We collected 16 populations of C. sinuosa in the northern coast of South China Sea (the latitude range spanning 18–22° N), including 5 populations in Guangxi province, 1 population in Guangdong province, and 10 populations in Hainan Island (Table 1, Fig. 1, Supplementary Fig. S1). The total genomic DNA was isolated using Hi-DNAsecure Plant Kit (TIANGEN, DP350-03) following the manufacturer’s instruction. After quality checking by OD₂₆₀/OD₂₈₀ ratio and 1% agarose gel electrophoresis, multiple molecular markers were amplified using these DNA template. The mt cox3, atp6, and cp rbcL partial sequences were amplified with pair primers F49 (5′-CATTTAGTNGAYCCWAGYCCTTGGC-3′)/R20(5′-AACAAARTGCCAATACCAKG-3′), F25P (5′-CCHTTAGAACAATTTBAAATACTYCC-3′)/R754P (5′-GCRTCRTTTATRTARATRCAACTTA-3′)(Leeetal. 2014) and F3 (5′-GCCTGAAGATGTGCAGAATCG-3′)/R1266 (5′-GCCTGAAGATGTGCAGAATCG-3′). The amplification procedures were carried out based on the method in Lee et al. (2013) and Song et al. (2019) as follows: for mt cox3 gene, the polymerase chain reaction (PCR) reactions consisted of an initial denaturation at 94°C for 4 min, followed by 35 amplification cycles containing denaturing at 94°C for 30 s, annealing at 53°C for 45 s, extending at 72°C for 1 min, and a final extension at 72°C for 10 min; for mt atp6 gene, the PCR reactions consisted of an initial denaturation at 94°C for 4 min, followed by 35 amplification cycles containing denaturing at 94°C for 30 s, annealing at 45°C for 30 s, extending at 72°C for 1 min, and a final extension at 72°C for 10 min; for cp rbcL gene, the PCR reactions consisted of an initial denaturation at 94°C for 4 min, followed by 45 amplification cycles containing denaturing at 94°C for 1 min, annealing at 48.5°C for 45 s, extending at 72°C for 1 min, and a final extension at 72°C for 10 min. Purification and sequencing of PCR products were performed using a BigDye Terminator Cycle sequencing kit and an ABI3730 automated sequencer (Applied Biosystems, Foster City, CA, USA).

The obtained sequence datasets were aligned with MUSCLE model in MEGA X, respectively (Kumar et al. 2018). In addition to the datasets for individual genes, we constructed a multi-gene dataset as concatenated cox3-atp6-rbcL. After manual adjusting for false or missing gaps, variable sites and parsimony informative sites in each marker were counted in DnaSP v5.0 (Librado and Rozas 2009). Molecular genetic diversity indices of 16 populations were inferred in Arlequin v3.5.2.2 (Excoffier and Lischer 2010), including haplotype diversity (h) and nucleotide diversity (π).

We reconstructed phylogeographical network of C. sinuosa based on the haplotypes of concatenated cox3-atp6-rbcL, cox3, atp6, and rbcL using Median-Joining algorithm in POPART (Leigh and Bryant 2015). Besides, we assessed the population genetic structure based on the 4 datasets using admixture model in Structure v2.3.4 (Pritchard et al. 2000). The K value of putative clusters was set from 1 to 8. To ensure the reliability and consistency, the analysis carried out 10 repetitive iterations per K with a burn-in of 100,000 Markov chain Monte Carlo (MCMC) chains followed by 500,000 chains using a correlated allele frequencies model (Falush et al. 2003). The log-likelihood LnP(K) and statistic ΔK (Evanno et al. 2005) was estimated to determine most probable K value by online StructureHarvester(http://taylor0.biology.ucla.edu/struct_harvest/). Based on the K results, the analysis of molecular variance (AMOVA) was conducted to explore the genetic differentiation of subdivisions using all datasets in Arlequin. Additionally, the population pairwise differentiation (F_st) was estimated in Arlequin with 1,000 permutations.

To check the population distinctiveness, the principal component analysis (PCoA) was operated in GenAlex v6.5 (Peakall and Smouse 2012) based on the population pairwise genetic distances with all datasets calculated in DnaSP v5.

The isolation by distance (IBD) model was estimated in order to assess whether the geographical distance acted as an important factor to influence population genetic structure. The genetic distances between pair populations were calculated measured as Slatkin’s linearized pairwise F_st [F_st = F_st / (1 − F_st)] (Rousset 1997). The log-transformed geographic distances between pair sampling sites along the coastline were estimated with the roadmap tool in QGIS (QGIS Development Team 2024). Also, the isolation by environment (IBE) model was estimated to analyze the correlation about environment variable and population genetics. The total 19 bioclimate variables were obtained in WorldClim 2 (https://worldclim.org/) (Fick and Hijmans 2017). Pearson’s correlation coefficients between bioclimate variables were calculated by “cor” function in R 3.4.1, and the highly correlated variables (|r| > 0.7) were removed, then 5 bioclimate variables (bio2, bio8, bio14, bio32, and bio38) were kept for IBE analysis. The environmental distance matrix was calculated based on pairwise Euclidean distances using ‘vegan’ package (Dixon 2003) in R. Partial mantel test using Pearson’s product-moment correlation were carried out to test the relation between genetic distances, geographical distances, and environmental distances with 10,000 permutations using ‘vegan’ package in R. The multiple matrix regression with randomization (MMRR) was implemented to evaluate the IBD and IBE model using ‘MMRR’ function in R.

Due to the lack of direct fossil evidence, we calibrated the molecular clock based on the timeframe proposed by Silberfeld et al. (2010), Choi et al. (2024), and Denoeud et al. (2024). For reckoning the divergence time of Colpomenia, we retrieved 63 sequences of 21 brown algae species in GenBank (Supplementary Table S1) and 3 sequences of an individual (HT1) in this study, and constructed a cox3-atp6-rbcL candidate dataset, aligned as 2,216 bp. We estimated the partition for each codon of the dataset and calculated optimal substitution models using Akaike information criterion in PartitionFinder v2.1.1 (Lanfear et al. 2017). Six partitions were specified and the optimal substitution model for all sites was determined as GTR + I. We reconstructed the time-calibrated phylogeny using log normal relaxed clock implemented in Beast v2.7.5 (Bouckaert et al. 2019). Yule model was selected for the tree prior, and 3 fossil nodes were added to help the divergence time correction: (1) Padina boryana and P. pavonica were set as monophyletic group with an age of Padina genus (99.6–145.5 Ma); (2) the age of Sargassaceae lineage (including Sargassum horneri, S. muticum and S. thunbergii) was taken for 13–17 Ma; (3) a monophyletic group including Nereocystis luetkeana and Pelagophycus porra with the age for 13–17 Ma. The runs of MCMC carried out 10⁸ generations with sampling each 1,000 generations, then 10⁵ samples were saved, and the first 25,000 samples were removed as burn-in. One consensus tree was constructed with maximum clade credibility in TreeAnnotator v2.7.5 (Drummond et al. 2012). The tree was visualized using FigTree v1.4.3 (http://tree.bio.ed.ac.uk/software/figtree/).

To examine the extent of evolutionary selection among populations, we carried out the neutrality test (Tajima’s D and Fu’s F_s) using cox3-atp6-rbcL, cox3, atp6, and rbcL datasets in DnaSP. Besides, we calculated the Tajima’s D and Fu’s F_s based on 5 lineage datasets of concatenated cox3-atp6-rbcL, respectively. To evaluates demographic historical dynamic patterns in different genetic subgroups, the coalescent extended Bayesian skyline plots (EBSPs) analysis was preceded with different genetic lineages in Beast v2.7.5 based on cox3, atp6, and rbcL datasets. Molecular clocks of these three genes were chosen as relaxed clock log normal with calculated substitution rates mentioned in the previous paragraph. MCMC chains were set as 8 × 10⁸ iterations, followed by 2 × 10⁸ iterations discarded as burn-in, with sampling every 80,000 iterations.

In the 16 populations of southern Chinese coast, we obtained 233 cox3 sequences (622 bp, 37 haplotypes) (GenBank accession Nos. PQ760306–PQ760342), 294 atp6 sequences (636 bp, 37 haplotypes) (GenBank accession Nos. PQ760184–PQ760220), and 289 rbcL sequences (1,143 bp, 19 haplotypes) (GenBank accession Nos. PQ760221–PQ760239), with 28, 29 and 8 parsimony information sites, respectively. And we established a multi-gene dataset consisting of 219 concatenated cox3-atp6-rbcL sequences (2,401 bp) with 74 haplotypes.

The molecular diversity indices showed that the populations in Guangdong and Hainan exhibited higher diversity (Table 1, Supplementary Table S2), while only 1 population in Guangdong showed higher haplotype diversity in the cp rbcL dataset (Supplementary Table S2). Most populations in Guangxi provinces showed lower haplotype and nucleotide diversity (Table 1, Supplementary Table S2). And 3 populations (Dongfang [DF] and Ledong [LD] in the southwestern Hainan and Wanning [WN] in southeastern Hainan) displayed the highest haplotype diversity by exceeding 0.9 (Table 1).

We found two main genetic clusters (Group I and II) and 5 subdivided genetic lineages in concatenated cox3-atp6-rbcL (Fig. 1), in which that Group II contains 3 populations in Sanya of southern Hainan (Xiaodonghai [DH], Luhuitou [HT], and Haitangwan [HW]), and Group I contains the rest 13 populations (Fig. 1A). We noted obvious regionalism to the lineage distribution (Fig. 1B). More precisely, the northern part of the study area was occupied by the lineage 1, and the eastern coast was mostly dominated by the lineage 2, also, the western part was mainly covered by the lineage 1 and 3, and the southern part was majored in the lineage 4 and 5 (Fig. 1B). Furthermore, two genetic clusters were also detected in cox3, atp6; and rbcL datasets (Supplementary Figs S2B, S3B & S4B). Four genetic subdivisions were recognized in mt cox3 (Supplementary Fig. S2C), and five subdivisions were discovered in mt atp6 (Supplementary Fig. S3C). Compared to the concatenated cox3-atp6-rbcL, the genetic distribution of these 2 datasets also showed similar regionalism (Supplementary Figs S2A & S3A). As for the cp rbcL, there were 2 genetic groups found (Supplementary Fig. S4C), consistent with the result of population genetic structure (Supplementary Fig. S4B), in which that one group was mainly distributed in the southwestern and southern Hainan, and the other group contained other populations.

In parallel, the PCoA analyses were conducted for assessing the population partitions and deviations. Consistent with the result of genetic structure (Fig. 1A), there were 2 main genetic groups found in the cox3-atp6-rbcL dataset (Fig. 2B). Three populations of Sanya in southern Hainan consisted of Group II, and other populations consisted of Group I. However, within the genetic groups, it was evident that the populations exhibited various patterns of aggregation and dispersion. In Group I, nine populations in Guangxi and eastern Hainan displayed high levels of aggregation, whereas the three populations in Guangdong and southwestern Hainan (DF, LD, and Guangdong [XW]) remained closely clustered. Conversely, the WN population in southeastern Hainan showed relatively distinctness from other populations. In Group 2, the HW population stood out as a significant outlier. Similarly, the result of mt cox3 and atp6 dataset showed 2 genetic groups among 16 populations (Supplementary Figs S5 & S6), resembling the result of concatenated dataset. Meanwhile, the two populations WN and HW were disjoint in respective genetic groups (Supplementary Figs S5 & S6). In cp rbcL, although there were 2 main genetic groups, the populations contained in each group were different with the other 3 datasets, and the population DF was a specific outlier in Group II (Supplementary Fig. S7).

We examined the genetic differentiations by AMOVA and F_st. The AMOVA analysis of all datasets suggested that the major genetic differentiations occurred among genetic groups and within populations (Supplementary Table S3). The genetic differentiation between populations was examined with all datasets (Fig. 2A, Supplementary Tables S4 & S5). The pairwise F_st values revealed significant differentiations among most populations (Fig. 2A). The populations in Sanya, southern Hainan (HT and DH) showed highest differentiation with other populations (Fig. 2A, Supplementary Tables S4 & S5). Besides, five populations of XW and southwestern Hainan (DF, LD, HW, HT, and DH) showed significant genetic differentiations in cp rbcL dataset (Supplementary Table S5), coincident with the result of population hierarchical structure (Supplementary Fig. S4).

When the significance threshold was specified as 0.01, the mantel test using IBD model showed that |r| = 0.28 and p = 0.001 (Supplementary Fig. S8), while the mantel test using IBE model showed that |r| = 0.18 and p = 0.018 (Supplementary Fig. S9). Consequently, the current population structure of C. sinuosa can be explained by the IBD model, indicating that geographical distance rather than environmental change was the main factor leading to the isolation among populations.

Based on concatenated cox3-atp6-rbcL, we reconstructed the divergence calibration within 21 Phaeophyceae species, and the result was comparable to the timeframe proposed in Choi et al. (2024) and Denoeud et al. (2024). The origin of C. sinuosa may occur at 26.73 Ma (22.52–31.18 Ma) before present, which was dated at the Chattian stage in late Oligocene (Supplementary Fig. S10). According to this divergence time, we compared the genetic distances among 3 Colpomenia species (C. sinuosa, C. peregrina and C. claytoniae), and the molecular clocks of the 3 markers were estimated as 0–635% Ma⁻¹ in cox3 (average 0.255% Ma⁻¹), 0–0.71% Ma⁻¹ in atp6 (average 0.247% Ma⁻¹), and 0–0.25% Ma⁻¹ in rbcL (average 0.082% Ma⁻¹), respectively.

The neutrality tests were examined with Tajima’s D and Fu’s F_s index at population level based on 3 datasets. The D and F_s were negative in most populations, suggesting potential population expansions in the evolutionary process, however, all D and F_s were positive only in population HW in Sanya, southern Hainan, suggesting underlying contraction (Supplementary Table S6). Meanwhile, the neutrality tests were computed at lineage level in concatenated cox3-atp6-rbcL. All the 4 lineages (except lineage 4 containing only 3 individuals) showed expansion possibility within the 2 tests (Supplementary Table S7).

Furthermore, the EBSP showed that all lineages have undergone dynamic expansions in Pleistocene (Fig. 3). The lineage 1 has experienced expansion at about 0.2 Ma ago and expanded more rapidly from about 0.1 Ma ago. There was slight expansion detected in lineage 2 from 0.35 Ma before present. The lineage 3 exhibited apparent expansion from about 0.4 Ma before present and the lineage 5 had slight expansion within a very short time period at about 0.1 Ma before present.

In this study, most populations displayed the pattern of high haplotype diversity and low nucleotide diversity (Table 1, Supplementary Table S2). According to the thesis proposed by Grant and Bowen (1998), the formation of such diversity pattern suggested that C. sinuosa in northern South China Sea have experienced rapid population growth in a short period because sudden expansion couldn’t support sufficient time to accumulate enough nucleotide mutations. Similarly, the star-like structure of haplotype network implied intraspecific expansion as well (Fig. 1C, Supplementary Figs S2C, S3C & S4C). However, most populations of cp rbcL dataset showed the pattern of low haplotype diversity and low nucleotide diversity (Supplementary Table S2), and compared to mitochondrial datasets, there were not so much subdivided genetic lineages showed (Supplementary Fig. S4). It reinforced that mitochondrial markers of C. sinuosa might have higher resolution in phylogeographic structure delineation than cp gene (Hu et al. 2017).

In this study, the genetic diversity distribution of C. sinuosa exhibited complex picture. Three populations in southwestern and southeastern Hainan (DF, LD, and WN) displayed highest haplotype diversity. Different situation occurs in a coral species Turbinaria peltata that the DF population exhibited lowest genetic diversity due to severe anthropogenic pollution (such as wharf building) (Wu et al. 2021). The contrasting situation suggested that C. sinuosa might have stronger adaptation to resist the negative environmental change compared with other sensitive marine species. The higher diversity pattern in southwestern Hainan appears to be similar with the findings of another brown alga Sargassum polycystum C. Agardh (Hu et al. 2018, Liang et al. 2022), which indicated that the population of LD in southwestern Hainan harbored the richest genetic diversity. In our study, the population of LD also showed high genetic diversity (Table 1, Supplementary Table S2), but the result could be distorted due to the limited sample size, so more extensive sample collection should be conducted for further understanding. Additionally, for S. polycystum, Hu et al. (2018) proposed a hypothesis that a unique historical climate refugium might be located at southwestern Hainan because the population of LD in Hainan had highest genetic diversity and supported the most widespread ancestral haplotypes. Moreover, the Yinggehai Basin in Ledong harbored rich endemic species diversity, and the pockmarks and mudflow gullies in Yinggehai Basin have helped to preserve unique diversity during the sea level dropping in Pleistocene (Hu et al. 2018). In our study, although there was only 1 lineage in the southwestern Hainan in the concatenated dataset, we found that the population of DF in southwestern Hainan held haplotypes of all lineages in genetic Group I in the mt cox3 dataset (Supplementary Fig. S2A). The populations in glacial refugia usually accumulated higher genetic diversity and particular haplotypes than recolonized populations (Maggs et al. 2008). Thus, our results partially supported the hypothesis by Hu et al. (2018). However, it still needs further sample collection and analysis. Moreover, the population of WN in southeastern Hainan had the highest genetic diversity (Table 1). Based on the structure result, this population held genetic compositions of both Group I and II (Fig. 1A, Supplementary Figs S2B, S3B & S4B), and in the PCoA result, the population WN was located in the middle position of 2 genetic groups (Fig. 2B, Supplementary Figs S5 & S6), which suggested that this population might be an intersection of genetic groups. But no haplotypes of lineage 3 has been found in this population (Fig. 1B), so it could be challenging to perceive it as a diversity center. We assumed the situation might result from weak genetic exchange between southwestern and southeastern Hainan.

Deep genetic structure and significant differentiation were detected in C. sinuosa datasets in the northern margin of the South China Sea. Two genetic groups were displayed clearly between the populations of southern Hainan and other regions in Guangxi, Guangdong and Hainan (Fig. 1, Supplementary Figs S2 & S3), which indicated that current genetic distribution might originate from different ancestral populations despite monophyletic origin. According to the IBD / IBE results, the genetic differentiation was influenced by geographical isolation rather than environmental variables (Supplementary Figs S8 & S9). Therefore, the current population genetic structure might be a consequence of historical isolation engendered by paleoclimate change. Historical vicariance caused by Pleistocene glaciation events may account for present phylogeographic pattern. During the glacial cycles in Pleistocene, the marginal seas in East Asia have endured huge sea level fluctuation, and the South China Sea became enclosed inland sea, only connected to the Pacific by Bashi Strait in the Last Glacial Maximum (Ni et al. 2014). Thus, we assumed a scenario that current genetic distribution pattern might be a mixture of separated ancestral populations, which were likely isolated due to declining sea levels during the Pleistocene ice age and subsequently expanded in the process of glacial eustasy (Hoarau et al. 2007, Sinclair et al. 2016, Bringloe et al. 2020). Although both EBSP analysis and neutral tests indicated that all the lineages of C. sinuosa have experienced demographic expansion in Pleistocene, the demographic process occurred in different period (Fig. 3), which implied that the ancestral populations in two genetic groups might experience multiple colonization and spread in turbulent historical oscillation, resulting in the sophistication of genetic differentiation. As the most ancestral lineage, lineage 5 maintained relatively long-term demographical stability with a minor expansion in recent 0.1 Ma. The contrasting phylogeographical process might be responsible for significant differentiation with other lineages, suggesting strong endemism and limited dispersal capability. However, this small lineage is restricted at Sanya in southern Hainan with low genetic diversity (Fig. 1). Additionally, although we detected expansion in lineage 5, the HW population in this lineage showed signs of population contraction in the neutrality test (Supplementary Table S6), suggesting potential bottleneck effect in the evolutionary process. Thus, it revealed a potential precarious situation and underscored the importance of seaweed diversity conservation of Sanya. Contrast to the post glacial expansion pattern of other seaweed species (Zhang et al. 2019b, Song et al. 2021, Assis et al. 2023), the C. sinuosa lineages presented pre-LGM (Last Glacial Maximum) demographic dynamics pattern (Fig. 3). Especially, the population expansion of lineage 1 experienced a significant acceleration during recent 0.1 Ma in late Pleistocene. This phenomenon might be attributed to the relatively warm climate during the last interglacial period (1.3–0.5 Ma before present in China) (Wang and Wang 1980), which might facilitate the dispersal and recolonization of C. sinuosa. This pre-LGM demographic pattern was also observed in other seaweed and marine organisms (Coyer et al. 2011, Ni et al. 2014). We assumed that C. sinuosa existed throughout the last glaciation of Pleistocene, and the LGM seemed to have influence only on its distribution and habitat, rather than demographic history. It implicated the importance of insight on pre-LGM tectonic events and marine population demographic genetic shifts.

At present, there were few researches discussing about the phylogeographical pattern of C. sinuosa (Cho et al. 2009, Lee et al. 2013, Martins et al. 2022). Cho et al. (2009) found that C. sinuosa could be divided into 2 genetic groups in northern and southern hemispheres, while Lee et al. (2013) proposed that there were 3 divergent groups at the global scale and C. sinuosa might originate from the large Indo-Pacific region. The phylogeographic studies of C. sinuosa across global oceans highlighted its complex genetic diversity and distribution were likely shaped by historical climate changes and oceanic barriers. Furthermore, Martins et al. (2022) have illustrated that the genetic structure of C. sinuosa along the Brazilian coast might be influenced by Quaternary climate change, coastal habitat heterogeneity and ocean current barriers. Notably, all three studies have also found that Pleistocene glaciation-induced climate change had a significant impact on the current genetic distribution patterns of C. sinuosa, which were comparable with our findings. Although our study is only based on a regional scale, we found more haplotypes than the Brazilian coast (Supplementary Table S2), and the demographic expansions of genetic lineages in northern South China Sea were roughly earlier than that in Brazilian coast (Martins et al. 2022). It might emphasize more intricate evolutionary history in the Western Pacific. In addition, when comparing current released haplotypes worldwide, we found that all the cox3 haplotypes in this study were newly discovered, but several haplotypes were comparable to those from Taiwan in China, the Philippines (Lee et al. 2013), and French Polynesia; similarly, most rbcL haplotypes were newly recorded, but one haplotype was identical with those from the Philippines (Santiañez et al. 2022) and French Polynesia. Therefore, integrating these newly discovered haplotypes into comprehensive study from the global landscape may offer novel insights into the evolutionary history of C. sinuosa. Given the sampling limitation of regional scale studies, it is crucial to study the phylogeographical pattern of C. sinuosa in further large-scale comparative research with integration of these previous research.

The dynamic fluid nature of the marine environment promotes connectivity among distinct populations, which helps to shape genetic structure, biogeography and speciation (Cowen and Sponaugle 2009, Assis et al. 2022). Multiple researches have indicated that complex hydrological factor of South China Sea could significantly shape the marine phylogeographical patterns by facilitate the genetic homogeneity (Geng et al. 2021, Sun et al. 2023). Meanwhile, many studies have demonstrated that the ocean currents of South China Sea and adjacent seas have significant impact on the population structure of some Sargassum species, which have strong ability of floating and dispersal (Zhang et al. 2019a). High gene flow in S. thunbergii might be affected by the Kuroshio current and coastal current in China (Li et al. 2017). In addition, the apparent genetic homogeneity of S. polycystum in southeastern Asia might be influenced by the facilitation of complex current system (Chan et al. 2013, Liang et al. 2022). The ocean currents of South China Sea were characterized by seasonal directional changes due to the monsoon effect, and the southwestern currents affected by northeast monsoon in autumn and winter might promote the southward dispersal of S. polycystum (Liang et al. 2022). In this study, since the highly differentiated genetic structure of C. sinuosa, potential dispersal barriers result from the coastal currents and Kuroshio currents in the South China Sea might also play a role on the current distribution pattern of C. sinuosa. In the field work we found the occurrence of C. sinuosa was mainly at the winter and early spring in the northern South China Sea, so we assumed that the gene flow among populations was mostly influenced by the seasonal current in winter and spring by northeastward monsoon. The haplotypes of lineage 1 and 3 were only distributed in the coastal Beibu Gulf (Fig. 1B), which might be related to the limited genetic exchange with eastern side of Hainan, affected by the restriction by semi-closed topography and counterclockwise coastal currents of Beibu Gulf in winter and spring (Supplementary Fig. S11) (Jia et al. 2023, Gao et al. 2024). Also, the westward costal current through the Qiongzhou Strait can block the eastward spread of lineage 1 and 3 (Shi et al. 2002, Yang et al. 2003). Moreover, other lineages might sustain population connectivity promoted by the northeastward Kuroshio current and South China Sea warm current, and we need to enlarge sample collection of C. sinuosa and utilize detailed analysis to study the ocean current’s influences on the seaweed genetic connectivity at larger scale.

Besides, several studies have suggested that environmental heterogeneity, such as sea surface temperature and nutrients variation, can have an impact on the genetic differentiation of marine organism in South China Sea, particularly evident in some coral species (Wu et al. 2021, Li et al. 2022, Jia et al. 2023). In our study, the IBE result indicated that the correlation between genetic variation and environmental variables was not significant. We speculated that C. sinuosa had a broad thermohaline adaption at regional scale compared to more sensitive marine species, thus little affected by the regional environmental changes (Oates 1985, Lee et al. 2014).

As a global marine biodiversity hotspot, the coastal ecosystem of South China Sea has faced severe stress of habitat degradation and biodiversity loss due to recent global climate change and human-mediated disturbances (Juinio-Meñez 2015). In this study, we investigated the genetic diversity of a widely distributed macroalga C. sinuosa and provided some scientific basis of the coastal seaweed diversity conservation in northern South China Sea. Firstly, we found that Sanya in southern Hainan contains two ancestral small lineages with low genetic diversity, which was deeply differentiated with other populations (Table 1, Fig. 1). It revealed that it should be regarded as an independent management unit (MU) and a priority for conservation. Secondly, our findings indicated that Yulinzhou, Dongfang (DF) in southwestern Hainan might serve as a potential central population in the mitochondrial dataset (Supplementary Fig. S2). Additionally, the PCoA result of cp rbcL dataset showed significant deviation from other populations (Supplementary Fig. S7). These results highlighted the importance of seaweed biodiversity conservation in this region. We recommended Yinggehai Basin involved Dongfang in southwestern Hainan as an independent MU, and increase the efforts for marine plant protection with the basis of marine nature reservation. Thirdly, the highest diversity was displayed at Wanning (WN) in southeastern Hainan. So we suggested that it should be involved in the conservation concern of marine plant resource in conjunction with more researches on other marine organism.

In conclusion, we detected deep divergent genetic lineages and complex diversity pattern of C. sinuosa in the northern margin of the South China Sea. Although at regional scale, we found different demographical expansion processes in different lineages in Pleistocene. The phylogeographical pattern might result from the paleoclimatic oscillations and sea level fluctuations in the Pleistocene glacial period, and the ocean currents along the South China Sea could also influence the population genetic structure. Based on the ancestral lineages and high genetic diversity, we proposed Dongfang in western Hainan and Sanya in southern Hainan need extra attention to seaweed diversity conservation and management. In the future, we’ll use C. sinuosa as a good example and shed light on macroalgae phylogeography under historical climate change based on extensive collection work.