Small, isolated populations are more susceptible to natural disturbances and  declines in genetic diversity. Calamus rhabdocladus has a scattered and fragmented natural distribution across South China, where only a few relict populations still exist. However, no population genetic studies have been conducted to date, and conservation management plans are lacking. In this study, we examined the genetic diversity, genetic differentiation, and environmental adaptability of all sampled individuals of C. rhabdocladus using reduced representation sequencing technology. We observed a moderate level of genetic diversity, with the values of unbiased observed heterozygosity (Ho), unbiased expected heterozygosity (He), Nei's gene diversity (Nei's), and Shannon's information index (I) being 0.180, 0.282, 0.279, and 0.436, respectively. Additionally, the GZ, HY, and FS sites exhibited higher genetic diversity than other sites. Large genetic differentiation was detected in C. rhabdocladus. Structure analysis, phylogenetic tree, and principal component analysis all clustered the samples into three distinct genetic groups, shaped by geographical barriers, heterogeneous environments, and germplasm reproduction characteristics. Redundancy analysis (RDA) identified the driest month (bio14) and the mean temperature of the wettest quarter (bio8) as the two most important environmental factors driving adaptive genetic variation. The latent factor mixed model (LFMM) identified 955 loci significantly associated with environmental adaptation and successfully aligned 469 genes from related species, which were enriched in functional categories involving cell structure, carbohydrate metabolism, and protein degradation. Overall, this study represents the first investigation into the genetic diversity and structure of C. rhabdocladus. Our findings will aid in the conservation and utilization of the species, providing a theoretical basis for future ecological and evolutionary research.

In this study, mature leaf samples from 47 wild individuals were sequenced using RAD-seq. To obtain high-quality sequencing data, the following filtering criteria were applied (Liu, 2021): (1) Reads without 4-8 bp barcode information at the 5' end were removed; (2) After trimming the barcode, reads lacking restriction enzyme recognition sites were filtered out; (3) Reads containing adapter sequences were discarded; (4) Reads with over 40% bases scoring less than 15 were also removed; (5) Finally, reads with more than 10% unknown bases ("N") were excluded. The filtered data were analyzed for read counts, total base counts, GC content, and quality scores (Q20, Q30).For SNP detection and filtering, Stacks v1.48 (Catchen et al., 2013) was used with the following parameters: a minimum coverage depth of 2 for forming stacks, a maximum of 2 mismatches within stacks, and a maximum of 3 mismatches between loci during SNP catalog construction (Liu, 2021). The steps included: (1) Inputting filtered reads into the ustacks program to cluster perfectly matching reads; (2) Compiling loci into a catalog with cstacks, merging alleles; (3) Matching stacks to the catalog using sstacks to determine allelic states for each individual; (4) Performing genetic statistics with the populations program, outputting SNP results in VCF format (Catchen et al., 2013). Based on previous analyses, we retained SNPs meeting the following criteria: (1) Biallelic loci; (2) Missing rate < 0.5 (Guo et al., 2023); (3) Minor allele frequency > 0.05 (Linck et al., 2019); (4) Heterozygosity ratio < 0.8. Finally, the VCF file was converted into multiple formats using PGDSpider v2.1.1.2 for subsequent statistical analyses (Lischer et al., 2012).