Reply to Renner: Meticulous investigation, not sequencing effort, leads to robust conclusion

We thank Renner (1) for summarizing and comparing our study (2) with that of Cui et al. (3). Renner questions the logic of our domesticated/wild classification. As already mentioned (2), without a genetic tool, plant collectors assigned the cultivar/wild status purely based on whether a plant was cultivated. However, many cultures, such as the Amis people in Taiwan, frequently consume wild vegetables and sometimes transplant them into gardens. A plant collector would record such plant as a “cultivar” despite it genetically being a wild accession (e.g., THMC170). All of our analyses followed the grouping from solid genomic evidence.

We agree that a broader sampling would be beneficial, if funding permits. Cui et al. have nicely shown that African cultivars are nested within the South Asian group, suggesting recent introduction. The lack of African samples therefore has little effect on our conclusion. Contrasting our study with that of Cui et al., Renner claims our study to be unconvincing because “it is unclear whether any truly wild populations were sampled” (1). To minimize elements of subjectivity in such debate, one has to combine data from both studies, which we did: The phylogenetic tree remains the same regardless of outgroup species, and all major clades were included in both studies. The wild group in our study is “TR group” in that of Cui et al., who considered TR too divergent to be included in further analyses (3). Cui et al. instead focused on contrasting the “wild” var. muricata versus only the large-fruited extreme of domesticated var. charantia. The distinction between these two groups, however, is morphological without genetic support (3), and many muricata were nested within charantia. While muricata was treated as the wild progenitor because some were basal to charantia (3), our reanalysis grouped those “basal muricata” with our admix-wild accessions, with strong evidence of domesticated-wild admixture and highest heterozygosity. Therefore, it is possible that their basal status results from forcing admixed individuals into a bifurcating tree, and muricata might be once-domesticated feral whose small-fruited alleles were inherited from ancestral polymorphism or introgressed from wild accessions.

Why such difference in divergence time estimation? With ancestral polymorphism, we refrained from using phylogeny-based methods (4). Assembly gaps from Cui et al.’s short-read contigs (N50 about 63 kb) might prevent reliable haplotype structure estimation, which is crucial for demography inferences. Our long-read contigs (N50 about 10 Mb) minimize such problem and allow us to confidently exclude repeat-rich centromeric regions from the analyses. While we used SMC++ (5) capable of including all samples, Cui et al. compared only one individual from each group using multiple sequentially Markovian coalescent (6), which is sensitive to phasing error in heterozygous individuals (5). Finally, we specifically excluded admixed individuals from demographic analysis.

In summary, the claim that we did not sample truly wild populations in contrast to Cui et al. is unfounded. Both studies have the same breadth of sampling, albeit focusing on different clades. The wild group in our study was regarded as another subspecies and excluded from further investigation by Cui et al., who instead focused on muricata, whose basal phylogenetic position and small-fruited phenotype likely resulted from wild-domesticated admixture (2).