Ecological and genomic signatures of the convergent evolution of planktivory in fossil and living reef fishes over deep time Authors: Aintzane Santaquiteria, Dahiana Arcila, Giorgio Carnevale, Melissa Rincon-Sandoval, Fernando Melendez-Vazquez, William T White, Carole C Baldwin, Mark W Westneat, Guillermo Ortí, James C Tyler, Ricardo Betancur-R Published: 2026 ABSTRACT AbstractTransitions from benthic to pelagic habitats often coincide with shifts in diet, yet the genomic basis and macroevolutionary dynamics of these changes remain poorly understood. Surgeonfishes and allies (Acanthuriformes), with their rich fossil record, dietary diversity, and compact genomes, provide a powerful model to investigate these transitions across deep time. We integrate genomic data from 80 extant species with morphological data from 32 fossils to construct total-evidence, tip-dated phylogenomic trees and reconstruct ancestral diets and geographic ranges. We identify at least seven independent origins of planktivory and four reversals, with transitions concentrated in the Tethys and Indo-Pacific oceans. Contrary to the view that planktivory represents an evolutionary dead end, we find higher reversal rates and lower extinction in planktivores. Time-dependent models support both climate and phylogenetic effects on planktivorous lineage diversification. Using a newly assembled chromosome-level genome and 44 additional assemblies, we apply phenotype-aware selection models to identify 39 genes under convergent positive selection in planktivorous lineages. Six of these genes show accelerated evolution tied to metabolism and dietary specialization, and also exhibit codon-level parallelism or convergence. Our findings illuminate the ecological, biogeographic, and genomic basis of repeated dietary shifts and challenge assumptions about evolutionary constraints in pelagic environments.Subject terms: Phylogenetics, Comparative genomics Introduction Evolutionary convergent adaptations often arise in response to shared ecological pressures or opportunities in similar environments1. The shift from benthic (bottom-dwelling) to pelagic (open-water) habitats represents a major ecological transition in aquatic organisms and is frequently accompanied by morphological and trophic specialization2–4. Among these, the evolution of a zooplanktivorous diet (planktivory) stands out as a deterministic shift influencing feeding strategies, ecological roles, and functional traits across diverse freshwater and marine fish clades. Notable examples include repeated benthic–pelagic transitions in lake fishes, such as African and Central American cichlids, whitefish, and three-spined sticklebacks, where bottom-feeding benthic forms have evolved into pelagic specialists adapted for planktonic feeding5–8. While these cases highlight striking divergence along this environmental axis in relatively recent evolutionary timeframes, similar transitions have also occurred at deeper phylogenetic scales. For example, snappers and fusiliers (Lutjanidae9), grunts (Haemulidae10), surgeonfishes (Acanthuridae11–13), and damselfishes (Pomacentridae14) have independently shifted from benthic to pelagic habitats multiple times over the last 60 million years (Ma). Although such ecological transitions are often associated with convergence in ecomorphological traits, recent large-scale studies have found little evidence for strong morphological convergence among reef fish planktivores15,16. Trophic shifts are also common among reef fishes, with dietary identity emerging as a key feature of ecological function and biodiversity patterns in reef ecosystems17. Planktivorous fishes, in particular, exhibit unique traits tied to their feeding strategy, such as filter-feeding. Although many species exhibit planktivory in their larval stages, only those with specific adaptations persist as obligate planktivores into adulthood18. Marine biodiversity is generally higher in benthic than pelagic environments19,20, suggesting that pelagic planktivory may constitute an evolutionary dead end; however, this view is not consistently supported across all groups. Adopting a water column lifestyle may offer short-term advantages, but in the long term, it can create an evolutionary trap due to reduced speciation, elevated extinction, or both9,21, particularly in structurally homogeneous pelagic habitats with fewer opportunities for ecological partitioning. While planktivores may initially diversify by exploiting vacant niches with little competition, long-term persistence in a resource-scarce and relatively uniform environment could limit diversification potential22. This hypothesis is supported in some groups, such as snappers and fusiliers9, yet remains ambiguous in others, including wrasses 23 and damselfishes, the latter showing a particularly strong asymmetry through an intermediate dietary ecotype14. Understanding these trophic shifts also provides insight into the adaptive capacity and evolutionary responses of organisms to environmental change, including both past paleoclimatic fluctuations and potential future impacts of climate change. Past fluctuations in ocean temperatures have been shown to influence the diversity of marine organisms, with drastic temperature changes followed by either extinction24–26 or increasing diversification27, but their role in shaping trophic transitions remains unexplored. While ecological factors are known to be tightly associated with trophic transitions, the genetic mechanisms underlying these shifts remain less understood (but see refs. 28–30). A central question is whether similar dietary specializations across lineages arise from convergent changes at the molecular level31–34. Genomic innovations have been linked to a range of ecological adaptations, including vision35, habitat use36, and feeding ecology37,38. Previous research, primarily in mammals and insects, suggests that dietary evolution is associated with gene family expansions39, changes in gene copy number40, or loss of gene function41 in pathways related to metabolism. Genes linked to appetite regulation, gastrointestinal function, enzymic digestion, and biosynthesis of fatty acid and cholesterol have been identified in zebrafish42 and clownfish29. As transitions to the water column are often associated with both a shift to a planktivore diet and morphological changes, previous studies in cichlids, whitefish, or sticklebacks have also identified genes linked to body elongation and caudal fin shape30,43, gill raker counts28,44,45, and jaw length46. However, comparable studies in marine fishes remain sparse, largely due to a lack of high-quality genome assemblies for reef lineages. In parrotfishes, for example, genomic analyses revealed expansions in detoxification genes, likely linked to their highly specialized herbivorous diet and biomineralized teeth47. Reef fishes in the order Acanthuriformes sensu stricto48, which includes the charismatic unicornfishes, surgeonfishes, and tangs (Acanthuridae), luvars (Luvaridae), and Moorish idols (Zanclidae), offer a unique opportunity to investigate dietary transitions across both extant and fossil species, as well as the genomic underpinnings of these transitions in living taxa. This is facilitated by their relatively small genomes ( ~ 700 Mb), well-documented ecologies, and an exceptional fossil record comprising over 30 species, which provides a temporal framework to examine dietary evolution49,50. While previous comparative studies have clarified the role of morphology and diversification dynamics in acanthuriform trophic evolution11,12,51,52, integrating fossil and extant data offers deeper insight into the clade’s macroevolutionary and biogeographic history. Two studies—one including 18 fossil taxa subjectively placed from photographic records alongside 41 extant species50, and another with 25 fossil taxa placed according to available literature alongside 72 extant species52, rather than explicit morphological datasets—proposed that surgeonfishes diversified in two distinct phases: first during the Paleocene–Eocene ( ~ 66–34 Ma), potentially due to ecological release after the Cretaceous–Paleogene (K–Pg) mass extinction, and again in the Oligocene–Miocene ( ~ 23–5 Ma), coinciding with the rise of extensive coral reef systems53. Trophically, acanthuriforms are broadly divided into non-planktivores (i.e., herbivores and detritivores feeding on turf algae, macroalgae, or detritus) and planktivores that feed primarily on zooplankton17. Although the group is widespread across tropical marine realms, planktivorous species are most common in the Indo-Pacific12,26,52. Prior studies have suggested that planktivory evolved at least five times independently within surgeonfishes11,12, however, these inferences were based on limited phylogenetic resolution from mitochondrial and nuclear markers and did not incorporate fossil data. Moreover, the genomic basis of trophic transitions in marine fishes remains poorly understood in a macroevolutionary context. Here, we investigated the tempo and mode of planktivory evolution in fossil and extant acanthuriforms. Specifically, we (1) estimated the timing, frequency, and geographic context of dietary shifts; (2) tested whether planktivory represented an evolutionary dead end; (3) assessed correlations between trophic transitions and paleoclimatic trends; and (4) identified candidate genes under positive selection associated with transitions to planktivory. We constructed time-calibrated phylogenies using a total-evidence dating approach that integrated genome-wide data from nearly 1000 loci in 80 extant species with 107 morphological characters scored across 32 fossil and 19 extant taxa. From these phylogenies, we reconstructed ancestral diets and geographic ranges, estimated diet-dependent diversification rates, and evaluated the impact of ancient ocean temperatures on trophic shifts. We also tested the hypothesis that acanthuriform trophic shifts have been shaped by past geological and climatic events and the extent to which transitions to a planktivorous diet represent an evolutionary dead end. To examine the genomic basis of planktivory, we generated a chromosome-level genome assembly for a surgeonfish representative and short-read assemblies for 44 additional acanthuriform species. Using phylogenetic genotype-to-phenotype (PhyloG2P) analyses54, we identified genes under positive selection that have convergently evolved across planktivorous lineages and along branches coinciding with this dietary transition. Results and discussion Phylogenomic total-evidence dating resolves acanthuriform relationships and reveals post-K–Pg diversification To infer acanthuriform relationships, we assembled three complementary datasets. The morphology-only matrix includes 107 characters newly scored for 32 fossil and 19 extant species, along with 5 extant outgroups. The molecular-only matrix contains 80 ingroup species (representing approximately 93% of the group’s diversity) and 9 outgroups. Of these, 65 species have data for 998 “FishLife” exon markers55, which also incorporate 22 legacy markers, whereas 24 species include data only for legacy markers. Finally, we created a combined dataset that integrates both morphological and molecular data. Using several complementary methods, we consistently retrieved the same family-level patterns of relationship from the genomic data. Concatenation-based maximum-likelihood trees built in RAxML v.8.2.11 (ref. 56; Supplementary Figs. 1 and 2), as well as coalescent trees inferred with ASTRAL-III (ref. 57; Supplementary Figs. 3 and 4), all resolve one well-supported acanthuriform clade in which Luvaridae branches first, followed by the sister pair Zanclidae + Acanthuridae. The combined morphology + molecular dataset yields the same topology (Supplementary Fig. 7). By contrast, morphology-only trees generated with TNT v1.558 and RAxML split Zanclidae and Acanthuridae and reposition Luvaridae, indicating that skeletal traits alone can be misleading at this phylogenetic depth (Supplementary Fig. 6). Specifically, Luvarus is grouped with the largely planktivorous Naso, a pattern that may reflect convergence associated with planktivory and illustrates how morphology alone can obscure true evolutionary relationships, reinforcing the importance of incorporating genomic data. Most genera remain monophyletic; the main exceptions are Ctenochaetus and Acanthurus (both surgeonfishes) consistent with previous studies52,59,60 and the fossil genus Tylerichthys. Gene-concordance analyses indicate that individual genes overwhelmingly support the molecular topology, reinforcing the robustness of the genomic signal (see Supplementary Fig. 5). We estimated divergence times using a Bayesian tip-dating (total-evidence) framework under the Fossilized Birth–Death model in MrBayes v.3.2.7a61. To account for topological uncertainty, we assembled five independent datasets integrating the 107 morphological characters and ~200 non-overlapping molecular markers for 112 taxa (80 living, 32 fossil; Supplementary Fig. 8 and Supplementary Table 2). Each tip-dating analysis ran for over a year to ensure convergence (effective sampling size or estimated sample size (ESS) values > 200, Supplementary Table 4). From each of the five runs, we retained 100 post-burn-in trees, yielding a total of 500 trees used to account for tree uncertainty in downstream comparative analyses. Two Eocene fossils, †Gazolaichthys vestenanovae and †Padovathurus gaudryi, previously assigned to Acanthuridae52, were placed within that family in our morphology-only phylogeny (Supplementary Fig. 6) but fell outside Acanthuridae in the combined morphological + molecular analyses, instead forming the sister group to Zanclidae and Acanthuridae. Given anatomical evidence supporting the latter arrangement62, we present comparative results based on this topology (the two Eocene fossils sister to Zanclidae + Acanthuridae: Scheme 1 hereafter), with alternative results (the two Eocene fossils sister to Acanthuridae: Scheme 2 hereafter) reported in the Supplementary Information (Supplementary Figs. 10, 14, 16, 17, 19–21 and Supplementary Data 3–5). The resulting trees (Fig. 1A and Supplementary Fig. 9) were topologically stable and broadly consistent with previous molecular11,48,51,52,59,60,63 and morphological64 studies. Diversification of fossil and extant acanthuriforms and evolutionary origins of planktivory. A Ancestral diet reconstruction based on a posterior distribution of 500 trees (999 loci) from Scheme 1 trees for 112 species (32 fossil and 80 extant) reveals multiple transitions between non-planktivorous and planktivorous diets across the acanthuriform phylogeny. Lineage-through time plot for each trophic guild show two major waves of diversification: one during the Paleocene–early Eocene and another in the early Miocene. The darker lines indicate the mean number of lineages over time for each trophic guild, and color indicates dietary strategy. B Ocean temperatures (purple) and phylogenetic signal (yellow) show cofounding effects in the evolution of planktivory. AIC weights are shown for climate-dependent (curves for tropical in red and global average temperature in black, illustrated in (A)) and independent models based on the 500 alternative trees as well as the MCC tree (indicated with a black arrow). C Model-averaged tip rates estimated with HiSSE show similar net diversification rates between trophic guilds, although non-planktivores experienced higher extinction rates. Box plots summarize speciation, extinction, and net diversification tip rates for extant species under the MCC tree, comparing non-planktivores (n = 61) and planktivores (n = 19). Box plots indicate median (middle line), 25th, 75th percentile (box), and 5th and 95th percentile (whiskers) as well as outliers (single points). P-values calculated using two-sided phylogenetic ANOVA with Bonferroni correction are shown at the bottom right of each panel. * p-value < 0.05. Arrows between fish silhouettes represent transition rates between dietary strategies estimated with HiSSE. Plio. Pliocene, Q. Quaternary, EB early burst, BM Brownian motion. Fish silhouettes were obtained from R package fishualize (https://github.com/nschiett/fishualize/tree/master). Source data are provided as a Source Data file. According to the time-calibrated trees, acanthuriforms originated in the Paleocene around 63.6 Ma based on the maximum clade credibility (MCC) tree (95% highest posterior density or HPD: 64–63.3 Ma), shortly after the K–Pg extinction. This estimate is notably younger than previous gene-only analyses, which placed the origin at approximately 80 Ma52. Luvaridae originated at 62.1 Ma, followed by Acanthuridae (unicornfishes, surgeonfishes, and tangs) at 61.7 Ma, and Zanclidae at 57.8 Ma. Within acanthurids, most fossil genera (e.g., Tauichthys, Proacanthurus) arose between 60–50 Ma, while extant genera diversified later: Naso at 26.4 Ma, Paracanthurus at 25.0 Ma, Acanthurus at 23.2 Ma, Zebrasoma at 17.9 Ma, and the clade uniting Acanthurus and Ctenochaetus at 11.6 Ma. These results confirm a post-K–Pg radiation of major lineages (Fig. 1A) and emphasize the value of integrating fossil and molecular evidence for reconstructing evolutionary history65. Multiple gains and losses of planktivory trace to early Tethyan origins and subsequent Indo-Pacific spread The recurrent phenomenon of trophic transitions from an herbivorous/detritivorous diet to a planktivorous one is not limited to surgeonfishes, unicornfishes, and luvars, but extends to various other reef fish groups, including damselfishes, groupers, snappers, triggerfishes, and wrasses18, as well as freshwater species flocks like cichlids, whitefish, and stickleback5,6,66. With the newly inferred time trees including fossil and living acanthuriform species, we reconstructed their diets to investigate the number of times planktivory evolved, as well as the regions and timing of these transitions. For extant species, we collected diet data from the primary literature11,12 and FishBase67. For fossils, we inferred diet based on paleoenvironmental information rather than tooth morphology, a common approach68,69 that can be misleading in this group. In extant species, both planktivorous and non-planktivorous fishes often possess multi-denticulate or conical teeth (Supplementary Fig. 11), limiting the reliability of dental characters for reconstructing trophic ecology. Fossils found in oceanic sediments were coded as planktivores, while those from limestones were coded as non-planktivores49,70. Ancestral diet reconstruction using the “make.simmap” function in the R package phytools71 (SIMMAP hereafter) was conducted under the best-fit equal-rates (ER) model for both phylogenies; ER was marginally favored for the extant-only tree (AICw = 0.51 vs. 0.48 for all rates different or ARD) and similarly for the tree including fossils (AICw = 0.55 vs. 0.45 for ARD). Under both models, we inferred six independent transitions to planktivory in the extant-only tree (Supplementary Figs. 12 and 15). When fossils were incorporated, the number of independent origins increased to at least seven (range: 7–9 across posterior distribution [PD] trees). These included one origin (1–3 in PD trees) in extinct lineages, three (2–3) in stem clades that include both fossil and extant species, and three (3–4) in extant-only clades (Fig. 1A and Supplementary Figs. 13 and 16). We also identified three reversals to non-planktivory in the extant-only tree, and four (3–5) when fossils were included. These trophic transitions span most of acanthuriform history, occurring between 62.1 and 3.6 Ma (Fig. 1A). To contextualize these transitions geographically, we conducted ancestral range reconstruction analyses in the R package BioGeoBEARS72. We evaluated 12 biogeographic models and ran analyses using either 1673 or 12 Ma74–76 as the final closure of the Tethys Seaway, with both timepoints yielding similar results (see Fig. 2 and Supplementary Fig. 18). Under the best-fit model (BAYAREALIKE+w; Supplementary Table 5), acanthuriforms were inferred to have originated in the Tethys Sea (Tet). From there, five lineages (from all genera except Ctenochaetus) expanded into the Indo-Pacific, and one lineage (Eonaso) dispersed into the Western Atlantic (WA). Subsequent dispersal from the Indo-Pacific included at least six events to the Tropical Eastern Pacific (TEP), one to the WA, and two to the Eastern Atlantic (EA). Two additional lineages crossed the Atlantic: (i) the ancestor of Acanthurus chirurgus, A. bahianus, and A. tractus, which dispersed from the EA to the WA; and (ii) an ancestral Prionurus biafraensis lineage that moved from the TEP to the WA. This biogeographic history aligns with a previous study on surgeonfishes and many other circumglobal shallow marine fish groups9,52,77,78. Although planktivory has evolved repeatedly across the world’s oceans, in acanthuriforms the earliest transitions (4–5 events) occurred in the Tethys Sea. More recent transitions (3–4 events) took place predominantly in the Indo-Pacific, particularly in the Western Indian Ocean (WIO), Central Indo-Pacific (CIP), and Central Pacific (CP) regions, characterized by complex reef topography that offers predator refuge and promotes high productivity through dynamic water flow79,80. No transitions were detected in the WA, EA, or TEP (Fig. 2). These results highlight the repeated and spatially structured evolution of planktivory in acanthuriforms, shaped by both deep-time biogeography and regional ecological opportunity. Historical biogeography of acanthuriforms and the geography of trophic transitions. Early transitions to planktivory occurred in the ancient Tethys Sea, while more recent ones took place in the Indo-Pacific Ocean. Ancestral ranges were estimated in BioGeoBEARS using the best-supported biogeographic model (BAYAREALIKE+w) applied to 20 trees subsampled across the five subsets from Scheme 1 and using the MCC tree as fixed topology. Boxes at internal nodes are color-coded by area or combination of areas with the highest likelihood. Boxes at the tips are colored according to the present-day distribution of each species. Branches undergoing a transition to planktivory are indicated in blue, while those reverting to non-planktivory are shown in green. The paleomaps (obtained from GPlates, Global_Scotese, https://www.gplates.org/) illustrate the location of the Tethys Sea at 49 million years ago (Ma), and the Indo-Pacific at present-day. WIO Western Indian Ocean, CIP Central Indo-Pacific, CP Central Pacific, TEP Tropical Eastern Pacific, WA Western Atlantic, EA Eastern Atlantic, Tet Tethys Sea. No evolutionary dead end associated with the evolution of planktivory Planktivory is considered the principal evolutionary destination among trophic transitions in reef fishes17, often assumed to represent a terminal trophic state. This expectation forms the basis of the evolutionary dead end hypothesis81, which in our system yields two testable predictions: (i) planktivorous lineages should exhibit lower diversification (either through reduced speciation, elevated extinction, or both) relative to non-planktivorous lineages; and (ii) trophic transitions should be directionally biased, with shifts into planktivory occurring more frequently than reversals to non-planktivory. To evaluate these predictions, we estimated diversification and transition rates between non-planktivory and planktivory by conducting state-dependent speciation and extinction (SSE) analyses using HiSSE82 on the MCC tree, including only extant species (80 tips). To incorporate model uncertainty, we model-averaged rates for all tips and nodes across five models using AIC weights (Supplementary Table 6). Our results indicate that both net diversification (p = 0.893 for the MCC tree and 0.001–0.995 for PD trees) and speciation rates (p = 0.893 for the MCC tree and 0.001–0.996 for PD trees) are similar between non-planktivorous and planktivorous species, while extinction rates are slightly lower in planktivores (p = 0.001 for the MCC tree and 0.001–0.895 for PD trees; Fig. 1C and Supplementary Table 7). Transition rates from non-planktivory to planktivory (0.010 for the MCC tree and 0.007–0.011 for PD trees) were lower than the reverse (0.026 for the MCC tree and 0.013–0.034 for PD trees; Supplementary Table 7). When we compared these results with transition rates from ancestral diet reconstructions, we found them to be highly consistent. Including fossils, SIMMAP estimated an ER rate of 0.015, with ARD rates of 0.012 from non-planktivory to planktivory and 0.025 for the reverse; excluding fossils, the ER rate was 0.012, with ARD rates of 0.010 and 0.025, respectively. These values closely match the HiSSE estimates (0.010 and 0.026), indicating that differences between approaches reflect model structure and sensitivity rather than conflicting evolutionary signals. Although the number of transitions from non-planktivory to planktivory has exceeded reversals over the course of acanthuriforms’ evolutionary history, it is noteworthy that planktivorous lineages are generally younger than their non-planktivorous counterparts. Consequently, younger lineages, such as Naso (unicornfishes), have experienced reversals to the ancestral condition (non-planktivory), leading to higher rates of transitions from planktivory to non-planktivory compared to the reverse direction. Some species in this genus, especially N. brevirostris and N. vlamingii, feed on benthic macroalgae as juveniles and become planktivorous later in life83,84, highlighting an exceptional dietary lability in Naso. These reversals occurred alongside similar diversification rates among both trophic guilds, contradicting our prediction based on an evolutionary dead end scenario. While planktivory is often considered an evolutionary destination17, some unicornfishes revert to their ancestral trophic states without further dietary specialization, representing a notable exception to this expectation. The transition to a planktivorous diet often coincides with a shift from bottom-dwelling to mid-water habitats, and it is conceivable that reliance on a specific dietary resource, especially under changing environmental conditions or competitive pressures, may constrain the adaptive potential of planktivorous species. Together, these findings suggest that planktivory does not represent an evolutionary dead end in acanthuriforms, but, rather, a dynamic state with the potential for ecological and evolutionary reversibility. Paleoclimate effects on the origin of planktivory We tested whether shifts in ocean temperatures over the last 64 million years influenced the trophic diversification in acanthuriforms, with a particular emphasis on the emergence of planktivorous lineages. During this period, global temperatures fluctuated significantly85, driven by paleogeographic, tectonic, and oceanographic changes86. To explore these patterns, we generated lineage-through-time (LTT) plots for planktivores and non-planktivores using 500 tip-dated trees in phytools71, and compared them with global average and tropical temperature curves86 (Fig. 1A and Supplementary Figs. 14 and 17). We then statistically assessed correlations by leveraging a time-dependent threshold analysis, fitting various paleoclimate-dependent and -independent diversification models87 to the MCC tree, incorporating species’ dietary guilds and using both temperature curves. These analyses revealed that acanthuriforms gradually started diversifying during the Paleocene and early Eocene (66–50 Ma), particularly benthic lineages whose trophic ecology (based on inference from extant relatives11) might have involved filamentous or macrophytic algae as well as detritus and microbes. The early Eocene period (56–48 Ma) marked a significant phase of elevated temperatures88, with ocean temperatures surging by 6 °C85. This warm interval encompassed both the Paleocene–Eocene Thermal Maximum (~56 Ma) and the Early Eocene Climatic Optimum (EECO; 53–49 Ma)89, the latter coinciding with Bolca (50.5–48.5 Ma) in the Tethy Sea, a rich site of reef fish fossils49,90. This climatic phase has been linked to elevated extinction rates in some marine fishes26 and foraminifera91, coincident with a coral reef crisis triggered by ocean acidification and a consequent decline in reef growth92. Several benthic-feeding acanthuriform lineages are restricted to this interval in the fossil record (Fig. 1A49,50), suggesting a contraction or turnover of benthic trophic lineages during this time. In contrast, zooplanktivorous species in the water column appear to have been less affected. Following the EECO ( ~ 49 Ma), global cooling set in and was accompanied by increased diversity in planktonic groups such as foraminifera, diatoms, and radiolarians93,94. This shift may have created new ecological opportunities for planktivores, facilitating their diversification, as supported by our results showing elevated origination rates for planktivorous lineages during the middle to late Eocene (44–30 Ma; Fig. 1A, B). However, it is important to recognize that strong taphonomic biases, such as the dominance of non-planktivorous acanthuriform fossils in Bolca deposits49,90, may exaggerate the apparent link between temperature shifts and extinction patterns. A second wave of diversification occurred in the early Miocene ( ~ 20 Ma), when acanthuriform planktivores, alongside other reef fishes, increased in lineage diversity, although non-planktivores exhibited a steeper diversification slope (Fig. 1A). This diversification coincided with new reef configurations, coral diversification, and major changes in Indo-Pacific Ocean circulation12,17,79,95,96. Additionally, persistently warm tropical temperatures over the past ~20 Ma85,97 increased algal turf cover, boosting food availability and supporting higher abundances of herbivorous reef fishes98. Although we observe a potential temporal correlation between the evolution of planktivory and major climatic transitions, our model fitting analyses indicate split support between climate-dependent and climate-independent models. Based on the MCC tree, the phylogenetic signal (lambda) model (AICw = 0.55 vs. 0.51) and the climate-dependent model (AICw = 0.44 vs. 0.48) both fit better than climate-independent models such as Brownian motion (AICw = 0.005 vs. 0.004) and early burst (AICw = 0.002 vs. 0.002; Fig. 1B and Supplementary Table 8). Across the 500 posterior trees, however, the lambda model received slightly stronger support, with AICw averages of 0.63 (global temperatures) and 0.82 (tropical temperatures), compared to 0.34 and 0.15, respectively, for the climate-dependent model (Fig. 1B and Supplementary Table 8). Very similar results were obtained under Scheme 2 fossil placements (Supplementary Figs. 14, 17, and 21 and Supplementary Data 3–5). Overall, these findings suggest that ancient climate shifts and lineage-specific evolutionary history both likely contributed to the timing and tempo of planktivory diversification in acanthuriform fishes. Signatures of convergence in positive selection following transitions to planktivory Understanding whether trophic transition convergence is mirrored at the molecular level (i.e., through repeated use of the same genes or pathways) could offer critical insights into the genetic mechanisms underlying adaptation to planktivory. To explore this, we investigated genome-wide signatures of positive selection associated with dietary transitions in acanthuriforms using PhyloG2P approaches. First, we sequenced, assembled, and annotated a chromosome-level reference genome of A. chirurgus (doctorfish tang, Acanthuridae) using PacBio long-reads, Hi-C, and IsoSeq sequencing technologies. This resulted in a high-quality phased genome with 15 putative chromosomes for each haplotype. Haplotype 1 has a size of 776.87 Mb, with scaffold N50 of 47.48 Mb, while haplotype 2 has a size of 664.96 Mb, scaffold N50 of 47.66 Mb (Supplementary Fig. 22). We assessed the genome completeness in BUSCO99,100 based on the Actinopterygii database (odb12). Before annotation, BUSCO showed a completeness score of 98.1% (Supplementary Fig. 23). When BUSCO was run on the annotated genome and predicted 23,548 protein-coding genes, the completeness was 85.7%. Second, we sequenced a total of 48 short-read genomes through Illumina. We assembled them in MaSuRCA v.4.0.8101 and enhanced scaffolding of each genome by mapping to the reference genome in RagTag v.2.1.0102 to reorder, reorient, and fill gaps. After quality control and removal of duplicated A. chirurgus, we retained 44 assemblies, whose BUSCO completeness scores ranged from 67 to 97.6% (Supplementary Fig. 23). To detect genes under positive selection associated with planktivory, we used the dN/dS ratio as implemented in HyPhy103. We analyzed 3926 single-copy orthologs (2542 derived from the annotated reference genome and 1384 from the BUSCO databases) filtered through OrthoFinder and rigorous quality control procedures (see “Methods”). We targeted selection along the five branches where transitions to planktivory were inferred based on ancestral diet reconstructions (Fig. 3). Our approach combined two complementary methods: the adaptive Branch-Site Random Effects Likelihood model (aBSREL104), which detects selection on specific branches, and BUSTED-Phenotype (BUSTED-PH105), which compares selection pressures between foreground (planktivore) and background (non-planktivore) lineages (Supplementary Fig. 23). This two-step filtering strategy allowed us to identify positively selected genes (PSGs) evolving under selection specifically during or following transitions to planktivory. After correcting for multiple testing using a false discovery rate threshold (p < 0.05) and combining both aBSREL and BUSTED-PH results, we identified a total of 302 PSGs along branches associated with planktivory transitions (Supplementary Data 4). Among this, 39 convergently underwent positive selection in at least two lineages (Fig. 3 and Supplementary Data 5). Convergent positive selection associated with transitions to planktivory. A Schematic representation of the aBSREL analysis used to identify positively selected genes (PSGs), with branches associated with transitions to planktivory shown in blue. For the genus Naso, two scenarios were analyzed: Scenario 1 captures selection occurring during the initial transition to planktivory (stem branch only), while Scenario 2 includes both stem and crown branches, capturing selection during and after the transition. The species for which we generated a reference genome (Acanthurus chirurgus) is shown in bold. The UpSet plot (right) illustrates the number of PSGs (combing both scenarios) shared across planktivorous lineages, with Naso exhibiting the highest degree of overlap, likely reflecting its extended evolutionary history as a planktivore (see main text). Source data are provided as a Source Data file. B Table listing the 39 PSGs convergently evolving in at least two lineages, along with their associated biological processes based on Gene Ontology annotations. Metabolic and cellular processes were the most frequently represented functional categories. Light blue cells indicate PSGs shared by two lineages; dark blue indicates convergence across three. Cell differentiation and growth function are grouped under the category “developmental processes”. Among the lineages analyzed, Naso exhibited the highest number of shared PSGs with other planktivorous taxa (Fig. 3A). However, rather than indicating stronger molecular convergence per se, this pattern likely reflects the extended evolutionary duration of planktivory in Naso, which includes multiple species and relatively long branches, thereby providing more opportunities for positive selection to accumulate. Similarly, elevated convergence observed between Naso and Luvarus, or Naso and Paracanthurus hepatus, may result from this same “age effect,” as both Luvarus and P. hepatus represent some of the most ancient, long-branched planktivore lineages. This interpretation is consistent with recent findings that emphasize the importance of time and clade age in shaping patterns of molecular convergence across ecological transitions38,45. Gene Ontology (GO) enrichment analyses based on the PANTHER106 and ZFIN107 databases revealed 10 biological processes associated with these PSGs (Fig. 3B). The most represented categories were metabolic (24 genes) and cellular processes (24 genes), followed by developmental/growth/cell differentiation processes (13 genes) and biological regulation (12 genes). Other notable functions included localization (10 genes) and response to stimulus (5 genes). Less frequent categories included multicellular organismal processes (3 genes), homeostatic processes (3 genes), and immune system processes (3 genes). All in all, these results suggest that dietary transitions to planktivory involved diverse physiologic and developmental pathways, rather than a single convergent genetic mechanism. Accelerated molecular evolution in metabolic genes linked to transitions to planktivory While genes related to metabolism have been implicated in dietary adaptations across various fish species, such as zebrafish42, goldfish108, parrotfish47, and clownfish29, it remains unclear whether these metabolic genes also exhibit signatures of rapid evolution in association with repeated shifts to planktivory. To address this, we tested whether the 39 convergently evolved PSGs showed accelerated or decelerated evolution in planktivorous lineages by estimating the relative evolutionary rates (RERs) in HyPhy. We detected ten genes with rate shifts associated with trophic transitions: nine exhibiting accelerated rates and one showing decelerated rates (Fig. 4 and Supplementary Data 5). Six of the accelerated genes are associated with metabolic processes and are likely involved in the digestion of zooplankton and small crustaceans, key components of a planktivorous diet (Figs. 3 and 4). Notably, erg28 (ergosterol biosynthesis 28), a gene involved in cholesterol biosynthesis and membrane homeostasis, exhibited the strongest signal of evolutionary acceleration, largely driven by Naso vlamingii, N. brevirostris, and N. annulatus109. Selection on erg28 may reflect adaptations to optimize energy metabolism and nutrient absorption under ecological pressures such as prey availability or predation risk. Similarly, nr5a5 (nuclear receptor subfamily 5, group A, member 5), a regulator of cholesterol metabolism and steroidogenesis expressed in the digestive tract and linked to appetite regulation, may reflect adaptation to a lipid-rich zooplanktonic diet110, with acceleration likely influenced by species such as Acanthurus nubilus, Luvarus imperialis, Naso minor, and N. thynnoides. Evolutionary rate shifts and convergence patterns in positively selected genes, including parallel and convergent evolution at selected sites. A The scatterplot shows relative amino acid substitution rates in foreground (planktivorous) versus background (non-planktivorous) branches for each convergently evolved positively selected gene (PSG). Dots above the diagonal indicate acceleration in planktivores. Dot color represents the FDR-corrected q-value (as –log₁₀(q)); bubble size reflects the magnitude of rate shift [max(foreground/background, background/foreground)]. The asterisk following the gene name denotes that the gene is involved in metabolic processes. Key metabolic functions linked to the accelerated genes are highlighted in blue. The Palette surgeonfish (Paracanthurus hepatus) represents an example of a planktivore species. The illustrations of the surgeonfish, copepod, and krill were done by Enara Santaquiteria. Source data are provided as a Source Data file. B The phylogeny maps sites under episodic positive selection (identified with MEME), highlighting codons showing parallel evolution (same amino acid change across lineages, in orange) and convergent evolution (different amino acid changes at the same site, in purple). Sites with both colors suggest contributions from both mechanisms (e.g., when three lineages are involved). Codon positions are shown for each labeled site. Other accelerated genes include prss, a serine protease involved in protein digestion and expressed in the digestive system29, which showed signals of selection in three Naso species. Slc1a4 (solute carrier family 1 member 4), a transporter of neutral amino acids like serine and alanine, abundant in zooplankton and expressed in the gut111,112, was accelerated in N. mcdadei and Paracanthurus hepatus. Tha1 (threonine aldolase 1), involved in amino acid metabolism and gluconeogenesis, is accelerated in N. hexacanthus, while kbtbd12 (kelch repeat and BTB domain containing 12), linked to protein ubiquitination and energy regulation, showed acceleration in N. vlamingii and N. annulatus107. Kcna2b (potassium voltage-gated channel, shaker-related subfamily, member 2b), though unrelated to metabolism, plays a role in neuronal signaling and action potential propagation and was accelerated in Luvarus107. Ttc36 (tetratricopeptide repeat domain 36), involved in developmental processes such as cilium assembly and otolith morphogenesis, also showed acceleration in Luvarus and N. vlamingii, despite no known digestive role107. Lastly, ap1ar (AP-1 complex-associated regulatory protein), involved in intracellular transport and vesicle trafficking (vesicle targeting from trans-Golgi network to endosomes), is accelerated in N. brevirostris. The only gene with a significant deceleration was pip4p2 (Phosphatidylinositol-4,5-Bisphosphate 4-Phosphatase 2), involved in lipid signaling, energy use, and nutrient uptake; this pattern was driven by A. nubilus, P. hepatus, and N. minor107. Altogether, the digestive functions and expression profiles of these genes highlight their likely role in facilitating dietary specialization during repeated evolutionary transitions to planktivory. We further examined whether any of the nine accelerated genes exhibited evidence of episodic diversifying selection at specific codons, and whether these changes reflected convergence or parallelism. Using MEME113 we identified 93 positively selected sites across these genes, with counts ranging from 19 in prss to 3 in erg28 (Supplementary Data 5). While many sites were flagged, only a subset showed clear evidence of convergence, parallelism, or both—potentially contributing to the repeated evolution of planktivory (notably, none were detected in erg28; Fig. 4). Site convergence refers to independent planktivory transitions resulting in different amino acid substitutions at the same codon, whereas site parallelism involves identical amino acid changes across transitions. Several genes showed both patterns. For example, kcna2b exhibited site convergence between Luvarus imperialis and Acanthurus nubilus, and site parallelism between L. imperialis and Paracanthurus hepatus. In prss, slc1a4, ttc36, and tha1, both convergence and parallelism were observed across L. imperialis, members of the genus Naso, and in some cases P. hepatus. Notably, prss and nr5a5 had the highest number of positively selected sites (6 and 5, respectively), most of which reflected site convergence. These results suggest that episodic diversifying selection has significantly influenced the evolution of these genes, with both convergent and parallel mechanisms shaping their sequence variation, potentially as adaptations to specific dietary niches or environmental pressures. More broadly, our findings highlight how macroevolutionary patterns of trophic specialization in acanthuriforms emerge from a dynamic interplay of ecological filtering, mutation bias, and lineage-specific evolutionary history, funneling adaptation along repeatable yet historically contingent molecular trajectories34,45,114. No statistical associations between repetitive elements and planktivory The quantity of repetitive DNA in a genome correlates with genome size115,116, and shifts in repeat content have been linked to adaptations to new environments117–119. For example, bursts of transposable elements (TEs) have been implicated in the diversification of clownfishes through their mutualistic interactions with sea anemones120. However, investigations specifically examining the relationship between repetitive genomic elements and dietary patterns remain lacking. To address this gap, we assessed the association between transposable element (TE) content and trophic guild in acanthuriforms. First, we categorized repeat elements in each genome into ten distinct classes or subfamilies. In the reference genome, repetitive sequences comprised 16.39% of the assembly, while short-read assemblies showed a range of 13.8–21.3% (Supplementary Fig. 24). Second, we performed a phylogenetic ANOVA using phytools71 to test for differences in TE content between trophic guilds. We found no significant association between overall TE content and diet (p = 0.989; Supplementary Fig. 24). Similarly, none of the individual repeat categories showed significant correlations with trophic guild (p = 0.178–0.975; Supplementary Fig. 25). These results indicate that repetitive elements do not systematically vary with dietary transitions to planktivory in acanthuriforms. This pattern aligns with findings from other recent studies, such as in catfishes45 and teleosts more broadly121, which similarly report weak or inconsistent associations between repeat content and ecological or behavioral traits. While TE expansions can occasionally correlate with environmental shifts or significant life history changes, such as mutualisms120, this pattern is not universal. Repeat expansions may occur under relaxed selection or during environmental stress, but they are not a consistent feature of ecological diversification122. In planktivory, dietary transitions appear driven primarily by selection on coding regions, with no detectable parallel changes in TE abundance. Although we cannot rule out the potential role of regulatory evolution, our results suggest that repetitive elements have played a minimal role in the trophic diversification of acanthuriform fishes. Final remarks This study provides a comprehensive view of the evolutionary dynamics of trophic transitions in both fossil and extant acanthuriform fishes by examining their associations with biogeographic patterns, diversification rates, paleoclimatic changes, and genomic signatures. By integrating morphological and genome-wide data, we inferred the most comprehensive tip-dated phylogeny to date for the group, accommodating uncertainties in both topology and divergence time estimates for downstream analyses. We found that planktivory evolved independently at least seven times, including four early Cenozoic transitions in now-extinct lineages. Notably, our biogeographic reconstructions reveal that early transitions occurred in the ancient Tethy Sea, while more recent ones are concentrated in the reef-rich Indo-Pacific, highlighting the role of spatiotemporal ecological opportunity in shaping trophic evolution. Similar net diversification rates between non-planktivores and planktivores, coupled with the fact that transition rates to planktivory are lower than reversals, suggest that a planktivorous lifestyle is not an evolutionary dead end in this group. Our findings also suggest that dietary shifts reflect both paleoclimatic fluctuations and lineage-specific histories, with temporal patterns aligning planktivore diversification with cooler intervals and benthic herbivore diversification with warmer periods, while models also support strong phylogenetic influences. Our genomic analyses reveal that repeated transitions to planktivory are accompanied not only by convergent adaptations but also by shared molecular signatures of positive selection and accelerated evolution, particularly in genes related to metabolism, digestion, and nutrient uptake. While both site convergence and parallelism were observed, the molecular pathways involved appear shaped by a complex interplay of evolutionary history, lineage-specific context, and time (lineage age). These findings emphasize that while convergent ecological pressures may prompt repeatable genomic responses, the molecular routes taken remain partly lineage-specific. Despite providing insights into the ecological and genomic basis of planktivory transitions in reef fishes, our PhyloG2P analyses have limitations. Notably, our focus on protein-coding genes overlooks regulatory elements or noncoding regions involved in dietary adaptation, and we did not identify any PSGs shared across all planktivory transitions. Future research can leverage the annotated reference genome generated here, alongside additional chromosome-level assemblies, to explore genomic features (e.g., rate acceleration in regulatory regions; genome expansions and contractions) that may contribute to trophic specializations, including genes like taste receptor41. Integrating skull shape data123–125 may also further reveal ecomorphological patterns and help link the genotype to the phenotype. Together, our findings enhance understanding of the ecological and genetic factors shaping dietary evolution in acanthuriforms and marine fishes more broadly. They offer a framework for investigating how convergent ecological pressures interact with genomic features to influence diversification across environmental gradients. More broadly, they highlight the utility of combining high-resolution phylogenomics with phenotype-aware selection models to illuminate the genetic signatures of dietary specialization and adaptive evolution across the Tree of Life. Method DNA extractions, exon capture, sequencing, and assembly We generated new exon capture data from tissue samples extracted from museum voucher specimens for a total of 57 acanthuriforms (Acanthurus tractus duplicated) and 9 outgroups from closely related families (Supplementary Data 1). High-quality DNA extractions were sent to Arbor Biosciences for library preparation and exon capture to target the 1105 single-copy exons developed for the FishLife project126 using the Eupercaria-specific probe set55, that also includes PCR-based 29 legacy markers (mtDNA and nuclear genes, see Supplementary Table 1) commonly used for fish phylogenetics127–129. Enriched libraries were sequenced using one lane of the Illumina HiSeq 4000 platform with paired-end 100 bp at the University of Chicago Genomics facility. The raw sequence data was assembled and aligned using the bioinformatic pipeline developed by Hughes et al.55 and available at https://github.com/lilychughes/FishLifeExonCapture/. The final step of the pipeline generated alignments in their correct reading frames for each exon using MACSE v.2.03130. We then removed any instance of single-taxon insertions, trimmed gappy edges, and filtered short sequences using the AlignmentCleanerCodons.py script from the GitHub link above. For species verification, we extracted COI and CYTB sequences for each sample and blasted them against the Barcode of Life Database and National Center for Biotechnology Information (NCBI) repositories using the “bold_identification” python script131. We also visually inspected all the alignments to adjust the reading frames, remove poor-quality reads, and correct misaligned sections in Geneious Prime v.2024.0132. Alignment summary statistics, such as the percentage of missing data, GC content, proportion of variable sites, and alignment length was assessed using the Python package AMAS133. After these quality control steps, we removed 9 low-quality markers and 98 markers with more than 50% missing data (present in fewer than 32 species). The final reduced molecular matrix assembled comprised 998 genes for 57 ingroup taxa (56 species out of the 86 extant acanthuriforms), representing all genera and families, along with 9 outgroup species as outlined above. Taxonomic sampling augmentation To increase the number of species, we generated an expanded matrix combining the 66 newly sequenced taxa (reduced matrix, Supplementary Data 1) with 59 species obtained from GenBank (Appendix 1). We individually aligned each of the 16 legacy markers downloaded from GenBank using MACSE and then merged them with the corresponding legacy marker alignments from our reduced matrix (Supplementary Table 1). After manual inspection to ensure correct reading frames, we retained 22 of 29 legacy markers: 20 from our matrix and 9 from GenBank, with 7 shared between both sources (Supplementary Table 1). We combined these 22 markers with the reduced matrix, placing mitochondrial markers at the end to generate a concatenated dataset of 1002 markers for 148 individuals. This total reflects the combination of our focal taxa and all available GenBank sequences, prior to duplicate removal. We subsequently performed an initial quality control assessment of phylogenetic placement using FastTree-2134 to identify potential contamination or misidentification. We then removed 58 redundant tips representing duplicate species across datasets. Following phylogenomic inference, Prionurus punctatus was renamed P. laticlavius based on its recognition as a junior subjective synonym135 and the duplicated P. laticlavius specimen containing only legacy markers was pruned from the trees. The placement of Acanthurus tristis could not be reliably inferred due to incomplete data (COI only, 202 bp) and was therefore excluded from all downstream analyses. The final expanded molecular matrix comprises 1002 genes for 80 ingroup species ( ~ 93%) and 9 outgroup species (65 spp. with FishLife exons and 24 spp. with legacy markers only). Phylogenomic inference For each assembled molecular matrix, reduced and expanded, we inferred maximum likelihood (ML) trees and multispecies coalescent species trees. First, we determined the best-fitting partition scheme for each matrix using PartitionFinder2136 based on a priori by-codon partitions for each protein-coding marker, and two partitions for each of the ribosomal markers (12S and 16S). We estimated concatenation-based ML trees in RAxML v.8.2.1156 using the best-fit partitioning schemes selected via the Bayesian Information Criterion and the GTRGAMMA model. Using the raxml-ng (extension of RAxML for supercomputers137), we ran 30 independent ML searches and used 100 nonparametric bootstrap replicates to assess edge support. To infer species trees while accounting for incomplete lineage sorting, we initially estimated individual gene trees in RAxML using by-codon partitions. All mtDNA markers were grouped into a single locus alignment, with by-codon partitions applied specifically for protein-coding genes and two partitions for 12S and 16S, as explained above. After inferring best trees from multiple runs and bootstrap support (BS) values, we collapsed gene tree branches with low BS ( < 33%). We then conducted multispecies coalescent species-tree analyses with multi-locus bootstrapping in ASTRAL-III57 using the collapsed gene trees as input to generate a species tree for each matrix. We also assessed gene concordance factors138 by calculating the percentage of gene trees in the data matrix that support a specific branch in the concatenation-based (RAxML) and multispecies coalescent-based (ASTRAL-III) species trees inferred for both datasets138. Integration of fossils and extant species We newly coded a morphological matrix consisting of 107 characters for 32 fossil and 19 extant acanthuriform species plus 5 extant outgroups (see Supplementary Note 1 for the list of osteological characters and character states). To assess the phylogenetic placement of each species based on morphology, we inferred trees based on parsimony and ML approaches. We estimated the parsimony tree in TNT v.1.558 using a driven-search strategy (sectorial ratchet, tree-fusing methodologies) with default parameters. The ML morphological tree was estimated using the MULTIGAMMA and Mk models with 30 iterations in RAxML. We combined the morphological and the expanded molecular matrices for a total of 112 fossil and extant ingroup species and nine outgroups. We estimated the combined matrix in RAxML using the MULTIGAMMA and Mk models, 100 bootstrap replicates, and six partitions: five for the molecular sequences (one for each codon position of all nuclear and mtDNA protein-coding markers, plus two for 12S and 16S) and one for the morphological dataset. Taxa with polymorphic character states were coded as missing (“?”) for RAxML, which cannot handle polymorphic characters. Total-evidence dating analyses and phylogenetic uncertainty We conducted divergence time estimations under a total-evidence, or tip-dating, framework using the Fossilized Birth Death model in MrBayes v 3.2.7a61. To account for topological uncertainty, we assembled largely independent subsets (randomly subsampled from the expanded matrix with genes only), each with enough number of genes to overcome sampling error. Specifically, we partitioned the complete dataset into 20 subsets (50 loci ×13 + 49 loci ×7), 10 subsets (99 loci ×7 + 100 loci ×3), and 5 subsets (199 loci ×3 + 198 loci ×2). To maintain consistent taxon representation across subsets, all shared nine anchor genes. We then ran ML trees using RAxML-NG for each subset, revealing substantial topological discordance, particularly among trees estimated with smaller numbers of loci (Supplementary Fig. 8). Given this pattern, we retained the five largest subsets (198–199 loci each) for downstream total-evidence dating analyses. Each genomic subset was combined with the morphological matrix (including fossil and extant taxa), yielding a total of 112 taxa per total-evidence analysis. The dataset was divided into five partitions (three codon positions, 12S + 16S, morphology). Molecular partitions were modeled under GTR + G, whereas the morphological partition was analyzed using the MK model. Fossil ages and calibration priors used for divergence time estimation are listed in Supplementary Table 2. To evaluate the effect of root calibration choice, we conducted two null analyses using one representative genomic subset. In the first, we applied a secondary calibration on stem Acanthuriformes (68–75 Ma), with Chaetodon striatus used as the outgroup (Supplementary Table 3). In the second, we omitted this root calibration entirely. These analyses yielded crown Acanthuriformes age estimates of 73.8 Ma (with root calibration) and 80.8 Ma (without root calibration), respectively. Because both estimates were older than those reported in recent large-scale phylogenetic studies (refs. 48,63,139; see Supplementary Table 3), we instead implemented a secondary calibration on crown Acanthuriformes using a uniform prior of 55.8–64 Ma. Speciation and extinction priors were estimated using the “fit.bd” function in phytools71, based on a well-sampled acanthuriform tree140. That tree was rescaled by adjusting its crown age from 77 Ma to 60 Ma to align with previously published age estimates (Supplementary Table 3). After independently estimating phylogenies using morphology alone (Supplementary Fig. 6) and using a dataset combining morphology and molecular information (Supplementary Fig. 7), we constrained the three acanthuriform families (Luvaridae, Zanclidae, and Acanthuridae), including both fossil and extant species, to be monophyletic. We implemented this constraint because the full total-evidence dating runs were computationally intensive and expected to require months to complete, and we sought to avoid recovering paraphyletic families despite strong prior evidence supporting their monophyly. We also observed incongruent placements for two fossils, †Gazolaichthys vestenanovae and †Padovathurus gaudryi. In the combined matrix, both fossils were resolved as sister to Zanclidae + Acanthuridae (Scheme 1, Supplementary Fig. 7), whereas in the morphology-only analysis (Supplementary Fig. 6) and in a previous study52 they were placed as sister to Acanthuridae (Scheme 2). Although two constraints were applied for tip-dating analyses, we favor the placement in Scheme 1 over Scheme 2. This interpretation is consistent with Tyler62, who noted that osteological differences between the families Acanthuridae and Zanclidae are minimal and may not warrant separation at the family level. From an anatomical standpoint, the distinction between these two families appears to reflect historical taxonomic practice more than strong morphological discontinuity. To address this topological ambiguity, we implemented both fossil-placement schemes as alternative constraints in the tip-dating analyses. This resulted in 10 MrBayes analyses (two schemes by five genomic subsets). Each analysis was run with eight independent runs and four Monte Carlo Markov chains (MCMCs) for over 350 million generations each, sampling every 10,000 generations. We used a sample probability of 0.94 and a relaxed clock model with the clock rate prior following a log normal distribution and independent gamma rate. The first 10% of trees sampled were discarded as relative burn-in and convergence of the MCMC was verified using the ESS criterion for each parameter in TRACER v.1.7141. After more than 14 months of total runtime, we found that 9 of the 10 analyses reached convergence where ESS values were close to or above 200 (Supplementary Table 4). Because these analyses ran for over a year and Subset 2 based on the Scheme 2 had not reached satisfactory ESS values, we removed it from all downstream analyses. We sampled ~2000 trees for Scheme 1 and ~2500 trees for Scheme 2, evenly distributed along the PD post-burn-in from each subset to have a total of 10,000 trees. For each scheme independently, we inferred a Maximum Clade Credibility tree (MCC tree) in TreeAnnotator v.2.7.5142. To obtain a PD of trees for phylogenetic comparative analyses, depending on the type of analysis and their computational time, we sampled either 100 or 4 trees from each subset, resulting in a total of 500 and 20 trees for each scheme. Diet and biogeographic data We compiled a discrete diet database for extant and fossil taxa (i.e., non-planktivore vs. planktivore). While extant species were coded based on the diet composition from the literature12 and FishBase67, fossil diet is typically determined based on their tooth morphology, paleoecology, and paleoenvironment68,69. However, for acanthuriforms, this task is particularly daunting given the complex tooth morphology in the group. We thus examined if there is a correlation between tooth morphology and diet for extant species only by codifying three types of teeth: conical, multi-denticulate, and brush-like. We then conducted a phylogenetic regression to determine their relationship, finding no significant evolutionary correlation (p = 0.842; Supplementary Fig. 11). We find cases where both non-planktivorous and planktivorous fishes displayed multi-denticulate or conical tooth morphologies. This result indicates that factors other than tooth morphology should be used to determine the diet of acanthuriform fossils. Therefore, we focused on the paleoenvironment of each fossil, categorizing fossils derived from oceanic sediments as planktivores and from limestones as non-planktivores49,70. Two fossils (†Gazolaichthys vestenanovae and †Padovathurus gaudryi) were coded as ambiguous. We also gathered geographic distribution data and built a presence/absence matrix by coding each extant and fossil species according to their geographic ranges, primarily based on the IUCN Red List143, Ocean Biogeographic Information System144, and Paleobiology database (https://paleobiodb.org/), as well as from the primary literature. We used a seven-region biogeographic scheme (based on refs. 145,146): WIO, CIP, CP, TEP, WA, EA, and Tet. Ancestral diet and ancestral range estimates To examine the frequency of shifts to planktivory and their timing, we conducted ancestral diet reconstruction analyses using “make.simmap” function in the R package phytools71. We fixed the root state to non-planktivory, considering that close acanthuriform relatives primarily feed on algae and benthic invertebrates, such as Chaetodontidae (butterflyfishes147) and Pomacanthidae (angelfishes78). We first assessed whether these trophic transitions fit into a model where shifts between non-planktivory (state 0) and planktivory (state 1), and vice versa, exhibit equal rates (ER), or if instead these transitions are different (ARD71). We then conducted stochastic character mapping using the best-fit model on the MCC tree and across 20 trees of each scheme. Transitions were counted when a nodal pie is >50% of state 0 and one of its descendant branches is >50% for state 1 and vice versa. Exceptions were made following the parsimony principle when a nodal pie with one state is between two nodes with the alternative state, resulting in fewer transition counts. For comparison purposes only, we also ran the analysis excluding all the fossils for MCC tree from Scheme 1. We also investigated where trophic transitions occurred. We conducted ancestral range reconstruction analyses in the R package BioGeoBEARS72 following the approach outlined in ref. 77. We evaluated 12 biogeographic models combining DEC148, DIVA149, and BayAREA150, with and without the jump-dispersal or founder-speciation event (j)151 and the dispersal matrix power exponential (w) parameters152. We analyzed each model using three time slices based on two major geological events: (i) prior to the closure of the Tethys Seaway (65–16 or 12 Ma), (ii) after the closure of the Tethys Seaway and prior to the last rising of the Panama Isthmus (16 or 12–2.8 Ma), and (iii) after the last rising of the Panama Isthmus (2.8–0 Ma153). Due to the ambiguous age of the total closure of the Tethys Seaway, the terminal Tethyan event, we ran separate biogeographic analyses assuming 1673 and 12 Ma74–76 as the final closure. We also accounted for connectivity between areas by implementing three different dispersal probability categories: 1 (high connectivity), 0.5 (intermediate separation), and 0.0001 (wide separation/no connectivity). We ran these analyses for both schemes using the MCC trees with fossils as input phylogenies. We then selected the best-fit biogeographic model based on the Akaike Information Criterion scores corrected for small sample size (Supplementary Table 5). The best-fit model was implemented in 20 time-trees, and the results were summarized by overlying average probabilities across compatible nodes on each MCC tree154. Diet-dependent diversification We estimated diversification and transition rates between non-planktivorous and planktivorous diets by conducting SSE analyses in HiSSE82. Because SSE analyses cannot handle non-ultrametric trees, for these analyses, we used extant-only trees (but see below). We fitted a total of five SSE models with and without unknown (“hidden”) states: null BiSSE model (rates independent of the trait), full BiSSE or BiSSE-like HiSSE (rates dependent on the trait), full HiSSE model (rates dependent on the trait with hidden states), CID-2 model (character-independent model with two hidden states), and CID-4 model (character-independent model with four hidden states). We defined sampling fractions for each trophic guild to account for missing taxa; non-planktivores 91% and planktivores 100%. As with ancestral diet reconstruction analyses, we also fixed the root to non-planktivory. We ran this analysis on the MCC tree and across 20 trees of each scheme. To incorporate uncertainties in model choice, we then model-averaged rates for all tips and nodes in the trees calculated under five models using AIC weights. We then used the R package ggplot2155 to generate box plots and assess differences in diversification rates between the two states for all tips and nodes. Finally, we statistically compared tip-associated lineage diversification rates between the trophic guilds conducting a phylogenetic ANOVA in phytools71 with Bonferroni correction. To incorporate fossil data into our diversification analyses, we estimated the number of lineages over the evolutionary history of this group for the two trophic guilds. We generated LTT plots in phytools71 using the 500 trees for each scheme. For comparison purposes only, we also ran this analysis excluding all the fossils for MCC tree from Scheme 1 with 100 simulations. Paleoclimate-dependent diet evolution We compiled paleoclimatic temperature curves spanning the last ~64 Ma of acanthuriforms’ evolutionary history to examine the association of planktivory with two temperature datasets: the global average ocean temperatures obtained from oxygen isotope data and the tropical ocean temperatures based on sea surface temperatures from tropical latitudes85. We assessed the association of the global and tropical paleoclimatic temperature curves with the evolution of planktivory using a modified time-dependent regression threshold model recently developed by Melendez-Vázquez et al.87. We fitted a total of four different evolutionary models for discrete data separately for both temperature curves: three climate-independent models, Brownian motion, early burst, and phylogenetic signal or lambda, and a climate-dependent model. We used 100 integrations (N = 100) and a climatic spline interpolation function based on 500 degrees of freedom (df = 500). For each curve, we fitted all these models using both the MCC tree and the 500 trees of each scheme. Sequencing and assembly of a chromosome-level genome for Acanthurus chirurgus We generated a phased chromosome-level genome for the Doctorfish tang A. chirurgus. We obtained flash-frozen muscle tissues from an individual caught using a hand net in the Florida Keys (24°59.564 N, 80°25.753 W), US, by Phillipp Rauch on the 29th of January of 2023. This specimen was collected under a valid Commercial Saltwater Products License with a Marine Life endorsement (MLD-299) issued by the Florida Fish and Wildlife Conservation Commission. The collection was conducted in accordance with applicable state regulations governing marine life harvest. The voucher specimen is deposited at Scripps Institution of Oceanography (SIO), collection number SIO 24-10. Subsequently, we outsourced the DNA and RNA extractions, library preparations, sequencing, assembly, and annotation to Cantata Bio LLC. High molecular weight DNA extraction using the Qiagen Blood and Cell Culture DNA Kit following the manufacturer’s protocol. DNA samples were quantified using Qubit 2.0 Fluorometer (Life Technologies, Carlsbad, CA, USA). The PacBio SMRTbell library ( ~ 20 kb) for PacBio Sequel was constructed using SMRTbell Express Template Prep Kit 2.0 (PacBio, Menlo Park, CA, USA) and the manufacturer recommended protocol. The library was bound to polymerase using the Sequel II Binding Kit 2.0 (PacBio) and loaded onto PacBio Sequel II. Sequencing was performed on PacBio Sequel II 8 M SMRT cells. PacBio CCS reads were assembled into scaffolds using Hifiasm v.0.15.4-r347156 with default parameters. Blast results of the Hifiasm output assembly against the nt database were used as input for blobtools2 v.1.1.1157 and scaffolds identified as possible contamination were removed from the assembly. Finally, purge_dups3 v.1.2.5158 was used to remove haplotigs and contig overlaps, yielding one de novo assembly for each haplotype. To improve genome architecture through scaffolding and read orientation, Dovetail Omni-C library sequencing was performed on an Illumina HiSeqX platform at ~30× sequence coverage159. Each Dovetail Omni-C library, chromatin was fixed in place with formaldehyde in the nucleus. Fixed chromatin was digested with and then extracted, chromatin ends were repaired and ligated to a biotinylated bridge adapter, followed by proximity ligation of adapter-containing ends. After proximity ligation, crosslinks were, the DNA purified. Purified DNA was treated to remove biotin that was not internal to ligated fragments. Sequencing libraries were generated using enzymes and Illumina-compatible adapters. Biotin-containing fragments were isolated using streptavidin beads before PCR enrichment of each library. The de novo assembly and Dovetail reads were then assembled with Omni-C HiRise, a software pipeline designed specifically for using proximity ligation data to scaffold genome assemblies160. Dovetail sequences were aligned to the draft input assembly using bwa (https://github.com/lh3/bwa). The separations of Dovetail read pairs mapped within draft scaffolds were analyzed by produce a likelihood model for genomic distance between read pairs, and the model was used to identify and break putative, to score prospective joins, and make joins above a threshold. Transcriptomic data was obtained by extracting RNA, preparing libraries, and sequencing. Total RNA extraction was done using the QIAGEN RNeasy Plus Kit following manufacturer protocols. Total RNA was quantified using Qubit RNA Assay and TapeStation 4200. Prior to library prep, we performed DNase treatment followed by AMPure bead clean up and QIAGEN FastSelect HMR rRNA depletion. Library preparation was done with the NEBNext Ultra II RNA Library Prep Kit following manufacturer protocols. Then these libraries were run on the Illumina NovaSeq 6000 platform in 2 × 150 bp configuration. Genome completeness was assessed using Benchmarking Universal Single-Copy Orthologs (BUSCO) v.6.0.099,100 based on the single-copy orthologs for ray-finned fishes database (actinopterygii_odb12). Scaffolds were assigned chromosomal numbers based on their length as no other surgeonfish genome was available at the time of assembly. Contiguity statistics of scaffolded assembly per haplotype was computed using Quast161. For the genome annotation, first repeat families found in the genome assembly of A. chirurgus were identified de novo and classified using the software package RepeatModeler v.2.0.1162. The custom repeat library obtained from RepeatModeler were used to discover, identify, and mask the repeats in the assembly file using RepeatMasker v.4.1.0163. Coding sequences from Acanthochromis polyacanthus (GCF_021347895.1), Chaetodon austriacus (draft genome164), and Chelmon rostratus (GCF_017976325.1) were used to train the initial ab initio model for A. chirurgus using AUGUSTUS v.2.5.5165, with six rounds of optimization. The same coding sequences were also used to train a separate ab initio model for A. chirurgus with SNAP (v.2006-07-28; https://github.com/KorfLab/SNAP). RNAseq reads were mapped onto the genome using the STAR v.2.7 aligner software166 and intron hints generated with the bam2hints tools in AUGUSTUS. Gene prediction in the repeat-masked genome was performed using MAKER2167, SNAP, and AUGUSTUS (with intron-exon boundary hints provided from RNA-Seq). To guide predictions, Swiss-Prot peptide sequences from the UniProt database168 were downloaded and used in conjunction with the protein sequences from A. polyacanthus, C. austriacus, and C. rostratus to generate peptide evidence in the MAKER2 pipeline. Only genes that were predicted by both SNAP and AUGUSTUS software were retained in the final gene sets. To help assess the quality of the gene prediction, AED scores were generated for each of the predicted genes as part of the MAKER2 pipeline. Genes were further characterized for their putative function by performing a BLAST search of the peptide sequences against the UniProt database. tRNAs were predicted using the software tRNAscan-SE v.2.05169. Sequencing and assembly of short-read genomes Using the libraries prepared at Arbor Biosciences, we sequenced genomes for a total of 48 acanthuriform species (average genome size 0.8 Gbp). We sequenced short reads aiming for 30× coverage using the Illumina NovaSeq 6000 S4 PE150 platform at the Oklahoma Medical Research Foundation sequencing facility. We pre-assembled the raw data in MaSuRCA v.4.0.8101 and removed three species due to poor sequencing quality. We then enhanced scaffolding of each genome (Supplementary Data 2) by using as a reference the chromosome-level genome for A. chirurgus in RagTag v.2.1.0102 to reorder, reorient, and fill gaps. Because A. chirurgus was represented by both a reference genome and a short-read genome, the latter was excluded from downstream analyses. Finally, we assessed genome completeness for the 44 species in BUSCO using the same procedure as for chromosome-level genome. This analysis also identified single-copy and duplicated genes within each species based on the ray-finned fishes database. Identification of orthologous sequences for positive selection analyses We first identified single-copy genes among the 23,548 protein-coding genes in the reference genome, along with three reference genomes from closely related species available in NCBI: Larimichthys crocea (Sciaenidae), Epinephelus lanceolatus (Serranidae), and Chelmo rostratus (Chaeotodontidae). We used OrthoFinder170 with protein sequences as input (translated with SeqKit v2.8.2171), identifying 6381 single-copy orthologs among these four species. We then mined these genes from the 44 short-read data using HMMER v.3.1172. All 6381 mined genes were aligned using MACSE130, followed by cleaning and filtering steps as described for the exon capture data. Because positive selection analyses are highly sensitive to alignment errors173, we performed exhaustive quality control. Specifically, we trimmed alignments in MACSE, allowing a maximum of 20% gaps per site, and excluded genes containing fewer than 70% of species ( ≤ 31 taxa) or shorter than 200 bp. Alignments were visually inspected in Geneious Prime v.2024.0132, to adjust reading frames, remove poor-quality reads, and correct misalignments. After filtering, we translated the remaining 4940 gene alignments into protein sequences and ran them in OrthoFinder to confirm that they still represented one-to-one orthologs. Using the pruned MCC tree from Scheme 1 to match the species with available genome data, we ultimately retained a set of 2542 single-copy orthologous genes for positive selection analyses. In addition to these mined genes, we also identified single-copy orthologs from the BUSCO Actinopterygii database (odb10) and AUGUSTUS as gene predictor. OrthoFinder identified 1269 BUSCO orthologs, of which 1250 passed quality control filters explained above. Following the initial round of peer review, two referees recommended updating our ortholog search to incorporate the most recent Actinopterygii BUSCO database (odb12). We therefore re-ran the pipeline using this updated version and recovered 1666 single-copy orthologs. To evaluate their correspondence with the previously compiled gene set (derived from reference genome and BUSCO odb10, see Supplementary Data 3), we compared not only gene identifiers, which differ between database versions, but also sequence similarity using Geneious Prime. Of the 1666 orthologs, 1532 were duplicated from our earlier set. We added the remaining 134 newly recovered loci to the BUSCO dataset, resulting in a total of 1384 loci for positive selection analyses. Analyses of positive selection associated with trophic shifts To identify instances of positive selection associated with planktivory, we apply alternative approaches implemented in HyPhy103 that use the dN/dS metric (the ratio of non-synonymous substitutions [dN] to synonymous substitutions [dS]). We interrogated a total of 3926 single-copy genes, including 2542 mined from the reference genome’s protein-coding genes and 1384 from the BUSCO databases. The MCC tree from Scheme 1 was pruned to retain only the species present in each gene alignment. We focused on convergently evolved PSGs by testing for positive selection exclusively along branches corresponding to transitions to planktivory, as inferred from ancestral diet reconstruction analyses (Fig. 3A). We first applied the adaptive Branch-Site Random Effects Likelihood method (aBSREL104). Although a full data-driven run across all branches is standard, it is computationally intensive. Thus, we adopted a hypothesis-driven approach by testing only the five foreground branches corresponding to dietary shifts (see Fig. 3A), greatly reducing computational burden. This approach allowed detection of genes under positive selection associated with the transition to planktivory (Scenario 1), selection acting after the transition (Scenario 2), and convergence across planktivore lineages. aBSREL is restricted to focal branches and cannot detect positive selection in non-focal branches; we therefore complemented these analyses with BUSTED-PH. BUSTED-PH distinguishes positive selection between foreground and background lineages by explicitly modeling phenotype-associated selection pressures. In this analysis, stem and crown planktivore lineages were designated as foreground branches, non-planktivores as background, and reversals to a non-planktivore diet as nuisance branches, therefore excluded from inference (Fig. 4B and Supplementary Fig. 23). After correcting for multiple testing (FDR p < 0.05) in both aBSREL and BUSTED-PH, we initially identified 1461 PSGs (382 from the genome dataset and 1079 from BUSCO). Of these, 993 genes (68 genome-derived and 925 BUSCO-derived) also showed evidence of selection in background branches and were removed. Visual inspection in Geneious Prime of the remaining 467 genes (314 genome + 154 BUSCO) confirmed quality for 160 genes without changes, while 307 required manual editing (e.g., trimming gappy regions). One gene was excluded due to a possible instance of out-paralogy, and Ctenochaetus striatus was removed from all alignments due to sequencing error. We then reran aBSREL and BUSTED-PH on the remaining 466 genes, repeating the same filtering pipeline. We examined the overlap between both approaches and removed six duplicated genes from the dataset. This resulted in a final curated set of 302 PSGs (197 genome + 105 BUSCO). Of these, 39 PSGs (15 genome + 24 BUSCO) were identified in at least two of the five planktivory transitions. We visualized shared convergence across lineages using an UpSet plot generated with UpSetR174. GO terms and functional categories were retrieved for each PSG using the PANTHER Classification System (https://www.pantherdb.org106) and the ZFIN database107 to determine gene function and expression profiles. We further tested whether the 39 convergently evolved PSGs exhibited accelerated or decelerated rates of evolution in planktivore lineages using the RER method in HyPhy, implemented via the RERconverge R package175. This approach identifies foreground branches with significantly different evolutionary rates relative to the background. As input, we used amino acid alignments for each gene and an unconstrained phylogram generated from all exons using RAxML. The unconstrained phylogram was pruned to match the species present in each gene alignment. To investigate site-specific signals of episodic diversifying selection, we applied MEME113 in HyPhy to the subset of genes showing accelerated rates. This analysis also allowed us to assess whether significant codons (p < 0.1) exhibited patterns of convergence, either via different amino acid changes at the same site or changes at different sites, or parallelism, defined as identical amino acid substitutions at the same site across the five planktivory transitions. For MEME, we used the same amino acid alignments and the RAxML-based phylogram used in the RER analysis. Variations in transposable elements linked to trophic shifts First, we used Tandem Repeats Finder for the identification of tandem repeats176. We then conducted both homology-based and de novo methods to detect TEs. The homology-based TE identification for all species was carried out using RepeatMasker and based on the Repbase library177. Additionally, the de novo TE annotation was conducted using RepeatModeler with default settings to create a de novo repeat library for each assembled genome. Subsequently, we employed the de novo repeat library in conjunction with RepeatMasker to predict soft-masked repeats for each species. The identified repeat elements were categorized into ten distinct class/subfamilies, encompassing long interspersed nuclear elements, short interspersed nuclear elements, long terminal repeats, transposons, rolling circle elements, small RNAs, satellites, simple repeats, low complexity repeats, and unclassified elements. To evaluate the correlation between the quantity of TEs (total and for each class/subfamily) and trophic guilds, we conducted a phylogenetic ANOVA in phytools71 using Bonferroni correction. Reporting summary Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article. Supplementary information Supplementary information Descriptions of Additional Supplementary Files Supplementary Data 1 Supplementary Data 2 Supplementary Data 3 Supplementary Data 4 Supplementary Data 5 Reporting Summary Transparent Peer Review file Source data Source Data Acknowledgements We are thankful to the Smithsonian Tropical Research Institute, to A. Graham and J. Pogonoski (CSIRO Australian National Fish Collection, Hobart); and to C. Huddleston and D. Pitassy (Smithsonian Institution National Museum of Natural History) for providing tissue samples, assisting with sample preparation, and shipping the tissue samples. We also thank R. Peterson and V. Rodriguez (The George Washington University) for conducting DNA extractions and coordinating their shipment. We thank M. McGee (Monash University) for helping with the preassembly of the short-read genomes. We thank J. Gallego-Garcia for guiding us in uploading the chromosome-level genome to NCBI. We also thank A. Selberg (Temple University) for valuable assistance in interpreting and visualizing the RER results. Finally, bioinformatic analyses were facilitated by the University of Oklahoma Supercomputing Center for Education & Research (OSCER). Author contributions Writing–original draft: A.S. and R.B.-R. Conceptualization: A.S., R.B.-R., D.A., J.C.T., and G.C. Investigation: A.S., R.B.-R., D.A., J.C.T., and G.C. Writing–review and editing: A.S., R.B.-R., D.A., J.C.T., G.C., G.O., F.M.-V., M.R.-S., M.W.W., W.W., and C.C.B. Methodology: A.S., R.B.-R., D.A., J.C.T., G.C., F.M.-V., M.R.-S., G.O., and M.W.W. Resources: A.S., R.B.-R., D.A., J.C.T., G.C., G.O., M.W.W., W.W., and C.C.B. Funding acquisition: A.S., D.A., R.B.-R., C.C.B., M.W.W., and G.O. Data curation: A.S., R.B.-R., D.A., J.C.T., G.C., G.O., M.W.W., C.C.B., and W.W. Validation: A.S., R.B.-R., and D.A. Supervision: A.S., and R.B.-R. Formal analysis: A.S., R.B.-R., D.A., M.R.-S., and F.M.-V. Project administration: A.S. and R.B.-R. Visualization: A.S. and R.B.-R. Peer review Peer review information Nature Communications thanks Remy Gatins who co-reviewed with Annabel Hughes and the other anonymous, reviewer(s) for their contribution to the peer review of this work. A peer review file is available. Funding This project was supported by National Science Foundation (NSF) grants DEB-1932759 and DEB-1929248, DEB-2225130 to R.B.-R., DEB-2015404 and DEB-2144325 to D.A., DEB-1457426 and DEB-1541554 to G.O., DEB-1541552 to C.C.B., and DEB-1541547 to M.W.W. A.S. was partially supported by the Bullard Dissertation Completion Award from the University of Oklahoma, USA. Additional support to R.B.-R. was provided by the UC San Diego Marine Sciences Research Award (internal grant RG114123). Data availability The genomic data generated in this study have been deposited in the NCBI database under the umbrella BioProject PRJNA1433022. Raw sequence reads for exon-capture and short-reads are under the BioProject PRJNA1418135. Accession numbers for exon-capture data and short-read genomes are provided in Supplementary Data 1 and 2, respectively. Museum collection numbers and the collected geographic locations for all sampled specimens can also be found in Supplementary Data 1 and 2. The voucher specimen of Acanthurus chirurgus used to generate the chromosome-level genome in this study is deposited at Scripps Institution of Oceanography (SIO), under collection number SIO 24-10. PacBio and Hi-C raw reads for the A. chirurgus genome are also available under BioProject PRJNA1418135. The phased haplotypes assemblies are deposited under BioProjects PRJNA1430673 (principal haplotype, accession number JBVQTB000000000) and PRJNA1430672 (alternate haplotype, accession number JBVQTC000000000). All data generated and analyzed during this study are available on Figshare (10.6084/m9.figshare.29594255)178, including exon and single-copy ortholog alignments, phylogenetic trees, input files for comparative analyses, and chromosome-level genome annotation files. Source data for the main figures are provided as a Source Data file and on Figshare. Source data are provided with this paper. Code availability All scripts used in this study are available on Figshare (10.6084/m9.figshare.29594255)178. Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Supplementary information The online version contains supplementary material available at 10.1038/s41467-026-73110-3. References Associated Data Supplementary Materials Supplementary information Descriptions of Additional Supplementary Files Supplementary Data 1 Supplementary Data 2 Supplementary Data 3 Supplementary Data 4 Supplementary Data 5 Reporting Summary Transparent Peer Review file Source Data Data Availability Statement The genomic data generated in this study have been deposited in the NCBI database under the umbrella BioProject PRJNA1433022. Raw sequence reads for exon-capture and short-reads are under the BioProject PRJNA1418135. Accession numbers for exon-capture data and short-read genomes are provided in Supplementary Data 1 and 2, respectively. Museum collection numbers and the collected geographic locations for all sampled specimens can also be found in Supplementary Data 1 and 2. The voucher specimen of Acanthurus chirurgus used to generate the chromosome-level genome in this study is deposited at Scripps Institution of Oceanography (SIO), under collection number SIO 24-10. PacBio and Hi-C raw reads for the A. chirurgus genome are also available under BioProject PRJNA1418135. The phased haplotypes assemblies are deposited under BioProjects PRJNA1430673 (principal haplotype, accession number JBVQTB000000000) and PRJNA1430672 (alternate haplotype, accession number JBVQTC000000000). All data generated and analyzed during this study are available on Figshare (10.6084/m9.figshare.29594255)178, including exon and single-copy ortholog alignments, phylogenetic trees, input files for comparative analyses, and chromosome-level genome annotation files. Source data for the main figures are provided as a Source Data file and on Figshare. Source data are provided with this paper. All scripts used in this study are available on Figshare (10.6084/m9.figshare.29594255)178.