Genotyping-by-sequencing of Illicium difengpi highlights its potential genetic diversity and conservation status

Abstract

Illicium difengpi is a critically endangered plant endemic to the karst mountains of Guangxi, China. Because of its medicinal and ecological values, many efforts were made on assessing the medicinal usage, natural history, ex-situ conservation, and stress tolerance of I. difengpi. Due to its extensive genome, whole genome sequencing has not been completed, and there has been limited research on its population genetics. Using genotyping-by-sequencing, we analyzed the single nucleotide polymorphism of 43 samples from 13 different sites, encompassing all the natural habitats of the species. The results showed that I. difengpi has low genetic diversity, which is exemplified by low heterozygosity, polymorphism information content, nucleotide diversity, and small effective population size (80–90). Also, the fixation index indicated that most subpopulations were highly differentiated, presumably due to their geographic isolation. This is a warning that I. difengpi is in real and immediate danger of extinction. Moreover, evolutionary relationship, principal component analysis, and Admixture all suggest that there is an obvious correlation between the geographic and genetic characteristics of I. difengpi. According to our results we suggest conservation strategies to safeguard its sustainability in the future.

Introduction

Illicium difengpi K.I.B et K.I.M. (Fig. 1A) is an endangered medicinal plant endemic to Guangxi Zhuang Autonomous Region, China. This perennial shrub grows almost exclusively on top of the rocky mountains of the karst landscape, with occasional distribution on hillside slopes of these areas (Fig. 1B). In local traditional medicine, the bark of I. difengpi (Illiciaceae, Fig. 1C) is used for treating the symptoms of rheumatic arthritis, muscle pain or fatigue and traumatic injury. Due to its medicinal value, I. difengpi is excessively harvested, as the karst mountains are unsuitable for farming, and the herbal trade helps alleviate poverty. Consequently, several medicinal plants have gone extinct in recent decades¹, and we must make every effort to prevent I. difengpi from suffering the same fate.

Fig. 1.

Morphology, habitat, and medical usage of I. difengpi. (A) Leaves and premature fruits. (B) A typical habitat of I. difengpi. Note the rocky mountain top it grows on. (C) Barks of I. difengpi are used in traditional medicine.

Apart from its medicinal value, I. difengpi plays an important role in karst ecology. Though karst areas in southern China receive a plentiful supply of rain, the rocky surface cannot hold water without vegetation coverage. This dry and barren rocky surface is hostile to most shrubs, but happens to be the favorite habitat of I. difengpi. In addition, this species can also resist high salinity and heat, which makes it ideal for studies related to stress tolerance as well as soil erosion prevention². However, its native population has retreated to western Guangxi, with some insignificant relicts in Guangdong, Yunnan, and Vietnam. For decades, I. difengpi remains on the National Key Protected Wild Plant List (Category II) of China with the label “Endangered”. The in-situ conservation of I. difengpi primarily includes legislations in habitat protection and market regulations, while the ex-situ conservation in Northeast Guangxi achieved initial success, with dozens of individuals transplanted and bred¹. However, due to its slow reproduction rate, it will be difficult for the population of I. difengpi to recover in the near future.

Generally, endangered species face the threat from mutational meltdown³, because a small population size can lead to intensified inbreeding, accumulation of deleterious alleles, and even extinction. Therefore, knowledge of genetic diversity is critical for protecting endangered species to guide conservation strategies. Although a method using a polymerase chain reaction (PCR) system by orthogonal design was developed^2,4, its population genetics remains largely unexplored, including the effective population size (N_e), F-statistics, heterozygosity, and other related parameters. The genus Illicium has an exceptionally large genome relative to its sister taxa (~ 25pg for 2 C DNA)⁵ and the whole genome sequence of any Illicium species has not yet been completed, meaning there is no reference genome available for alignment or comparison. This problem can be resolved by the recently developed method of genotyping-by-sequencing (GBS)⁶, which is capable of identifying single-nucleotide polymorphisms (SNPs) across the whole genome with low cost and high resolution in order to directly reflect population genetic diversity. Recently, this technology was successful in sequencing the SNPs of various species^7,8.

In this study, we collected 43 individual I. difengpi samples from 13 locations that encompass its entire natural habitats. Using genotyping-by-sequencing (GBS) and subsequent bioinformatic analyses, we assessed the population genetic diversity of I. difengpi based on single nucleotide polymorphisms (SNPs). The findings contribute to our understanding of the natural history and conservation regarding this rare, endemic herb.

Materials and methods

Tissue sampling and quality control

In this study, we collected individual I. difengpi samples during June to October, 2023. Most of sampled individuals grew on their natural habitats (Fig. 2A; Table 1), except for those from Daxin (X, cultivated from local wild strain) and Yuanqu (Y, grown in Guangxi Botanical Garden of Medicinal Plants as ex-situ conservation). Samples E and T were both from Tian’e county, but they were collected and sequenced at different seasons. As a result, they were grouped together in the calculation of genetic diversity, though labeled with separate letters on the figures.

Fig. 2.

Sampling and habitat coverage of I. difengpi. (A) Collecting sites of this research. See Table 1 for interpretation of the abbreviations. (B) Natural habitats of I. difengpi (recreated from Wang et al., 2021¹).

Table 1.

General information on I. difengpi sampling.

Abbr.	No.	Loc.	Coord.	EL. (m)
A	4	Du’an	108°10′35.7960″~36.4188″, 23°57′59.9220″~58′0.6672″	458.5 ~ 470.0
E, Ta	2	Tian’e	107°9′27.6408″~31.86″, 24°59′32.8920″~25°1′1.62″	812.1
F	2	Fengshan	107°2′46.4820″~46.9140″, 24°46′12.7884″~13.8072″	1,069.7 ~ 1,108.9
G	2	Gesheng	107°38′56.3280″~56.5044″, 24°8′13.4844″~13.5636″	744.2 ~ 754.7
H	5	Huaiyuan	108°28′52.4064″~52.554″, 24°35′17.9736″~18.1464″	126.8 ~ 142.8
J	5	Jingxi	106°18′52.5620″, 23°2′45.2849″	750
L	2	Longzhou	106°35′46.5252″~47.4324″, 22°22′51.0888″~54.3288″	560
M	4	Mashan	108°25′39.2268″, 24°0′9.1512″	99.5
N	5	Napo	105°51′48.4048″, 22°56′56.7853″	882
P	2	Pairu	107°24′24.7324″, 22°34′36.5652″	476
Q	2	Qibainong	107°44′39.7284″, 24°7′16.7376″	903.8
X	6	Daxin	107°12′9.6570″, 22°47′14.5413″	~ 0
Y	2	Nanning	108°22′26.22″, 22°51′26.46″	79

^a E and T were from the same county, but collected at different seasons.

All samples were identified as I. difengpi by Baoyou Huang (the corresponding author), and the voucher specimen was deposited in the Guangxi Chinese Medicinal Materials Herbarium (affiliated with Guangxi Botanical Garden of Medicinal Plants, the deposition number is 451026140604002LY). Permission for collection of I. difengpi tissue samples was granted from the local district authorities in Guangxi (Ref: 2021AC19020), and this study is complied with relevant institutional, national, and international guidelines and legislation. The sampling was non-destructive, with only one or two leaves or buds taken, and this did not harm the plant. As soon as the tissues left the plant, they were stored in 2 mL Eppendorf tubes, frozen in dry ice and then transferred to a -80℃ refrigerator in the laboratory.

The DNAsecure Plant Kit was used to extract genomic DNA from each sample by following the manufacturer’s instruction. The DNA content was evaluated by 1% w/v agarose gel electrophoresis and spectrophotometric analysis (NanoDrop, Thermo Fisher Scientific). 43 samples from 13 locations were qualified for sequencing (Table 1).

Library construction and sequencing

Due to the absence of a reference genome, GBS⁶ was the ideal method for this study. Tissue samples were shipped to Shanghai OE Biotech Co Ltd. (Shanghai, China) for DNA sequencing and bioinformatic analysis. In brief, genomic DNA was digested with restriction enzymes MspI and PstI-HF, and then the resulting cut sites were ligated with barcoded PstI-HF adapter and common MspI adapter, respectively, by T4 DNA ligase (all from New England BioLabs, NEB). After ligation, fragments below 300 bp were recovered with Sera-Mag SpeedBeads (GE Healthcare Life Sciences), and the beads were separated from the supernatants by using a magnetic stand. After elution from the beads, the DNA was amplified through PCR with a forward primer specific to the barcoded adaptor and a reverse primer with homology to the common adapter. The PCR products were examined by using agarose gel electrophoresis to select GBS libraries with concentrations higher than 5.0 ng/µL. Then, the GBS libraries (100 ng) were sequenced on an Illumina Nova platform (PE 150).

Read processing and SNP discovery

The raw reads were examined for quality control and then filtered based on the barcode and PstI restriction site. The resulting clean reads were clustered by the ‘ustack’ module in Stacks program (v1.34)⁹ and processed with ASustacks (GBS)⁶ to filter out rare (less than 50%) and repetitive (similarity level equal or higher than 98%) representative tags. The loci assembling was performed with programs, such as cstacks, sstacks, tsv2bam, gstacks et al. GATK (v3.8-1)¹⁰ identified out all the variants, including SNPs and Indels. Variant filtering was performed by vcftools (v0.1.13)¹¹ with the following criteria to removed loci or SNPs: (1) the sequencing depth was less than 4. (2) the minor allele frequency (MAF) was less than 0.01. (3) values were missing in more than 20% of the samples.

Population genetics analysis

When assembling genotypes with markers, the missing sites were labeled as “–.” The evolutionary relationship was constructed by neighbor-joining method in treebest (version 1.9.2)¹². Admixture (version 1.3.0)¹³ was used to determine the optimal value of population number (K) and population structure. Principal component analysis (PCA) with respect to the SNPs was performed by plink2 (version 2.0)¹⁴. The maps and PCA figures were generated with R¹⁵.

N_e estimation

The above sequencing operations and bioinformatic analysis were carried out by Shanghai OE Biotech Co Ltd. However, further procedures were needed to determine the N_e, and therefore the SNP sequences were sent to Genepioneer Biotechnologies Co. Ltd. (Nanjing, China) for N_e estimation. Two approaches were used for this purpose: currentNe¹⁶ and GONE¹⁷.

Because there were too many loci in the sequencing data, Genepioneer also performed a further SNP filtering using vcftools for the whole I. difengpi population and the subpopulations of A, H, J, and N. The filtering was based on the following parameters: maf = 0.05, min-mean DP = 90, max-missing = 0.7. These were to remove SNPs rarer than 5% or with a missing rate greater than 70%, and to keep SNPs that had a coverage rate of more than 90% on that locus.

Both programs for N_e estimation were run according to instructions from the articles introducing them^16,17. GONE used 1.00 cM per Mb for genetic distance (default setting), since the genome mapping of I. difengpi is not available. For the same reason, currentNe assigned 1.00 Morgan as the genome size for the chromosome information, and this is also the default setting of this program.

Results and discussion

Sample collection

The location of 43 samples in this research (Table 1; Fig. 2A) completely overlapped with the natural distribution (Fig. 2B) of I. difengpi.

Generally, a large sample size would enhance the confidence in the results. However, it is extremely difficult to obtain a large sample size for I. difengpi, because it preferentially grows on cliffs or mountaintops. These locations are hard to reach, especially with tissue preservative materials, such as dry ice or liquid nitrogen. In addition, the habitats of I. difengpi were fragmented, and there were usually less than 10 individuals on each location. Moreover, not all samples returned a satisfactory amount of DNA, presumably due to the richness of polysaccharides and polyphenols in the cytoplasm of I. difengpi. The chemicals interfered with the DNA extraction, and most of our DNA samples ended up as C category or lower. Therefore, these 43 samples, obtained through multiple rounds of extraction, represent the best we could achieve given the current resources and conditions.

Although the sample size may not be abundant for each location, many studies have shown that a large sample size is unnecessary to study the genetic diversity with SNPs. For example, a study on simulated data pointed out that sample size can be as small as n = four to six when accurately estimating F_st¹⁸, while an empirical study on a nonmodel plant without a reference genome reported that the sample size could be as small as 2 individuals¹⁹, given that there were enough SNPs (> 1500). Another study based on microsatellite data concluded that, for estimating heterozygosity, 25 to 30 individuals were enough²⁰, and the loci involved in this study were less than 10. Therefore, 43 samples as a whole population were enough for measuring genetic diversity. For each subpopulation, although the sample size was small, the data were also calculated from thousands of loci and SNPs. Thus, we reported the results for both the whole population and the subpopulations.

DNA sequencing and SNP discovery

For these 43 samples, the sequencing process returned a total of 1,732,609 sequences, with an average sequencing depth of 28.47 folds and a total data size of 24.7 Gb. The percentage of high-quality reads in most samples was greater than 95%, with the lowest being 89.80%. Shanghai OE Biotech Co Ltd. identified 2,030,644 raw SNPs, from which 421,491 were filtered for bioinformatic analysis. From these, 2,921 loci were further screened out by Genepioneer Biotechnologies Co. Ltd. to perform N_e estimation and population history reconstruction.

Evolutionary relationship among the samples

Except for the samples from Fengshan (F) as the most basal group and those from Nanning (Y) as ex-situ conservation, the population of I. difengpi formed 3 clusters (Fig. 3): Northwest (A, E, G, H, M, Q, and T), Central West (P and X), and West (J, L, and N). The branches of all samples were roughly of the same length, indicating that no specific strain was under particularly fast or slow evolution. The bootstrapping values strongly support that samples from the same location form a monophyletic group, and those from adjacent locations were more closely related, pointing to a lack of hybridization or migration between different habitats.

Fig. 3.

Evolutionary relationship of the 43 individuals from 13 locations. (A) Evolutionary tree with bootstrapping values and branch lengths. (B) Evolutionary tree on map showing how strains from various collecting sites were related. Note that F was the most basal one among all strains. Y was left out because it was from ex-situ conservation, and the evolutionary tree revealed that it was most closely related to A.

The basal position of the Fengshan strain could be informative when studying the origin of I. difengpi. Instead of locating at the center of the species’ natural range, this possible origin is on the northern boundary. An early study on platypus (Ornithorhynchus anatinus) reported a similar pattern²¹. Given the limited dispersal ability and special habitat requirements of I. difengpi, it’s likely that a historically large and widespread population experienced decline and fragmentation, resulting in isolated survivors. Consequently, strains that once connected Fengshan and the southwestern habitats may have become extinct.

PCA

The PCA test also confirmed that the Fengshan strain (F) is an outgroup to other strains (Fig. 4A), while the second axis (PC2, 5.2% of variance) separated J, N, L, E/T and other samples, and PC3 (4.72% of variance) further discriminated P and X from the background. Notably, individual data points lined up along the PC2 axis corresponding not only to their locations (Fig. 4B), but also to the evolutionary history (Fig. 3B): the points of West spread out on PC2 since the West cluster is the outgroup to Central West and Northwest, so there are more evolutionary differences to separate the West cluster; the points of Central West and Northwest distribute more tightly because they are newly evolved branches, and their positions on PC2 is consistent with their geographic locations.

Fig. 4.

Principal component analysis (PCA) based on SNP data. (A) 3-D PCA of the first three principal components (PCs). Note that Fengshan was isolated far away. (B) The positions of data points on the graph of PC1 vs. PC2 matched their geographical locations.

Population structure

In Admixture analysis, the cross-validation (CV) errors consistently increased with K (possible number of subpopulations, Fig. 5A). Here, the smallest CV error was found at K = 1, indicating that it was optimal to treat all the samples as one subpopulation. Although this agrees with the low genetic diversity of I. difengpi (see below), other K values were needed to further visualize the population structure. Therefore, stacked bar plots of K = 2, 3, and 4 are shown (Fig. 5B) because their CV errors were next in line. The result from all three possible values of K was that strains (J and N) on the west edge form a distinct “pure west” cluster, while L and P contain genetic elements from the flanking West (J and N) and Central West (X) clusters. As for the Northwest branch, the strains T/E showed intermediate characteristics between the Northwest (pink in Fig. 5B) and the Central West (X), which also agreed with their phylogenetic and geographic positions.

Fig. 5.

Admixture analysis based on SNP data. (A) Cross-validation errors of the potential number of subpopulations (k). (B) Stacked bar plots of K = 2, 3, 4.

Genetic diversity

Some critical parameters regarding the within-(sub)population genetic diversity of these 43 I. difengpi samples are summarized in Table 2. The p-values of Hardy-Weinberg proportions test (HW-p) of all strains were greater than 0.05, which means that the whole population and all the subpopulations had reached the Hardy-Weinberg Equilibrium (i.e., allele and genotype frequencies remain constant over generations). The expected and observed heterozygosity (H_E and H_O) were both extremely low, pointing to a lack of genetic diversity. For the whole population, H_O was slightly smaller than H_E, suggesting that there was a certain degree of inbreeding in the population. These findings were in agreement with the low polymorphism information content (PIC) and nucleotide diversity (π) values.

Table 2.

Within-(sub)population genetic diversity parameters of I. difengpi.

Loc.	HW-pa	HEb	HOc	PICd	πe	N A f	N E g
A	0.973	0.0634	0.0637	0.0515	0.0751	1.1831	1.104
TE	0.9837	0.0536	0.0615	0.0422	0.0741	1.1282	1.0945
F	0.9998	0.0476	0.0929	0.0359	0.0672	1.0974	1.094
G	0.9999	0.0328	0.0637	0.0247	0.0491	1.0672	1.0645
H	0.9559	0.0734	0.0671	0.0606	0.0825	1.2418	1.1161
J	0.9581	0.0704	0.0732	0.0572	0.0794	1.2118	1.1163
L	0.9833	0.0624	0.0726	0.0494	0.0858	1.152	1.1085
M	0.9717	0.0589	0.0657	0.047	0.0687	1.1557	1.1007
N	0.9534	0.0748	0.0676	0.0615	0.0843	1.2404	1.1197
P	0.9954	0.0496	0.0742	0.0389	0.0697	1.1172	1.0882
Q	0.9999	0.0359	0.0703	0.0271	0.0522	1.0733	1.071
X	0.9607	0.0722	0.0752	0.0589	0.0795	1.2263	1.1177
Y	0.985	0.0553	0.0657	0.0436	0.0772	1.1332	1.0972
All	0.6732	0.0991	0.0696	0.0883	0.1004	2	1.1331

^a HW-p, p-value of Hardy-Weinberg proportions test. If HW-p > 0.05, the (sub)population is in Hardy-Weinberg equilibrium.

^b H_E, expected heterozygosity.

^c H_O, observed heterozygosity.

^d PIC, polymorphism information content. The polymorphism is low when PIC < 0.25.

^e π, nucleotide diversity.

^f N_A, observed number of alleles.

^g N_E, effective number of alleles.

The effective number of alleles (N_E, not to be confused with N_e, the effective population size) is the number of alleles on a locus that an idealized population (equal frequency for all alleles) would have for this population to have the same amount of homozygosity as the actual population at that locus. N_E can be calculated by finding the inverse number of the actual population homozygosity. The closeness of N_E to the observed number of alleles (N_A) measures the degree of how evenly distributed those alleles are at that locus, which was detected from all the I. difengpi subpopulations (Table 2).

Conversely, the among-subpopulation genetic variation is fairly high (measured by pairwise F_st value, lower left triangle of Table 3, values above 0.15 means high diversity between groups). We also calculated the Reynolds’ genetic distance (DR) based on DR= -ln(1-F_st) and listed the results in the upper right triangle of Table 3. Note that this is not contradictory to the above findings, because they measure different aspects of population genetics. H_E/H_O, PIC, and π measure the general genetic diversity of the whole population, with no reference to population structure. F_st measures the portion of variation among subpopulations in the total variation. Thus, F_st value can still be high in the background of low genetic diversity, and this could be the result of geographical barriers isolating these subpopulations.

Table 3.

Pairwise F_st and DR values of I. difengpi^a

Subpop.	A	TE	F	G	H	J	L	M	N	P	Q	X	Y
A	–	0.1870	0.5224	0.3781	0.1081	0.2985	0.1618	0.2305	0.2276	0.2641	0.3467	0.2454	0.0876
TE	0.1705	–	0.6028	0.4978	0.1461	0.2846	0.1021	0.3005	0.1956	0.2724	0.5168	0.2506	0.1223
F	0.4069	0.4527	–	1.0675	0.4384	0.5208	0.5245	0.6223	0.4513	0.7531	1.0335	0.5133	0.59
G	0.3149	0.3922	0.6561	–	0.3002	0.4556	0.4186	0.5046	0.3593	0.6873	1.0273	0.4009	0.4522
H	0.1025	0.1359	0.3549	0.2593	–	0.2642	0.1286	0.1891	0.2013	0.2091	0.2797	0.2133	0.0701
J	0.2581	0.2477	0.4059	0.3659	0.2322	–	0.21	0.3694	0.1989	0.3256	0.4673	0.3018	0.2737
L	0.1494	0.0971	0.4081	0.342	0.1207	0.1894	–	0.2675	0.1232	0.193	0.4492	0.1892	0.0896
M	0.2059	0.2596	0.4633	0.3962	0.1723	0.3089	0.2347	–	0.2976	0.3746	0.4778	0.3121	0.2219
N	0.2035	0.1776	0.3632	0.3018	0.1824	0.1804	0.1159	0.2574	–	0.2347	0.3734	0.2386	0.1864
P	0.2321	0.2385	0.5291	0.4971	0.1887	0.2779	0.1756	0.3124	0.2092	–	0.6923	0.2685	0.245
Q	0.293	0.4036	0.6442	0.642	0.244	0.3733	0.3619	0.3799	0.3116	0.4996	–	0.4041	0.4405
X	0.2176	0.2217	0.4015	0.3303	0.1921	0.2605	0.1724	0.2681	0.2122	0.2355	0.3324	–	0.2243
Y	0.0838	0.1151	0.4457	0.3638	0.0677	0.2395	0.0857	0.199	0.17	0.2173	0.3563	0.2009	–

^a F_st values are listed in the lower left while DRs are in the upper right; the p-values of all F_st comparisons are close to 0.

N_e estimation

N_e is an indispensable parameter in population genetics because it measures the degree of genetic drift and inbreeding. In this study, N_e can be interpreted as that its inverse number (1/N_e) is the probability of two randomly chosen alleles coalescing in the previous generation. Here, both softwares infered N_e from LD and their assumptions or limitations are listed in Table 4. The software GONE gave an estimate of 90.09, and currentNe returned a result of 80.80 (Table 4). In addition, the historical N_e (previous five generations and beyond) provided by GONE is four orders of magnitude larger, and this overestimate was most likely due to migration between subpopulations²². The data are not shown because they are obviously artifacts. However, considering the natural history of I. difengpi from previous works¹ and its genetic diversity from this study, the trend of N_e declining is very likely to be real.

Table 4.

The N_e estimates of I. difengpi^{a, b}.

Method	Ne point estimate	Confidential interval	Remarks
GONE	90.0974	N/A	It assumes no migration between subpopulations and persistently closed or isolated population23.
currentNe	80.80	77.87–83.84 (50%); 73.84–88.41 (90%)	It assumes the SNPs mapping is known22.

^a Number of individuals: 43; Number of SNPs (loci): 2921;

^b Number of SNP pairs (independent comparisons): 4,264,660.

We also requested Genepioneer to calculate the N_e in GONE for four subpopulations of A, H, J, and N, for which we had relatively more samples at each location. Three subpopulations showed a smaller N_e as expected, and only N returned an abnormally high N_e presumably due to some rare alleles or artifacts (Table 5). Because GONE tends to overestimate N_e²², the true values could be even smaller. In addition, like the whole population, all four subpopulations experienced recent N_e drop from 5 generations ago (data not shown for the same reason as above).

Table 5.

N_e estimates in GONE for the four selected subpopulations.

Subpopulation	Further filtered SNPs	Ne point estimate
A	1122	54.9138
H	1355	43.8087
J	1239	22.295
N	1338	524.891

There is a concept of 50/500 rule in ecology and population genetics^24,25, stating that 50 and 500 are the N_e thresholds for short- and long-term persistence, respectively. As a comparison, N_e estimates of some common plants, such as European aspen (Populus tremula) and sunflower (Helianthus annuus), range from 70k to more than 800k²⁶. In contrast, the ironwood tree (Ostrya rehderiana) is critically endangered with only five surviving wild individuals. A recent study showed that its N_e has declined for one million years, and it accumulated more deleterious mutations than its congener (O. chinensis) with a stable N_e. During the long history of population declining, O. rehderiana gradually adapted to oppose the effect of inbreeding depression by purging lethal recessive mutations, but functional extinction of this species seems inevitable²⁷. Similarly, a low value of N_e sounds the alarm for immediate attention to the current conservation status of I. difengpi, because its genetic diversity could be too low to respond to demographic or environmental stochasticity.

Conservation strategy

The threat to I. difengpi population mainly comes from human disturbance and habitat lost¹. Therefore, except for protection from law enforcement, population genetics should be considered. This study revealed that I. difengpi experience the impact of genetic drift common among endangered species, characterized by low genetic diversity (e.g., low heterozygosity, PIC, π, and N_e) and highly differentiated subpopulations (high F_st). In this case, in situ conservation alone may not be sufficient, because inbreeding and mutational meltdown can draw a species with a small N_e into the extinction vortex^3,25. Meanwhile, outcrossing could contribute in bringing up I. difengpi population, on the condition that hybrids can have higher fitness and adaptability. Recently, outcrossing has been tested on an endangered shrub, feather-leaved banksia (Banksia brownii). In this study, moderate outcrossing (within-source) was found to be beneficial in terms of significantly higher plant volume compared to inbred offspring. However, outcrossing to a higher extent (between-source) did not show a significant difference, and this could be because of either insufficient instances or a certain amount of outbreeding depression²⁸. As such, outcrossing of I. difengpi should be guided with a pilot trial and more knowledge of hybrid weakness for this and other closely related species. Nowadays, transplanting I. difengpi from West (Jingxi, J) to Northeast Guangxi (Guilin) returned positive results¹, but to our knowledge, hybridization experiments of this species has yet to be made.

Conclusions

In this study we used GBS technology to study the genetic diversity of I. difengpi, an endangered medicinal plant native to the karst areas in Guangxi. The results show that I. difengpi suffers from low genetic diversity, which is exemplified by the low heterozygosity, PIC, π, and N_e. In particular, its N_e was estimated to range between 80 and 90, which suggested that it could fall into the extinction vortex in the near future. In addition, phylogenetic studies, PCA, and Admixture analysis indicated that the genetic elements of I. difengpi match the geographic distribution of its subpopulations. These subpopulations were found to be isolated and highly differentiated, so our findings accordingly recommend that future research could attempt to perform outcrossing. If outcrossing can bring hybrid vigor, it will add to the current effort for the conservation of this species.

Author contributions

BQ directed the research and wrote the paper; BYH, BQ, YH, YL, and XYG organized and prepared for the collection; BYH took the photos; BYH and BQ collected the samples; BQ, BYH, and YFH preserved and stored the samples; BQ, YH and YL revised the draft. All authors reached the final agreement for publication.

Funding

This work was supported by the Guangxi Science and Technology Program [grant number Guike AD21238011, from the Science and Technology Department of Guangxi Zhuang Autonomous Region, China], and Guangxi elite team of medicinal plant conservation.

Data availability

Sequence data of this study have been deposited in the National Center for Biotechnology Information with the accession code PRJNA1192014.

Declarations

Competing interests

The authors declare no competing interests.