea(Camellia sinensis)[43] have utilized SNP markers to construct DNA fingerprints, although their application remains limited.

In research related to agricultural crops and forestry, the random forest algorithm is widely utilized for plant classification [44,45] and variety identificatio [46,47,48,49]. Furthermore, when integrated with remote sensing imagery, this algorithm proves valuable for detecting plant diseases [45] and predicting yields [50,51]. The random forest algorithm holds considerable potential, offering the ability to enhance the efficiency and accuracy of target identification.

Slash pine has been planted in China for nearly a century, but the management of elite variety resources is still lacking. This study establishes a DNA fingerprinting and technical procedure for identifying improved slash pine varieties, which helps protect the intellectual property of improved varieties. Initially, we screened 3502 core SNPs representing 29 improved varieties by controlling for SNP polymorphism parameters. Subsequently, employing a random forest algorithm, we reduced the SNP count to 50 and validated their identification rate in a mixed population of improved and unimproved varieties, achieving a validation rate of 98.64%.We obtained results comparable to the previously calculated average Het, MAF, and PIC using the 51 K liquid-phase probes, indicating that the 50 candidate SNPs selected in this study are informative and highly representative. As more improved varieties receive national approval and authorization, we will continually expand the sample size in the fingerprint database to build a more accurate and efficient identification platform for improved slash pine varieties. Additionally, further optimization of model parameters and feature engineering will be essential to enhance identification accuracy.

5. Conclusions

In this study, we employed a 51K liquid-phase probe to detect SNPs in 29 improved slash pine varieties. Through a rigorous filtering process and random forest algorithm, we identified 50 SNPs for constructing a DNA fingerprint map. These SNPs were validated in a population consist of improved varieties and unimproved varieties, achieving an impressive 98.64% accuracy in variety identification. Evaluations of heterozygosity, polymorphism information content, and minor allele frequency using these 50 SNPs further confirmed their reliability. Our research has established a precise and reliable DNA fingerprint, poised to significantly enhance the rapid screening of superior slash pine germplasm in future breeding programs.