Population structure and genetic diversity in Eucalyptus pellita based on SNP markers

Eucalyptus pellita has the characteristics of rapid growth and high resistance. However, there is little research on molecular breeding of E. pellita, which is essential to shortening breeding life and selecting quality varieties. Therefore, a crucial step before selective breeding can be carried out to increase the wood quality of E. pellita is identifying genetic diversity and population structure using single nucleotide polymorphism (SNP) markers. In this study, the genetic diversity of 1st generation 196 E. pellita families from 23 geographically defined was assessed using 1,677,732 SNP markers identified by whole genome resequencing. SNP annotation showed that the ratio of non-synonymous to synonymous coding mutations was 0.83. Principal component analysis (PCA), phylogenetic tree, and population structure analysis permitted the families to be categorized into three groups, one of which (G2) contains most of the Indonesian (IDN) and Papua New Guinea (PNG) families. Genetic relationship analysis showed that IDN was closely related to PNG. Genetic diversity analysis showed that He, PIC, I, and H mean values were 0.2502, 0.2027, 0.3815, and 0.2680, respectively. PCA analysis classified various provenances in QLD into two categories (G1 and G3). The genetic diversity of G3 was higher than that of G2. The results of genetic differentiation (Fst) showed that PNG region was divided into two groups (PNG1 and PNG2), the Fst (0.172) between QLD and PNG2 region was higher than QLD and PNG1, and the Fst (0.024) between IDN and PNG1 is smaller than IDN and PNG2. A Mantel test revealed a positive correlation between the genetic and geographic distance of E. pellita. This study has a certain reference value for genetic identification, germplasm preservation, and breeding of E. pellita. Also, it provides a basis for subsequent association analysis to explore excellent alleles and introduction.


Introduction
Eucalyptus pellita, a fast-growing tree with strong disease resistance and a high survival rate, has been introduced and cultivated in numerous countries, such as China, Brazil, and Western Samoa.E. pellita thrives after introduction since the cultivated climate is comparable to that of the native distribution area.Currently, Australia (Brawner et al., 2010), Indonesia (Leksono et al., 2008), and Vietnam (Harwood and Nambiar, 2014) have advanced E. pellita seed orchards to the second or even the third generation.The average annual yield of E. pellita is 16-18 m 3 /ha in Sumatra, Indonesia, while the average annual yield of Eucalyptus in China is 15-28 m 3 /ha (Harwood and Nambiar, 2014).Before 2014, Eucalyptus plantations in China were in a developmental and exploratory stage.Many individuals and investors recognized the economic benefits of Eucalyptus and ventured into Eucalyptus forestry.However, the problem of poor Eucalyptus varieties also limited Eucalyptus yields.According to data from 2009 to 2018, domestic Eucalyptus annual yields have increased rapidly, reaching as high as 39.43 m 3 /ha.The main reason for this increase is that China began to place significant emphasis on the development of artificial Eucalyptus forests.
E. pellita is a humid and subtropical forest species.It has two naturally occurring regions: southern New Guinea (NG) and North Queensland (QLD).E. pellita was first discovered in 1864 by John Dallachy at Rockingham Bay, south of Innisfail in Queensland, and described in the same year by Baron Ferdinand von Mueller (Harwood, 1998).By the end of the 1980s, E. pellita was thought to be limited to Australia, with widespread distribution in northern QLD and New South Wales.Similar to E. pellita, Eucalyptus Scias is found in Lanzhou, New South Wales, however, its descriptions of leaves, buds, and fruits are very different from that of the species (Johnson and Hill, 1990).The E. pellita of Cape York in northern QLD are similar to the NG population and in some respects have fewer fruits and leaves than the southern Australian population.Further research may admit that the E. pellita populations in NG and Cape York in QLD originated independently of each other (Harwood, 1998).However, few studies on genetic diversity and population structure of E. pellita have been reported, which is of great significance in distinguishing the relationship between the three provenances.
Initially, a large number of introduced species and highgeneration improvements were carried out in Brazil, Southeast Asia, and other countries, with the primary goal of studying the genetic variation of growth, adaptability, wood characters, disease resistance, and genetic gain between generations (Brawner et al., 2010).The majority of the genetic materials investigated were from natural sites, and phenotypic traits analysis effectively distinguished between QLD and NG provenances.In plantings of E. pellita in eastern Colombia, the provenance of NG performed noticeably better than that of QLD (Nieto et al., 2016).Similarly, when planted in humid tropical environments, the NG outperformed the QLD in terms of survival rate, growth rate, and morphological characteristics (Harwood et al., 1997).In addition, the internal provenance of QLD and NG was also different.For the survival study of introduction, E. pellita was introduced to Urbano Santos in Brazil.The survival rates of Northeastern Coen and South Helenvale at 3.5 years of age were 29.8% and 43.8%, respectively (Harwood, 1998).Although the geographical location of the various provenances is established, the necessity for genetic relationships between provenances remains to be determined.
In China, E. pellita is mostly utilized as a hybrid parent to develop excellent Eucalyptus clones and to conduct theoretical research on cross-breeding for fast growth, disease resistance, and insect pests.Long-term artificial selection and domestication also introduce a slew of issues, the most prominent of which is the limited genetic diversity of breeding materials, which severely homogenizes variations and makes it harder and harder to produce novel, ground-breaking varieties (Zhang et al., 2012).The selection of parents in cross-breeding is aided by knowing the genetic background of the breeding materials; this increases the effectiveness of producing new varieties with evident heterosis.
Genetic diversity plays an important role in heterosis and breeding programs.Therefore, the genetic diversity within and between Eucalyptus populations is routinely assessed using different marker techniques such as morphological (Byrne et al., 2016), biochemical (Kirst et al., 2005), and molecular markers (Shang et al., 2019).Morphological markers have been extensively utilized to assess genetic diversity because they are inexpensive, fast, and easy to measure.They are, however, highly influenced by the environment, and several other factors limit their ability to estimate genetic diversity (Yaman, 2021).Because molecular markers are stable, polymorphic, easily obtainable in the genome, and insensitive to environmental factors, they are still valuable tools for measuring genetic diversity (Govindaraj et al., 2015).Therefore, molecular markers, including RFLP, RAPD, ISSR, SSR, and SNP based on single nucleotide differences, are the most ideal methods to explain biological genetic diversity.
The study of genetic diversity and population structure is essential in identifying genetic relationships among germplasm resources.Plant improvement initiatives benefit from genetic heterogeneity among populations as well as genetic relationships between them.Identification of populations with high levels of genetic variation will be a valuable resource for broadening the genetic base because it makes it possible to identify good alleles for traits (Yang et al., 2020;Aesomnuk et al., 2021).To comprehend and use accessible gene bank resources, a variety of approaches can be employed to identify E. pellita genetic diversity.For example, early isoenzyme markers technology discovered considerable genetic variations between NG and QLD provenances, with NG provenances having much lower heterozygosity (House and Bell, 1996).Similarly, RFLP was also used to evaluate the genetic resources of Eucalyptus, though the results were stable, reliable, and repeatable (Moran et al., 2000).However, it has some drawbacks, such as complicated operation, long duration, high cost, and large demand for DNA.
So far, studies on E. pellita based on SNP markers are scarce, and earlier techniques are prohibitively expensive and may yield inaccurate results.Therefore, SNP markers are being promoted with the advent of next-generation sequencing (NGS) technology.SNP molecular markers have been widely used to study genetic diversity due to their richness, wide genomic coverage, availability of neutral variation and selected loci, rapid and high-yield genotyping, and low error rate (Helyar et al., 2011).Based on SNP markers, complex traits were anatomized by QTLs (Quantitative traits Loci) or LD (linkage disequilibrium) analysis to provide a basis for Eucalyptus breeding (Resende et al., 2017;Butler et al., 2022).
Whole-genome resequencing is the process of sequencing the entire genome sequence of different individuals or species with known genome sequence, to analyze the differences between different populations or individuals (Risch and Merikangas, 1996).A large number of SNP, InDel, and SV can be discovered by comparing the sequenced sequence of a single individual with the known genome sequence of that species or related species (Catanach et al., 2019).In this study, 196 E. pellita resources from New Guinea and Australia were sequenced using whole-genome resequencing technology, and SNP sites were detected, screened, and typed.Genetic diversity, population structure, and genetic differentiation of these materials were further analyzed.It will lay the foundation for conservation and utilization of Eucalyptus resources, gene mapping of important traits, polymeric breeding based on molecular markers and further genome-wide association analysis.

Germplasm
The provenances of the 1 st generation of E. pellita breeding populations are from Queensland (QLD, AUS) and New Guinea (NG) Island, which are separated by the ocean with 150 km as the closest shoreline.Generally, the provenances of NG are divided into Indonesian provenances (IDN) and Papua New Guinea provenances (PNG).QLD is divided into Cape York provenances and northeast Queensland provenances.In this study, 196 E. pellita germplasm resources were collected, including 5 from IDN, 55 from PNG, and 133 from QLD.All seeds were randomly planted in Fusui, Guangxi, China, and Suixi, Guangdong, China.Each family was divided into 23 provenances according to geographical location, as detailed in Supplementary Table S1.After gathering young leaves from the top canopy of surviving trees in mature forests, the leaves were swiftly placed in an incubator with ice for sample and sealed in a zippered bag containing silica gel.The samples were brought indoors and stored in the refrigerator at -24°C.

DNA extraction
The CTAB method was used to extract DNA (Gan et al., 2003).Use Nanodrop2000 Spectrophotometer (Thermo Fisher Scientific) to determine the concentration and quality of the total genome DNA.DNA libraries with a mean insert size of 350 bp were constructed, and 125-bp paired-end reads were generated using an Illumina HiSeq 4000 instrument.Library preparation and sequencing were carried out at the Biomarker Technologies Corporation (Beijing, China).

SNP and InDel calling
Illumina platform was used for sequencing, and Raw image data files were identified by CASAVA bases to form Raw reads.Subsequently, Clean reads were obtained through quality control analysis, base quality distribution analysis, and sequencing data filtering.BWA software (Li and Durbin, 2009) was used to compare Clean reads with E.grandis genomes (https://www.ncbi.nlm.nih.gov/assembly/GCF_016545825.1), and the results were formatted by SAMTOOLS software (Li et al., 2009), and then reweighted by PICARD software (http://broadinstitute.github.io/picard/;v1.94).The ratio of pairs, genome coverage distribution, insert-size distribution, and variation were also analyzed.SNPs and InDels within the 196 families were called using the HaplotypeCaller module in GATK (McKenna et al., 2010).They were filtered with the following parameters: QD < 2.0 || MQ < 40.0 || FS > 60.0 || QUAL < 30.0 || MQrankSum < -12.5 || ReadPosRankSum < -8.0 -clusterSize 2 -clusterWindowSize 5. SNPs with minor allele frequency (MAF) of lower than 5% in the population were filtered out.Software SnpEff (Cingolani et al., 2012) was used to annotate variation and predict the impact of variation.The position of mutation locus on the reference genome can be obtained by SnpEff analysis and functional annotation.

Population structure analyses
Based on the genetic data of the population, the genetic distance was calculated and phylogenetic tree was constructed based on the distance matrix.MEGA5.2 (Tamura et al., 2011) software was used to construct phylogenetic trees for each sample, with the Kimura 2parameter model, 1000 bootstrap replicates, and neighborjoining method.
EIGENSOFT was used for principal component analysis to obtain sample clustering (Price et al., 2006).Principal component analysis (PCA) analysis can be used to know which sample relationships are relatively close and which sample relationships are relatively distant and can play an auxiliary role in evolutionary analysis.
Admixture software was used to analyze and study the population structure (Alexander et al., 2009).The pre-set subgroup number (K value) of the population was clustered, with a range from 1 to 10.The optimal clustering number was determined according to the minimum value of cross-validation error rate.
GCTA software was used to estimate the genetic relationship between two individuals in natural populations (Yang et al., 2011).In this study, the mean of the expected variances of SNP markers was used to correct the labeled expectation variances, namely the Gmatrix, to obtain a heat map of kinship.

Genetic diversity and LD analysis for 23 provenances
Nei's genetic diversity index (H), polymorphic information content (PIC), minor allele frequency (MAF), expected heterozygosity (He), and observed heterozygosity (Ho) of the 23 provenances and inferred groups were calculated by PowerMarker V3.25 (Liu and Muse, 2005).PopGenome R package (Pfeifer et al., 2014) was used to calculate the nucleotide polymorphism (Pi) and Tajima' D for 23 provenances.GenAlEx V6.5 (Peakall and Smouse, 2006) was used to calculate the pairwise genetic differentiation coefficient (Fst) between provenances to study population differentiation.PopLDdecay (Zhang et al., 2019) was used to evaluate LD value and plot LD decay for each chromosome.

AMOVA analysis, Mantel test, and genetic difference for three population
Based on Qual > 30,000 in the original data, 12,475 SNP markers were selected for subsequent analysis.R language was used to generate SNP format 0,1,2 from the filtered data, where 0 represented homozygote reference,1 represented heterozygote, and 2 represented homozygote substitution.The GenAIEx V6.5 software (Peakall and Smouse, 2006) was used to perform AMOVA to divide the total level of genotypic variance into variance within and between populations.The Mantel test was used to correlate the genetic distance and geographic distance matrices between populations by GenAIEx V6.5 software.In addition, these filtered SNPs were used to perform Fst and Nei's genetic distance analysis for three populations by GenAIEx V6.5 software (Fufa et al., 2022).chromosomes; for instance, chromosomes 1 and 7 exhibited a concentration of differentiation at their ends, whereas other chromosomes displayed uneven variations.

Population structure analysis
According to integrity > 0.8 and MAF > 0.05, all the above SNP markers were filtered, and a total of 1,677,732 SNPs with high consistency were obtained.Based on the SNPs filtered, admixture software was used to analyze the population structure of the samples, and the sample population number (K value) was assumed to be 1-10 for clustering (Figure 3A).According to the valley value of cross-validation error rate, the optimal population number was 3, indicating that 196 E. pellita families can be divided into three groups, which means they come from different original ancestors.The clustering with K values ranging from 1 to 10 and the cross-validation error rate corresponding to each K value is shown in Figure 3B.The population structure divided the families into Q1-Q3 (Supplementary Table S2), and most of the families from PNG and IDN were assigned to Q1. Families from the QLD region included 85 in Q2 and 47 in Q3.
As can be seen from the kinship heat map of E. pellita group (Figure 3C), as a natural group, the kinship between the families of the 1 st generation of E. pellita was not close and their kinship value was mostly low.However, the families in the upper left corner of the figure were relatively close, and most of these families come from the PNG region.In the white-blue shade in the upper left corner, PNG had 45 families, IDN had 4 families, and QLD had 2 families (Supplementary Table S3), which meant that the genetic distance between these 51 families was close.It also implied that IDN was closely related to PNG region.
To further verify the accuracy of the above results, principal component analysis and cluster analysis were performed on SNP data.According to the geographical location of each family (Figure 4A) and the distribution of each family in PCA (Figure 4B), they could be divided into 3 groups, namely, group 1-3.196 families from 23 provenances clustered into 3 groups and only provenances S14339 were relatively discrete and did not belong to the 3 groups, indicating that this population mainly came from 3 branches.For example, the provenances from IDN and PNG belong to Group 2 and are relatively independent from other provenances.The families of Group 2 are concentrated in NG Island (5°S-10°S), and the provenances from QLD are also divided into two groups according to latitude.Group 3 is located between 15°S and 17°S, and Group 2 is located in an area above 17°S (Figure 4A).Provenances from different regions have specificity and can be distinguished, and the relationship between provenances and SNP can be better understood by combining the traits of provenances from the three regions.
To better understand the clustering of each family, we constructed a phylogenetic tree for 196 families (Figure 5).From the roots of the developing tree, all provenances can be divided into three large clades (Group A-C).As mentioned above, PNG and S18774, and A18750.Most of the families of Group A branch were located in QLD, while most of the families of Group B originated from northern QLD.

Genetic diversity analysis
According to the genetic diversity parameters of the firstgeneration E. pellita population (Table 1), the overall expected heterozygosity of all provenances was 0.2502.The expected heterozygosity of QLD, PNG, and IDN were 0.2547, 0.2428, and 0.2232, respectively.The provenance with the highest expected heterozygosity was A18197 while the smallest provenance was S16120 with little difference among provenances.The observed heterozygosity of provenances A17854, A17860, A18314, A18596, A18598, A18774, and A20659 was higher than the expected heterozygosity, suggesting that these populations may have experienced distant hybridization.The Nei's diversity index, Shannon-Wiener index, and polymorphism information content of all provenances were 0.2680, 0.3815, and 0.2027, respectively.They all follow the same pattern, with the highest values in the QLD region, followed by the PNG region, and the lowest values in the IDN region.Overall, the genetic diversity is highest in the QLD region and lowest in the IDN region.The highest genetic diversity was A18197 and A18599, and the lowest was A17854 and S16120.
In the population predicted by PCA (Figure 4), the genetic diversity of Group 3 was greater than that of Group 1, and that of Group 1 was greater than that of Group 2, of which Group 1 and Group 3 were from the QLD area.In addition, the average nucleotide polymorphism across all sources is 0.000669, indicating that the 23 germplasm sources from the three major regions have not undergone artificial selection.This is consistent with their natural population attributes.Tajima's D test analysis also suggests the presence of a higher frequency of low-frequency variants within various provenances, possibly due to the imbalance caused by genetic drift or selection.

Population differentiation analysis
The population differentiation analysis of the 1 st generation of E. pellita population (Table 2) showed that the average Fst of IDN and PNG was only 0.039, while the average Fst of IDN and QLD was 0.123, indicating that the differentiation between IDN and QLD was greater than that between IDN and PNG.Based on the geographical distribution of the three regions (Figure 3A), we speculated that QLD was the origin of E. pellita, which spread from QLD to PNG and then to IDN.
According to the Fst results, PNG was divided into PNG1 and PNG2.The Fst between PNG1 (including provenances A18197 and A18199) and QLD was 0.067 and 0.102, respectively, with an average of 0.085, while the between PNG2 (including provenances A20659, S16120, S16121 and S16122) and QLD was 0.175, 0.178, 0.173 and 0.161 respectively, with an average of 0.172, which is twice as much as that between PNG1 and QLD, indicating that the relationship between PNG and QLD wasn't single.If PNG provenances were only transmitted from QLD, there would be no significant difference between them and the Fst of QLD.Combined with the speculation of Harwood (1998), who observed the characters of E. pellita from various sources and speculated that PNG and QLD were two independent populations, we concluded that QLD and PNG2 were two independent origins of E. pellita, and that PNG1 was transmitted from QLD. IDN is transmitted from PNG, but it is more likely to be transmitted from PNG1 based on the perspective of Fst coefficient.
The alternative hypothesis holds that PNG2 is the only source, which then spreads to PNG1 and QLD.In this case, the average Fst of PNG2 and QLD, as well as the average Fst of PNG2 and PNG1, are not that different.However, this theory is invalid since PNG and QLD are independent based on the observation of phenotypic traits.The value range of Fst is 0-1.The maximum value of 1 indicates that the populations are completely differentiated, while the minimum value of 0 indicates that there is no differentiation between the two populations.It is generally believed that genetic differentiation among populations is very small when Fst is 0-0.05.Between 0.05 and 0.15, there was moderate genetic differentiation among populations.At 0.15 to 0.25, there was a large genetic differentiation among populations.There was large genetic differentiation among populations above 0.25 (Weir and Cockerham, 1984).According to the average Fst values of the first-generation population from all major sources in this study, there was a large differentiation between QLD and PNG2, a moderate differentiation between QLD, PNG1, and IDN, and a small differentiation between other sources.

Linkage disequilibrium analysis
Association maps based on linkage disequilibrium (LD) provide a powerful strategy for genetic profiling of complex quantitative traits.The distance to threshold decay (R 2 ≈ 0.2), and halfmaximum decay (R 2 = 0.1) was calculated within all contigs and all samples in a global analysis, with genome-wide LD decaying to the threshold level within 5.8098 kb, to half-maximum within 33.6428 kb.In addition, the distance of chromosomal average LD decay to the threshold level ranged from 4.5364 to 7.1236 kb, and the distance to the half-maximum ranged from 18.7877 to 53.0599 kb (Figure 6).Among these chromosomes, Chr05 showed a faster LD decay, while Chr01 showed a slower LD decay.

Genetic relationship analysis of E. pellita population
Table 3 shows the results of AMOVA for the 196 E. pellita using 12,475 SNP markers.The results showed that the genetic variation among population accounted for only 10% of the total variation, while the genetic variation within the populations accounted for 90%.
All Fst comparisons between populations showed the values between PNG and QLD were higher than others, while Nei's genetic distance values between IDN and QLD were higher than others (Table 4).In general, Nei's genetic distance is more suitable for the study of genetic diversity within a population.Thus, a smaller genetic difference (Fst = 0) between IDN and QLD indicates the presence of higher gene flow or similarity in gene frequencies.
A mantel test was performed to obtain a correlation coefficient between genetic distance and geographic distance of E. pellita.The SNP marker-based correlation coefficient was Rxy 0.09 (P < 0.05), indicating a positive correlation between genetic distance and geographical location.The provenances of PNG were divided into PNG1 and PNG2, of which 2 provenances with * were PNG1 and 4 provenances with ** were PNG2.One provenance of PNG1 and PNG2 was respectively listed in the table with the Fst of QLD.

SNP annotation
In the coding region, the ratio of non-synonymous to synonymous coding mutations was 0.83, which was similar to that of Arabidopsis thaliana (0.83) (Clark et al., 2007), but lower than that of Cajanus cajan (1.18) (Varshney et al., 2017) and Chinese plum (1.32) (Wei et al., 2021), indicating that there were fewer mutations causing protein changes in the coding region compared other species.The ratio of non-synonymous to synonymous coding mutations is less than 1, suggesting that negative selection is an important evolutionary force affecting E. pellita.

Analysis of E. pellita population structure
The genetic structure of E. pellita was analyzed using various methodologies including population structure analysis, PCA, and phylogenetic tree construction, thereby providing complementary information.The 1 st generation of E. pellita families from three sources: QLD, PNG, and IDN.According to the analysis results, the 23 1 st generation provenances can be divided into three large groups.While the division of these groups is not fully compatible with the three large E. pellita sources, the families that come from the same source are essentially able to get together.This rule was also supported by the findings of the genetic distance calculation, which showed that there was a significant genetic distance between provenances in different locations and that there was a correlation between genetic distance and geographic position.Similarly, the clustering results of E. moluccana and its putative subspecies are not entirely consistent with geographical distribution (Flores-Renterıá et al., 2021).

Analysis of genetic diversity
Genetic diversity analysis is of great significance to the evaluation and utilization of plant germplasm resources and the breeding of new varieties.It is also an important component of Linkage disequilibrium for the E. pellita chromosomes.Different chromosomes were represented in different colors, and the enlarged image was used to clearly show the decay rate of different chromosomes.
plant genetics, breeding, conservation, and evolution.In recent years, the majority of studies on the genetic diversity of Eucalyptus plants have primarily focused on SSR molecular markers, with a lack of emphasis on SNP markers.However, it is important to note that different molecular marker technologies yield varying results.For instance, when using SSR molecular markers, the He value for E. Cloeziana is 0.682 (Lv et al., 2020); whereas when utilizing SNP molecular markers, the He value ranges from 0.2677 to 0.3487 for E. urophylla (Yang et al., 2020).Similarly, the genetic diversity value (He) of SSR markers was significantly higher than that of SNP markers in palm trees (Jara-Arancio et al., 2022).These findings highlight the influence of different molecular labeling techniques on parameter values.
It is generally considered that PIC ≥ 0.5 is highly polymorphic, 0.25 ≤ PIC < 0.5 is moderately polymorphic, and PIC < 0.25 is low polymorphic (Botstein et al., 1980).Previous studies based on SSR markers showed that the PIC values of E. urophylla (Lu et al., 2018), E. camaldulensis and E. tereticornis (Arumugasundaram et al., 2011) were highly polymorphic (greater than 0.5).In this study, the mean values of total He and PIC of the 1 st generation of E. pellita were 0.2502 and 0.2027, indicating a moderately low polymorphism according to Botstein's theory (Botstein et al., 1980).However, the PIC obtained by SNP markers is at a low level, which may be because each SNP is bi-allelic in nature.Theoretically, the maximum PIC of a single SNP is 0.5 (Eltaher et al., 2018).Similarly, other species studied based on SNP markers had lower PIC, such as corn (0.19) (Adu et al., 2019), cowpea (0.27) (Seo et al., 2020), sorghum (0.24) (Silva et al., 2020).In conclusion, the genetic diversity of E. pellita in this study is theoretically not low.
In addition, the difference between the observed heterozygosity and the expected heterozygosity reflected the rationality of SNP markers selection and the genetic diversity of population (Liu et al., 2022).In this study, the He and Ho difference of the three origins (QLD、IND、PNG) of the first-generation population was small, which could accurately estimate the genetic structure of the E. pellita population, indicating QLD had a relatively high genetic richness than others.

Analysis of genetic differentiation
Through to the 1 st generation group differentiation index (Fst) analysis, this study for E. pellita has two natural origins (QLD and PNG) that provided evidence, at the same time, this study also continued to develop the inference, considered one of the big provenances PNG1 spread by big provenance QLD, this assumption is supported by other index, for example, genetic distance.Population differentiation index is one of the important indexes to analyze the origin of species.Many scholars use this index to analyze the origin of species.According to the analysis of Cajanus cajan (Varshney et al., 2017), South Asian and sub-Saharan African populations had the lowest Fst values (0.102) compared to South Asian and South American populations (0.126) and South Asian and Central American populations (0.167).Fst values correlated with geographical distance between populations.These Fst values suggest that dal spread from South Asia to sub-Saharan Africa and finally to South and Central America.
The average Fst of the 1 st generation natural E. pellita population was 0.0913, while the Fst of the 2 nd and 3 rd generations of E. urophylla based on SSR markers was 0.03 (Lu et al., 2018).It can be seen that the Fst of natural populations of E. pellita was in the middle level.Similarly, there were Eucalypts with low genetic differentiation index, for example, E. camaldulensis with Fst values of 0.044-0.065(Butcher et al., 2002), E. globulus was only 0.08 (Jones et al., 2002), E. marginata was 0.045 (Wheeler et al., 2003).But some of the other Eucalyptus species had higher Fst values, for example, the Fst of E. curtisii was 0.30 (Smith et al., 2003), E. angustissima was 0.136 (Elliott and Byrne, 2004).E. morrisbyi was 0.19 (Jones et al., 2005).

Genetic relationship analysis of three population
The number of SNPs determines the accuracy and reliability of the results.A dataset consisting of 12,745 SNPs was employed for analysis.In comparison to recent studies on other tree species, such as avocado (384 SNPs) (Wienk et al., 2022) and palm (1038 SNPs) (Jara-Arancio et al., 2022), a larger number of filtered SNPs was used in this study, despite their original datasets containing millions or even billions of SNP loci.Thus, differences in analysis results between populations can be attributed to different filtering parameters applied to SNPs.
Considering the filtered SNP loci, AMOVA reveals that 90% of the variation is within populations.This percentage is notably higher compared to other tree species.For instance, in the case of the Fujian tea tree, 66% of the variation is within populations, but these samples have a closer geographic distribution (Liu et al., 2022).In this study, particularly for the QLD population, samples from different provenances span a larger geographic range, leading to greater differences within populations.This highlights a certain correlation between the geographic distribution of provenances and population variation.
To validate the potential association between different populations, Fst and Nei's distance were employed to assess genetic differences between populations.In this study, Fst values between PNG and QLD were relatively high, while Nei's distance values between IDN and QLD were higher.This suggests that different analysis methods can yield distinct results.This phenomenon has been observed in other species as well (Wienk et al., 2022;Osundare et al., 2023), possibly attributed to either lower levels of population differentiation or the presence of complex genetic structures.In addition, the Mantel analysis results indicated a moderate correlation between the molecular matrix and the geographic matrix, with a Mantel test value of only 0.09 in this study.For comparison, the Pinus bungeana population in central China showed a significantly higher Mantel test value of 0.6 (Tian et al., 2022), and the rice population in northern India exhibited a value of 0.525 (Choudhury et al., 2023).A commonality in these cases is that the samples have relatively concentrated local distributions.As previously reported (Tian et al., 2022), introducing a more dispersed population would lead to a decrease in Mantel test values.Therefore, despite the broad geographic distribution of E. pellita, there is still a discernible correlation with genetic distance.

Conclusion
Although the provenance differences of the 1 st generation of E. pellita have been indirectly studied by predecessors, the genetic diversity, population structure and genetic differentiation based on SNP markers in E. pellita have not been studied.In this study, we conducted rigorous whole-genome resequencing of 196 families of E. pellita.A series of analyses were carried out based on the quality screening SNP, and it was concluded that the 1 st E. pellita population could be divided into three groups.The IDN and PNG provenances were classified into one group, and the QLD provenances were divided into two groups, and the QLD provenances had high genetic diversity.Our analysis of genetic differentiation reveals that QLD was the origin of E. pellita.Analyzed 12,475 screened SNPs, showed that the genetic variation among the population accounted for only 10% of the total variation, while the genetic variation within the populations accounted for 90%.The results of this study help to clarify the population structure and provide a valid inference for the population origin of E. pellita.The results lay the foundation for genome-wide association analysis.

FIGURE 2 SNP
FIGURE 2 SNP and InDel distribution plot for genotyping-by-sequencing of 196 E. pellita accessions.SNP and InDel are represented by orange and blue bars.

3
FIGURE 3 Population structure and genetic relationship.(A) In each population structure, each individual was represented by a line of different colors, and which subgroup the variety belonged to was inferred according to the proportion of colors; (B) Cross-validation error rate for each K value; (C) The smaller the value of kinship between the two samples, the darker the blue shaded part.

TABLE 1
Genetic diversity parameters of various provenances and inferred groups of E. pellita.

TABLE 2
The 1 st generation of E. pellita group as the major source of differentiation index (Fst).

TABLE 4
Pairwise PhiPT values (below diagonal) and Nei's minimum genetic distance (above diagonal) between populations within region for 196 E. pellita assessed using GenAlex software.