Natural hybridization is known to play a vital role in speciation; however, the mechanisms underlying the early stages of natural hybridization remain unclear. Where two plant species come into contact, two driving forces may balance the dynamic consequences of hybridization: fusion by hybridization-mediated gene flow, and separation by reproductive isolation (RI) (Ma et al., 2010a, b; Chang et al., 2022). Therefore, the early stages of hybridization may lead to several possible outcomes. One common outcome of hybridization is the formation of a hybrid swarm, including hybrid genotypes produced from gene flow between the two parents (e.g., backcrosses to one or both parents) with or without intermediate genotypes (e.g., F₁s, F₂s). Another outcome is the persistence of only intermediate genotypes, i.e., only F₁s, or F₁s plus F₂s. In this second outcome, later intermediate genotypes (e.g., F₃s, F₄s …) may also occur, although their occurrence is challenging to identify given the limitations of current analytical methodologies.

Rhododendron L. (Ericaceae) is the largest genus of seed plants in China and has important cultural, horticultural, and ecological value. Rhododendron has undergone complex evolutionary radiation (Shrestha et al., 2018; Xia et al., 2022; Shen et al., 2024). Frequent natural hybridization is known among species in the subgenera Hymenanthes (Zhang et al., 2007, 2020; Ma et al., 2010a, Ma et al., 2010b) and Pseudorhodorastrum (Zheng et al., 2025). Previous studies have reported that most Rhododendron species are self compatible, and geitogarmy pollination might contribute to their reproductive success (Ma et al., 2015). However, these studies have relied on a limited number of markers; in addition, none of these studies employed whole genome resequencing data, which covers nearly all nucleotide variations in the genome (Ma et al., 2022). Simulations have clearly shown that species with lower divergence require more markers to accurately estimate admixture, due to shared polymorphisms that decrease the number available diagnostic markers and thus lower the precision of estimation of admixture proportions (McFarlane and Pemberton, 2019).

One of the best-known subgenera within the genus Rhododendron is Tsutsusi, owing to the tens of thousands of horticulturally significant varieties, which are prized for their vibrant flowers and adaptability. Notably, there are 58 species (figures include both varieties and forms) of Tsutsusi in China, accounting for 72.5% of the world's Tsutsusi species (Geng, 2014; Dai et al., 2020). To date, however, no case of natural hybridization has been reported involving any species of subgenus Tsutsusi in China. Rhododendron strigosum is a newly described species endemic to Jinggangshan, Jiangxi Province (Liu, 2001). Due to the variable number of stamens (6–10), rose-colored flowers, and the distribution, which is sympatric with R. simsii (red flowers with 10 stamens) and R. hypoblematosum (pale purple flowers with 5 stamens), we hypothesized that R. strigosum might have hybrid origin and could result from natural hybridization between R. simsii and R. hypoblematosum (Fig. 1A–C).

In the present study, we sampled 80 individuals from the type location of Rhododendron strigosum (including 35 individuals of R. simsii, 25 of R. hypoblematosum and 20 of R. strigosum) in Jianggangshan, Jiangxi Province (JGS), and 14 R. strigosum individuals from Lushan Botanical Garden, Jiangxi Province (LSBG), where R. strigosum were cultivated under ex-situ conservation. To explore the genotypes of seedlings obtained from seeds following natural pollinations in the ex-situ conserved R. strigosum individuals in LSBG, a further 11 seedlings were sampled. Based on whole-genome resequencing data analysis for the total of 105 samples, we aimed to: (ⅰ) test the hypothesis that R. strigosum is of hybrid origin; (ⅱ) determine the genetic make-up of R. strigosum with respect to classes (i.e. F₁, backcross or complex hybrid derivatives) and (ⅲ) integrate these results to reasonably explain the possible causes and consequences of this first case of natural hybridization in Subgenus Tsutsusi in China.

In total, ~2.62 TB of WGS data were generated for the 105 accessions from Rhododendron simsii, R. hypoblematosum and R. strigosum, with an average depth of ca. 45.15 (Table S1). Using the chromosome-level R. simsii reference genome, the mapping rate of these raw reads was found to be 94.13% (Table S1). The putative hybrid R. strigosum had the highest heterozygosity rate (~0.0122) when compared with its two parental species R. simsii (~0.0067) and R. hypoblematosum (~0.0084). R. strigosum also had the highest values of the two nucleotide diversity parameters π and θw (π: 0.0279; θw: 0.0297) compared with R. simsii (π: 0.0200; θw: 0.0265) and R. hypoblematosum (π: 0.0234; θw: 0.0223). Pairwise F_ST values among the three species were highest between R. simsii and R. hypoblematosum (0.2450), and lowest between R. simsii and R. strigosum (0.0981).

The PCA plot based on 2, 320, 655 SNPs showed that the first principal component accounted for 49.79% of total variation, and separated Rhododendron simsii from R. hypoblematosum. The putative hybrid R. strigosum samples were intermediate between the two parental species (Fig. 1D). The second principal component (21.01% of total variation) further divided R. strigosum samples into two categories, implying distinctive genetic composition within R. strigosum.

ADMIXTURE analysis based on 2, 320, 655 SNPs showed that all samples morphologically identified as Rhododendron hypoblematosum had extremely high genetic support (q extremelyK = 2). Similarly, all morphologically identified R. simsii samples (except YJP14) also had strong genetic support (q strong geK = 2). For the putative hybrid R. strigosum samples collected in JGS, all individuals exhibited genetic admixture between R. simsii and R. hypoblematosum (Fig. 1E). Intriguingly, with the exception of FJS02, the genomic composition of all the remaining R. strigosum individuals from JGS exhibited intermediate genotypes (0.521 ≥ q ≥ 0.467). Of the 25 R. strigosum samples collected from LSBG, all 14 ex-situ conserved samples exhibited the same genetic pattern as those from JGS. Of the remaining 11 plants obtained from seeds resulting from natural pollinations in the ex-situ conserved R. strigosum individuals, all had a distinct genetic composition, with a higher proportion of genetic material from R. simsii than from R. hypoblematosum. At K = 3, the 20 R. strigosum samples collected in nature (JGS population) maintained the same genetic composition as K = 2, whereas the ex-situ conserved R. strigosum samples collected from LSBG exhibited a unique genetic component. Among these 25 samples, 13 of the 14 ex-situ conserved R. strigosum samples showed this unique genetic pattern, whereas most seedlings propagated from seeds of these ex-situ conserved R. strigosum showed admixture between these ex-situ conserved R. strigosum and R. simsii.

We obtained 75, 744 ancestry-informative SNPs fixed (F_ST = 1) between Rhododendron simsii and R. hypoblematosum. Due to the constrained loci number of the software, NewHybrids analysis based on 337 ancestry-informative SNPs showed that all morphologically and admixture-identified R. hypoblematosum and R. simsii ‌individuals (except YJP14, which was identified as an F1) were assigned to the corresponding parents with very high posterior probabilities (Fig. 1F). With the exception of the individual FJS02, which was identified as a backcross hybrid to R. hypoblematosum, all the remaining collected R. strigosum samples from both JGS and those conserved ex-situ in LSBG were identified as ‌F1 hybrids. Of the 11 progeny propagated from R. strigosum seeds, 10 individuals were identified as backcross hybrids to R. simsii, and one individual was identified as an F2 (Fig. 1F).

Rhododendron hypoblematosum was found to have the highest number of deleterious mutations, whereas the F₁s and the first backcross generation (BC1) ranked second and third, respectively. R. simsii had the lowest number of deleterious mutations. We further divided these mutations into two states, i.e., homozygous and heterozygous deleterious mutations. R. hypoblematosum also had the highest ratio of homozygous deleterious mutations, followed by R. simsii, BC1 and F₁s. As expected, the ratio of heterozygous deleterious mutations had ‌the opposite trend, i.e., F₁s > BC1 > R. simsii > R. hypoblematosum (Fig. 1G–I).

Here, we report the first study case of natural hybridization in subgenus Tsutsusi in China, identified using whole genome sequencing analysis. We confirm the hybrid origin of Rhododendron strigosum and reflect its species status because the wild individuals of this taxon are all F1s, with the exception of individual FJS02, which was identified as backcross to R. hypoblematosum. The pattern of hybridization we found in our results is congruent with that observed in other cases of natural hybridization in Rhododendron. For example, R. agustum (in subgenus Hymenanthes) was reported to be mainly composed of F1s in its type location (Zha et al., 2010), but this composition could vary in other locations, for instance, in Baili Nature Reserve (Zhang et al., 2017). Nearly all wild R. strigosum individuals (34/35) were F1s, indicating very constrained gene flow and, in nature, nearly complete reproductive isolation between R. hypoblematosum and R. simsii. Notably, for the ex-situ conserved R. strigosum samples, there are still some genetic components not explained by the two parents and it is not clear whether these distinct genetic components have an ancient origin before the divergence of these two parents (Fig. S1). Thus, it is probable that an unknown parental lineage joined the formation of the ex-situ conserved R. strigosum samples in JGS.

An interesting finding is that F1 abortion (e.g., the abnormal meiotic behavior in hybrids arising from natural hybridization, between Ligularia paradoxa and L. duciformis, Pan et al., 2008) was not the main reason for selection against other types of hybrids. We were able to propagate seedlings from seeds arising from the natural pollination of Rhododendron strigosum. Furthermore, our study provides direct evidence for the low probability of F2 hybrid formation: for the 11 seedlings examined, 10 were BC1 and only a single individual was identified as an F2. The formation of F2s is believed to be rare due to the fact that once F1s have formed, they are generally surrounded by a large proportion of parental plants, and there is therefore a higher chance of backcrossing between the F1s and their parents than of self-pollination within the F1s (Arnold, 1997). Moreover, even if F2 seed was produced, lethal effects (as described in the Dobzhansky-Muller model of hybrid incompatibility: Dobzhansky, 1936; Muller, 1942) would remove some of the recombinant genotypes, while this is likely to be less of an issue for backcrossed genotypes.

We have also evaluated mutation loads among parental species and hybrids, which is currently still rare in natural hybridization studies despite the fact that whole genome sequence data has become highly accessible. We tentatively conclude that Rhododendron hypoblematosum had both the highest number of homozygous deleterious mutations and the highest ratio of homozygous deleterious mutations to synonymous mutations, exceeding that of the F1s. Theoretically, if a large number of parental individuals contributed to the formation of the F1 generation, the deleterious mutation count of the F1 should exceed that of either parent (Sun et al., 2023). The observed intermediate level could be explained by the involvement of a limited number of parental individuals in the formation of R. strigosum.

In summary, using whole genome sequencing data, we report the first study case of natural hybridization in subgenus Tsutsusi in China. Nearly complete reproductive isolation was detected between the two parental species as > 97% of the examined Rhododendron strigosum plants were F1s, and thus the species status of R. strigosum should be rejected. We also revealed the rare formation of F2s after detecting seedlings from seeds of R. strigosum under natural pollination. Although the number of deleterious mutations in F1s can be high, these can be masked in heterozygotic state. However, those effects particularly relating to fitness could be exposed to natural selection once backcrossed and/or once F2 hybrids formed, as large amount of these masked deleterious mutation would become homozygous, which may also provide valuable insight into genetic rescue programs for conservation biology.

Acknowledgments

This study was supported by the National Natural Science Foundation of China (U23A20160, 32360336) and Guizhou Provincial Key Technology R&D Program (Qian KeHe ZhiCheng [2023] YiBan035).

CRediT authorship contribution statement

Xiaoling Tian: Writing-original draft, Visualization, Software, Formal analysis, Data curation. Ningning Zhang: Writing-original draft, Formal analysis. Xiaohua Li: Investigation. Zhong Wang: Investigation. Heng Shu: Investigation. Chunying Zhang: Investigation. Yongpeng Ma: Conceptualization, Writing – review & editing. Yupeng Geng: Writing-review & editing, Supervision.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

Supplementary data to this article can be found online at https://doi.org/10.1016/j.pld.2025.12.001.