A comparison of Heliconius erato populations found that DNA differences and differences in chromatin accessibility, a readout of which DNA regions are open for gene activity, tracked each other most clearly at an intermediate stage of genetic separation. The pattern could mark a period when reproductive barriers are taking shape, but it does not show that any specific region causes them. Across 15 pairwise comparisons, the association strengthened as mean genome-wide Fst rose from about 0.07 to 0.24. The reported association test gave an F statistic of 158.1 and a P value of 5.6 times 10 to the minus fifth. Comparisons involving H. e. chestertonii were consistently nonsignificant.
In the two hybrid-zone comparisons, the evidence was less decisive. KS tests, which compare the shapes of two distributions, did not find a significant shift in Fst values within unique ATAC-seq peaks. For H. e. demophoon and H. e. hydara, the test statistic D was 0.06 with adjusted P = 0.2; for H. e. notabilis and H. e. etylus, D was 0.07 with adjusted P = 1. In practical terms, the few strictly population-specific accessible regions did not stand out as more differentiated than the broader genomic background in either pair.
A broader view of the genome
To look beyond a single reference genome, researchers built a linear pangenome from five newly assembled H. erato genomes and the previously published H. e. demophoon genome. A pangenome is a combined reference that captures sequence found across populations. The resulting comparison covered six populations and used Fst and Dxy scans in 1-kilobase windows, with a density-based method to distinguish clusters of elevated differentiation from isolated noisy windows. That fine-scale approach was designed to catch signals that can be blurred in 50-kilobase comparisons.
That scan found a patchwork of differentiated regions. The demophoon and hydara comparison contained 91 regions covering 2.943 megabases, or 0.427% of the pangenome. The etylus and notabilis comparison contained 76 regions covering 2.916 megabases, or 0.423%. Both comparisons also showed elevated absolute divergence relative to genomic background, a combination the authors interpreted as consistent with reduced gene flow. They cautioned that isolated 1-kilobase Fst windows are noisy and should not be treated on their own as evidence of selection.
Adding sequence to the reference changed the technical picture. The pangenome was 688 megabases, and pangenome mapping generally exceeded 98%, compared with about 92% to 97% for the single reference. Yet the pangenome call rate was lower, about 0.37 to 0.45 versus 0.62 to 0.74 for the single reference, with lower variant density and sequencing depth. The total number of detected variants remained broadly similar, but increased missing data and stringent filtering may have excluded some variants.
Most regulatory differences were shared
The regulatory analysis used ATAC-seq, which identifies stretches of DNA that are accessible to the cell's gene-control machinery. After pangenome processing and quality filtering, the researchers used a conservative consensus rule requiring agreement across samples in a population. Of 29,633 peaks, 29.7%, or 8,799, were shared across all populations. Only 1.48%, or 439, were unique to one population, while 68.82% were polymorphic, meaning their accessibility varied. The result was a large shared regulatory core with many variable sites and few strictly unique ones.
The differentiated regions included established wing-pattern genes and additional candidates linked by annotation to sexual dimorphism, immunity, chromatin regulation, circadian timing, chemosensation and reproduction. The study listed 54 genes in the demophoon and hydara comparison and 88 in the etylus and notabilis comparison, including four phenoloxidase-linked genes. The label candidate matters: location and annotation are suggestive, not functional validation.
The strongest warning against reading too much into uniqueness came from structural-variation checks. A single reference can make a missing sequence look population-specific, so the team tested 439 initial candidate peaks with multi-individual, phylogenetic and resequencing approaches. Only 41% remained restricted to the focal population after phylogenetic screening, and just 9.1%, or 40 peaks, met the strictest nonhomology criteria. Resequencing classified the candidates as 75% shared, 23% variable and 0.5% exclusive. The estimated amount of population-specific nonhomologous regulatory sequence therefore depended strongly on sampling depth and classification rules.
A signal with a narrow window
Forward simulations helped put the uneven signal in context. They indicated that Fst-based searches for barrier loci are most informative at intermediate divergence. High gene flow can make selected and neutral regions look alike, and advanced divergence can reduce the contrast again. In the high-gene-flow condition, with a migration rate of 0.1, the two Fst distributions were indistinguishable. These results come from a simplified model, so they describe what the statistic can detect under the tested assumptions, not a complete reconstruction of natural population history.
Taken together, the sequence, chromatin and structural-variation results support the authors' interpretation of a multilocus, genome-wide pattern consistent with incipient reproductive isolation. But the study identifies associations and candidates, not a proven mechanism: no single gene or accessible region was functionally shown to alter mating, hybrid viability or offspring fitness. The authors describe estimates of nonhomologous regulatory elements as approximate and call for broader population sampling and multi-individual genome assemblies. The central message is therefore twofold: early divergence may be spread across many parts of the genome, and the apparent uniqueness of those regions can shrink when the reference and sampling are expanded.
Paper data and sources
Original title: Multilocus basis of incipient reproductive isolation in hybridizing populations is revealed by pangenomic and epigenetic divergence.
Authors: Angelo A Ruggieri, Francesco Cicconardi, Nicolò Bellin et al.
Journal/Repository: Science advances
Status: Peer-reviewed
First online: 2026-08-21
DOI: 10.1126/sciadv.adz6665
Original paper · Full text