Gene loss and vesicular transport remodeling underpin heterotrophic adaptations of scleractinian corals

Abstract

Background

Azooxanthellate scleractinian corals, which lack symbiotic dinoflagellates, occur from shallow waters to deep-sea environments. In deep-sea benthic ecosystems, they serve as foundation species and are considered highly vulnerable to environmental disturbances. However, their genomic diversities and the genetic basis of their fully heterotrophic lifestyles remain poorly understood.

Results

In this study, the genome of a deep-sea solitary coral, Polymyces sp., was sequenced. Integrated with the published genomes of the deep-sea colonial coral Desmophyllum pertusum and two shallow-water azooxanthellate corals Dendrophyllia cribrosa and Cladopsammia gracilis, the adaptive mechanisms of fully heterotrophic lifestyles were investigated. For genomic diversity comparisons, Illumina short-read data for deep-sea (Polymyces sp., D. pertusum, and Madrepora oculata) and zooxanthellate corals (Porites australiensis) were generated. The results indicated that reduced genomic diversity was found in azooxanthellate scleractinians compared with zooxanthellate relatives. Signatures of recent inbreeding were detected in Polymyces sp. and D. cribrosa, potentially contributing to their low recent effective population sizes and the endangered status of D. cribrosa. Lineage-specific gene losses affecting circadian rhythm, immunity, and autophagy were also detected in the above two species, suggesting overly streamlined energy-saving trade-offs that may impair functional flexibility. Interestingly, the cosmopolitan deep-sea D. pertusum retained complete light-sensing and circadian clock complements, whereas deep-sea Polymyces sp. with evidence of recent inbreeding showed substantial reductions, suggesting that photosensitivity and biological clocks may remain important for ecological success even in the deep-sea environment. Convergent adaptations were supported by concordant patterns of gene-family contraction and positive selection shared among the four azooxanthellate scleractinians. The contractions were concentrated in functions related to photosensitivity, lipid metabolism, and mitochondrial processes, consistent with life in dark and oligotrophic habitats, whereas significant signatures of positive selection were detected on vesicle transport-related genes, highlighting the key roles of vesicle-mediated endosomal pathways in a photosymbiosis-free lifestyle.

Conclusions

This study provides a genomic perspective on the evolution of heterotrophic strategies in scleractinian corals and highlights conservation concerns for nonsymbiotic corals under accelerating global stressors.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12915-026-02582-y.

Background

The frameworks built by corals are renowned as the largest reservoirs of biodiversity on earth. They function as complex living systems that shelter, feed, and sustain an incredible variety of marine organisms [1]. Symbiosis with photosynthetic dinoflagellates in scleractinians is thought to confer a competitive advantage over other benthic invertebrates. Through efficient nutrient and energy exchange, this symbiotic relationship promotes ecological success in tropical areas [2]. However, the lifestyles of scleractinians are diverse. Approximately half of the extant scleractinian species are azooxanthellates [3], which survive in a completely heterotrophic mode without the benefits of photosymbiosis. These corals are mostly absent from tropical habitats but achieve cosmopolitan distributions in relatively dark areas [4]. Some even exhibit bathymetric records extending to depths greater than 6000 m in the deep sea [5]. From the perspective of evolution, the most recent common ancestor of Scleractinia lacked photosymbionts and emerged at approximately 460 Ma [6]. Photosymbiotic association with Symbiodiniaceae arose later at approximately 300 Ma and was subsequently gained independently multiple times across scleractinian lineages. Once established, these symbiotic relationships were seldom lost, with only a small number of lineages (e.g., some species in Thalamophyllia, Dactylotrochus, Madracis, Phyllangia, Paracyathus, Cladocora, and Astrangia) showing putative secondary losses [6].

Zooxanthellate scleractinians are mixotrophic. They acquire nutrients by transferring photosynthates from endosymbiotic microalgae or by capturing prey and particles [7]. The molecular mechanisms underlying the establishment and persistence of coral–dinoflagellate symbioses have been investigated in many studies. The main aspects include the recognition and phagocytosis of symbiotic dinoflagellates, the exchange and trafficking of metabolic nutrients between the host and symbionts, and biomineralization processes coupled with photosymbiosis [8]. For extant azooxanthellate scleractinians, many of which were potentially derived from lineages that never acquired photosymbionts [6], the genetic components involved in recognizing, establishing, and managing photosymbiosis may not be gained or maintained under strong selective constraints. Consequently, these genes can exhibit signatures of relaxed selection and even gene repertoire reduction or losses. However, the evolutionary trajectories and corresponding consequences of these genomic changes remain poorly resolved in scleractinians.

Global change represents the dominant global threat to the persistence of stony corals. Approximately 30% of assessed species are categorized as endangered or critically endangered on the International Union for Conservation of Nature (IUCN) Red List. However, whether deep-sea scleractinians suffer from a risk comparable to that of shallow-water species has not been evaluated. On one hand, because of greater human contact and increased environmental variation, shallow-water stony corals appear to face a greater risk of extinction than their deep-sea relatives [9]. On the other hand, the increased accessibility makes the related data more abundant for shallow-water species. Thus, the extinction risk for deep corals is likely underestimated [10]. A comparison of population dynamics and genomic diversity between shallow-water and deep-sea scleractinians is essential to reveal their differences of vulnerability under global change, which in turn can provide a scientific basis for the preservation of coral biodiversity.

Although azooxanthellate corals occupy a vast expanse of the ocean depths, genomic resources for deep-sea scleractinians remain limited. To date, only a single deep-sea scleractinian genome (Desmophyllum pertusum) has been reported [11]. Moreover, most azooxanthellate scleractinians are solitary [12], whereas all the available azooxanthellate genomes are generated from colonial taxa, leaving a major gap in representative genome resources and constraining comprehensively comparative inferences regarding the adaptation of a fully heterotrophic lifestyle. In this study, a deep-sea stony coral (Polymyces sp.) that exhibited a solitary morphology is sequenced, thereby serving as a necessary reference that expands taxonomic and morphological representations for azooxanthellate scleractinians. At the genomic level, four azooxanthellate scleractinians (including two deep-sea species Polymyces sp. and D. pertusum as well as two shallow-water species Dendrophyllia cribrosa and Cladopsammia gracilis) spanning Robust and Complex clades were compared with their zooxanthellate relatives. Both shared and distinct adaptive strategies among the azooxanthellate corals were revealed. With the analyses of historical demography and genomic diversity integrated, this study offers complementary perspectives on the evolutionary trajectories of scleractinians and informs future conservation strategies.

Results

Genome sequencing, assembly, and annotation

During a 2019 expedition to the western Pacific Ocean, a solitary coral specimen Polymyces sp. (Fig. 1a) was collected at the depth of 970 m from a seamount (10°04.69′ N, 140°12.19′ E) on the Caroline Ridge using a suction sampler mounted on the ROV Faxian deployed from R/V Kexue. High-molecular-weight DNA extracted from this sample was used to generate PacBio HiFi data, yielding 54.68 Gb of HiFi reads from two SMRT cells. After being assembled with Hifiasm 0.20.0 [13], a draft genome of 682.95 Mb was generated, which was at the higher end of the scleractinian genomic size range (234.35–1096.36 Mb) (Additional file 1: Table S1). This assembled genome size closely matched the estimated value of 678.3 Mb resulting from the genome survey (Additional file 2: Fig. S1). Comprising 1,024 contigs, the assembled genome showed a contig N50 of 5.58 Mb and a contig L50 of 36, indicating better continuity than those of other stony corals (Additional file 1: Table S1). Benchmarking universal single-copy ortholog (BUSCO) analysis with the metazoan (odb10) gene set revealed that the genome of Polymyces sp. covered 94.9% of the “core” metazoan genes, indicating that the assembly was as complete as those of other stony corals used for comparative analyses (Additional file 1: Table S1). A total of 25,537 gene models were predicted in the genome of Polymyces sp., of which 14,797 exhibited significant homologies to the validated proteins in the Swiss-Prot database. Additionally, 10,296 gene models were assigned putative functional annotations on the basis of matches identified in the Nr or Pfam domain databases.

Fig. 1.

Deep-sea solitary coral Polymyces sp., evolutionary process of transposable elements and macrosynteny analysis. a A picture of Polymyces sp. freshly after collection; scale bar: 1 cm. b Polymyces sp. exhibited recent bursts of insertion activities for the transposable elements of the LINE, SINE, and DNA groups. c, d Oxford dot plots between scleractinian genomes (x-axis) and the 17 presumed cnidarian ancient linkage groups (ALGs) (y-axis). The scleractinian genomes assembled at the chromosomal level are marked with asterisks

Insertion activities of repeats and genomic macrosynteny

The repeat sequences accounted for 58.75% of the genome, with a length of 401.25 Mb, among which transposable elements (TEs) accounted for 324.98 Mb. Analyses of the TEs indicated that compared with the other three transposon categories, DNA transposons accounted for the greatest number of TEs in Polymyces sp. (Additional file 1: Table S2). Moreover, DNA transposons and SINEs showed significant insertion activity after 10 Ma, and LINEs exhibited a burst of insertion at approximately 25 Ma (Fig. 1b). Such synchronized recent bursts of TEs reflected increased structural plasticity of the genome for Polymyces sp., which could have been caused by successive episodes of unfavorable environmental perturbations during the Neogene (2.58 ~ 23.04 Ma) [14].

Macrosynteny analyses of the chromosome-level genomes revealed that the azooxanthellate stony coral (represented by Dendrophyllia cribrosa with high assembly quality) and the zooxanthellate coral (represented by Catalaphyllia jardinei and Acropora cervicornis) shared the same evolutionary trajectories in terms of the fusion and splitting events with respect to the cnidarian ancestral linkage groups (ALGs). Merging of cnidarian ALGs was observed on chromosomes 1, 8; 4, 13; 6, 16; 9, 14; 10, 12; and 9, 11, and 17. Splitting events occurred in cnidarian ALGs 1, 4, 9, and 14 (Fig. 1c). For the deep-sea colonial coral Desmophyllum pertusum, fusions of chromosomes 8, 13, and 15, 17 were detected, revealing novel combinations of ALGs that had not been shown in shallow-water scleractinian genomes (Fig. 1d, Additional file 2: Fig. S2). These unique chromosomal rearrangements may have been induced by the high pressure of the deep-sea environment [15, 16]. However, for the deep-sea solitary coral Polymyces sp., no such novel fusions were discovered (Fig. 1d), reflecting divergent evolutionary responses of chromosomal architectures for different deep-sea scleractinian species.

Phylogenomic analysis, molecular clock, and mutation rates

A phylogenomic tree was constructed with 644 single-copy orthologous genes based on 152,730 concatenated amino acid sites. The resulting topology recovered the canonical division of scleractinians into Robust and Complex clades, with azooxanthellate scleractinians including both shallow-water and deep-sea lineages. The deep-water colonial coral D. pertusum was in the Robust clade, whereas the deep-sea solitary coral Polymyces sp. was assigned in the Complex clade. The placement of Polymyces sp. in a different major clade from D. pertusum underscores its value as a representative deep-sea reference genome and supports the notion of independent deep-sea colonization events occurring within Scleractinia [17]. The shallow-water azooxanthellate scleractinians Cladopsammia gracilis and Dendrophyllia cribrosa were clustered in a derived subclade that is sister to Polymyces sp. within the Complex clade (Fig. 2a). In light of the recent result from ancestral trait reconstructions based on exon/UCE information (including 274 scleractinian species) [6], the four azooxanthellate taxa analyzed here are inferred to fall within lineages that never acquired photosymbionts [6].

Fig. 2.

Phylogenomics and demographic history. a Molecular clock inferred from 644 single-copy orthologs of 14 scleractinian taxa and two corallimorpharian taxa as outgroups; the deep-sea species are colored blue, and the azooxanthellate species are shaded mauve pink. The positive numbers in red adjacent to each taxon indicate the number of expanded gene families, and the negative numbers in blue indicate the number of contracted gene families. The divergence time of each branch is indicated at the node. The four nodes with magenta rose dots represent the placements for time calibration. The node bars indicate 95% confidence intervals of the estimated divergence time. Abbreviations for the x-axis: C, Carboniferous; P, Permian; T, Triassic; J, Jurassic; K, Cretaceous; Pg, Paleogene; and Ng, Neogene. b, c Demographic history of deep-sea and shallow-water scleractinians. The phylogenetic placement of Madrepora oculata based on mitochondrial genomes is shown in Additional file 2: Fig. S3. The time intervals of the Pliocene Reef Gap (PRG, 5.33–7.25 Ma) and last glacial maximum (LGM: 21–18 kiloannum ago (kya)) are shaded in gray

Molecular clock analyses indicated that divergences of deep-sea scleractinians occurred at 176.7 Ma (95% CI: 155.6–217.6 Ma) and 127.5 Ma (95% CI: 70.4–159.2 Ma) for the Robust and Complex clades, respectively (Fig. 2a). Intriguingly, these divergence intervals largely coincided with those estimated for deep-sea actiniarians in our previous study (177.9 Ma [95% CI: 161.3–199.3 Ma] and 110.4 Ma [95% CI: 89.0–135.6 Ma]) [14], suggesting possible concurrent evolutionary responses of Hexacorallians to ancient deep-sea perturbations. Notably, the crown age of Scleractinia (259.9 Ma [95% CI: 234.5–289.8 Ma], Fig. 2a) estimated in this study is younger than the deeper timescales (~ 400 Ma) inferred from the recently published order-wide analyses based on exon/UCE [6]. This difference is expected given the much smaller available genome representatives applied in this study, which may underrepresent early-branching lineages, as well as limit the ability to implement several curated fossil constraints that can be confidently assigned to extant clades. Therefore, the divergence times in this study primarily served as an internally consistent temporal framework for comparisons among the focal lineages, while future genome resources that better capture early-branching lineages (e.g., Micrabaciidae and Gardineriidae) and key calibration taxa (e.g., Caryophyllia) will enable more direct cross-study calibration and improve precision of deep-node age estimates.

On the basis of divergence times, mutation rates were estimated for each species. The deep-sea stony corals exhibit slower evolutionary rates (7.94 × 10⁻¹⁰ substitutions per site per year for Polymyces sp. and 7.10 × 10⁻¹⁰ for Desmophyllum pertusum) than most shallow-water species (8.32 × 10⁻¹⁰ to 1.42 × 10⁻⁹), except Astreopora myriophthalma (7.70 × 10⁻¹⁰) (Additional file 1: Table S3). Such reduced mutation rates in deep-sea scleractinians are consistent with patterns observed in other deep-sea faunas [18], which may be attributed to the relatively stable environment in the deep sea.

Historical demography

The sequencing data used for the demographic dynamics analyses provided at least 54.33 × coverage of the reference genomes (Additional file 1: Table S4). For the deep-sea coral Madrepora oculata, the reference genome of its congener Madrepora sp. (GCA_046129775) was used, with a mapping rate of 97.11%. With respect to the remaining species, the reference genomes were derived from the same species as those used for generating the Illumina short reads, with the mapping rates ranging from 90.06 to 98.94% (Additional file 1: Table S4).

PSMC analyses indicated that deep-sea stony corals shared similar demographic histories during ancient periods (Fig. 2b). Their effective population sizes (Ne) increase steadily across the Pliocene Reef Gap (PRG, 5.33 ~ 7.25 Ma), followed by a decline at approximately 2 Ma. Similar trajectories can be seen in the demographic histories of shallow-water Porites australiensis and Dendrophyllia cribrosa (Fig. 2c). However, for the shallow-water stony corals Catalaphyllia jardinei and Stylophora pistillata, their trajectories are much more dynamic. The Ne of both corals reached a maximum at ~ 5 Ma, earlier than those of the other stony corals (~ 2 Ma). The subsequent valleys of their Ne nearly coincided with periods when the other stony corals reached their peaks, indicating a competition of niches for them at that time. With decreasing Ne in the other stony corals, short recoveries were observed for C. jardinei and S. pistillata (~ 2 Ma), followed by significant decreases.

In terms of more recent periods, the expansion of populations for deep-sea Desmophyllum pertusum and Madrepora oculata at 60–80 kiloannum ago (kya) resulted in their Ne significantly surpassing that of deep-sea solitary coral Polymyces sp. (Fig. 2b), whose Ne was as low as those of D. cribrosa and C. jardinei (Fig. 2c), two shallow-water endangered species listed on the Endangered and Protected Wild Species List of Korea and as vulnerable species on the IUCN Red List, respectively [19, 20]. The Last Glacial Maximum (LGM) has been shown to have decreased Ne for many marine faunas and may have even led to mass extinction [21, 22]; however, it did not appear to influence the Ne of the stony corals in the present study.

Genomic diversity

Genome-wide variations among the scleractinians were evaluated. The results revealed that the typical shallow-water stony corals Porites australiensis and Stylophora pistillata presented the highest genome-wide heterozygosity (approximately 0.01 heterozygous sites/bp). The genomic heterozygosities of the deep-sea species Desmophyllum pertusum and Polymyces sp. were slightly lower (approximately 0.009 heterozygous sites/bp) (Fig. 3a). Although both D. pertusum and Madrepora oculata are common framework-forming species and often co-occur in the deep sea [23], relatively lower heterozygosity was detected in M. oculata (~ 0.0065 heterozygous sites/bp). The overall genomic diversities of the above three deep-sea stony corals were significantly greater than those of the shallow-water endangered species (C. jardinei and D. cribrosa), suggesting relatively secure genetic statuses for these deep-sea populations. Scanning of heterozygosity in sliding windows across whole genomes further confirmed the above results (Fig. 3b).

Fig. 3.

Genomic diversity in the scleractinians. a Comparison of genome-wide heterozygosity in scleractinians. Deep-sea species are colored blue, and typical shallow-water scleractinians are colored black, with endangered species highlighted in red. b Sliding-window heterozygosity of scleractinian genomes. The x-axis does not span the complete genome length because windows containing more than 80% uncallable sites were excluded from the analysis. The results for Stylophora pistillata are available in Additional file 2: Fig. S4. c ROH of the scleractinian genomes. The ROH are binned by the size of the run, with the y-axis showing the amount of sequence contained in ROH of that size

The stony corals with different genomic diversities also showed differences in the distribution patterns of runs of homozygosity (ROH) (Fig. 3a and c), a parameter used here to evaluate inbreeding levels. P. australiensis and S. pistillata are shallow-water common species. They showed the highest genomic heterozygosity and only contained sequences in short ROH (< 1 Mb) with the least lengths (1123 and 4921 kb vs. 24,392–136,482 kb of other species, respectively). In the medium ROH 1–5 Mb category, the endangered species C. jardinei and D. cribrosa presented the greatest lengths (383,601 and 227,325 kb vs. 0–86,292 kb, respectively). Within the long ROH > 5 Mb group, in addition to the endangered D. cribrosa, whose length was 5453 kb, the deep-sea solitary coral Polymyces sp. also contained a long ROH, with a length of 32,469 kb. The detection of the long ROH category in the genomes of Polymyces sp. and D. cribrosa indicated recent inbreeding in both species.

Positive selection and relaxed selection

Positive selection analyses were conducted on four azooxanthellate scleractinians, including the deep-sea framework-building coral Desmophyllum pertusum, the deep-sea solitary coral Polymyces sp., and two shallow-water corals Cladopsammia gracilis and Dendrophyllia cribrosa. A total of 135, 134, 106, and 83 genes showed significant signals of positive selection for the above four species (Additional file 1: Table S5). Further analyses revealed that the proportions of positively selected genes (PSGs) were closely linked to the modulations of photosymbiotic relationships, including the PSGs involved in the processes of immunity, apoptosis, autophagy, vesicular transport, and energy and nutrient metabolism, with the vesicular transport category containing the greatest number of PSGs (Fig. 4a–d). These processes highlight core genetic panels that facilitate the adaptation to an azooxanthellate lifestyle for scleractinians. The endangered species D. cribrosa presented the fewest PSGs associated with the above functional categories, suggesting limited genomic adaptations and potentially underlying its vulnerability and endangered status. Although the extracted RNA qualities could be potentially impaired by the transition of specimens from deep-sea to shallow-water environments during sampling, gene expression analyses indicated that nearly all photosymbiosis-related PSGs were detectably expressed (TPM > 1; Additional file 1: Table S6), suggesting that these PSGs are transcriptionally active and have a functional role involved in the metabolic processes of azooxanthellate scleractinians. Among the PSGs, sorting nexin 14 (SNX14, which is required for autophagosome clearance) and vacuole membrane protein 1 (VMP1, which is required for autophagosome formation) were shared between Polymyces sp. and C. gracilis. The kinesin-like protein KIF28P (which is involved in vesicle transport along microtubules) was shared between C. gracilis and D. cribrosa. Furthermore, common positively selected sites shared between different species were also identified (N422A, D448Q, and E449S in KIF28P; R81L and K85S in VMP1; and E709S in SNX14). These parallel evolutionary sites were located within or near the regions where changes in protein conformation were detected (Fig. 4e, Additional file 2: Figs. S5 and S6), indicating that the functions of these proteins could be altered accordingly.

Fig. 4.

Positively selected genes (PSGs) in four azooxanthellate scleractinians. The PSGs associated with azooxanthellate adaptations were selected and classified according to their functions (a–d). Genes in each category with the strongest evidence of positive selection (highest log of adjusted P values from the LRT tests) are labeled. The PSGs shared between different species are labeled and represented by star symbols. The domain architectures of KIF28P differ between the zooxanthellate and azooxanthellate corals C. gracilis and D. cribrosa (Kinesin_assoc vs. DUF4175). AlphaFold models show altered local secondary structures near the positively selected sites. These structural alternations suggest potential changes in KIF28P interactions. The colors in the 3D structures correspond to the different domains (e)

The azooxanthellate lifestyles may lead to a relaxation of selective constraints on symbiosis-related genes. To investigate this, RELAX selection analyses implemented in HyPhy 2.5.64 [24] were performed. The results revealed that 301–539 genes were under relaxed selection with k < 1 for the four azooxanthellate stony corals (Additional file 1: Tables S7–S10), among which the endangered D. cribrosa showed the fewest genes under relaxed selection. KEGG enrichment of the gene sets revealed that the oxidative phosphorylation pathway (ko00190) was the only significantly enriched pathway shared among the four stony corals (Additional file 2: Fig. S7), indicating that the corals with azooxanthellate lifestyles may have adopted a distinct strategy for energy metabolism. Other shared genes with signals of relaxed selection related to photosymbiosis included those involved in apoptosis, such as JKAMP (shared between C. gracilis and D. cribrosa) and JUN (between D. pertusum and Polymyces sp.), as well as genes associated with vesicle transport, including KIF28P (shared between C. gracilis and D. cribrosa), Cltc (between D. pertusum and D. cribrosa), and Trappc1 (between Polymyces sp. and C. gracilis) (Additional file 1: Tables S7–S10).

Notably, KIF28P exhibited evidence of both positive and relaxed selection, suggesting that this gene may have simultaneously undergone the loss of ancestral functions (relaxed selection) and the acquisition of novel functions (positive selection). This hypothesis is further supported by the results of the Pfam annotation. In azooxanthellate lineages (C. gracilis and D. cribrosa), KIF28P shows a domain replacement, with the canonical Kinesin_assoc domain in zooxanthellates replaced by DUF4175 (Fig. 4e). AlphaFold predictions indicate changes in the corresponding spatial structure. The proximity of positively selected sites to the remodeled domain boundary is consistent with a possible rewiring of KIF28P molecular interactions. However, the functional consequences remain to be experimentally validated.

Gene family expansion and contraction

Variations in gene family size are recognized as major drivers of adaptive evolution [25]. Gene family comparison by CAFE v4.2.1 among the scleractinians revealed that the deep-sea stony corals Desmophyllum pertusum and Polymyces sp. exhibited the greatest number of contracted gene families (Fig. 2a), suggesting that gene family reduction or loss may have facilitated their adaptation to deep-sea environments.

The functional classification of the contracted families highlighted several categories associated with adaptation to the azooxanthellate mode of life (Fig. 5, Additional file 1: Tables S11–S14). Gene families involved in photosensitivity, lipid metabolism, and mitochondrial energy metabolism are contracted across all analyzed azooxanthellate scleractinians, implicating shared genomic changes to adapt to dark and energy-limited environments. Interestingly, GFP-like chromoproteins (GFPLs), a gene family implicated in shielding excessive light exposure on symbiotic corals [26], were completely undetected in the two deep-sea species (Polymyces sp. and Desmophyllum pertusum), with a distinct contrast to the shallow-water corals that harbored GFPL repertoires composed of 3–15 genes (Fig. 5). Within the adaptation-related functional categories, the two species with recent inbreedings exhibited the most extensive family contractions or losses (22 families across eight categories in Polymyces sp. underwent family size reductions or complete family losses; 11 families in six categories in D. cribrosa were completely lost). These included the functional categories related to circadian rhythm regulation, immunity, apoptosis, and autophagy that were exclusively contracted in Polymyces sp. or D. cribrosa, the functional losses of which may bring disadvantages to their existence.

Fig. 5.

Gene family contractions and losses that are potentially involved in azooxanthellate lifestyles. The lost gene families are marked with asterisks. The category names colored in red indicate those shared among the four species. The names of the deep-sea species are colored blue, and those of the shallow-water azooxanthellate species are colored orange

GO enrichment of the expanded gene families indicated that DNA metabolism (GO: 0006259) and DNA integration (GO: 0015074) were the top biological processes shared among the azooxanthellate stony corals (Additional file 2: Fig. S8), reflecting that host-autonomous DNA synthesis and genomic integrity may serve as primary necessities in the absence of symbiont-supplied photosynthates.

Light-sensing and circadian rhythm

Under darkness in the deep sea, whether animals can sense light and regulate light-entrained rhythms has been explored in several deep-sea faunas [27–31]. However, contradictory discoveries have been reported [27, 28, 31]. To explore the scenarios in deep-sea scleractinians, the identities of critical genes involved in photosensitivity and circadian rhythm regulation were validated through phylogenetic analysis (Additional file 2: Figs. S9–S12), given that the gene complements (e.g., opsins, CRYs, and CLOCK) involved with the above processes belong to large and complex gene families.

In this study, the photoreceptive protein opsins were identified in both deep-sea species (Polymyces sp. and Desmophyllum pertusum) (Fig. 6a). The key functional site K296, which is responsible for chromophore (11-cis retinal) binding, was conserved across all the detected opsins, and the majority of the other functional residues were also well preserved (Fig. 6b). Although Polymyces sp. possessed only a single opsin (anthozoan-specific opsin I (ASOI) [31]), its homolog in the stony coral Acropora palmata (Acropsin 3) has been functionally verified to respond to light and activate G protein-based signaling [32], indicating the potential of Polymyces sp. to sense light. The detectable expression of ASO transcripts in both deep-sea species (ASOI in Polymyces sp. and ASOI plus ASOII in D. pertusum) further indicated that these genes remained transcriptionally active and likely functional in the deep-sea environment (Additional file 2: Fig. S13). Notably, despite the presence and expression of ASOI transcripts, Polymyces sp. presented the greatest number of contracted gene families associated with phototransduction and visual cycle pathways (Figs. 5 and 6c), leading to the postulation that the light-sensing capacity in Polymyces sp. was likely substantially diminished relative to that in D. pertusum and other shallow-water azooxanthellate scleractinians.

Fig. 6.

Opsins and phototransductive pathways in scleractinians. a Gene numbers of opsins in the scleractinians of Robust and Complex clades. The photoreceptor opsins in cnidarians were classified into three groups, namely, cnidopsin, anthozoan-specific opsins I (ASO-I) and ASO-II, based on phylogenetic clustering. b Alignment of opsins from deep-sea scleractinians and those from shallow-water Astreopora myriophthalma as a reference. Conserved functional sites and motifs are indicated by arrows and dotted rectangles, respectively. c Schemes of the postulated phototransduction cascade. The genes whose numbers decreased in deep-sea scleractinians are colored pink

To further explore whether ASO homologs are expressed across different developmental stages, publicly available transcriptomes of the shallow-water coral Pocillopora damicornis were analyzed [33, 34]. The results indicated that ASOI and ASOII homologs exhibited comparatively higher expression during larval stages (Additional file 2: Fig. S13), suggesting that ASO-mediated photosensitivity may contribute to larval dispersal during development. Although larval stage transcriptomes are not yet available for deep-sea scleractinians, the presence and transcriptional activity of ASO detected in both deep-sea adults lead to speculation that their larvae may likewise retain a photoreceptive capacity, especially when encountering light during pelagic dispersal (e.g., vertical dispersal through upward transport into shallower illuminated water layers).

The canonical circadian clock is principally established by a transcription-translation feedback loop [35], whose components (cryptochrome (CRY), CLOCK, CYCLE, and TIM) were detected among all the shallow-water stony coral genomes (Additional file 2: Fig. S14a). The deep-sea D. pertusum, like other shallow-water species, possessed all these complements (Additional file 2: Figs. S10–S12), whereas Polymyces sp. retained only TIM, the function of which, however, appeared unrelated to circadian rhythm regulation in Hexacorallia (Additional file 2: Fig. S12) [36]. Therefore, D. pertusum is likely to maintain a functional circadian rhythm, and Polymyces sp. may have lost this regulatory mechanism. In addition to canonical CRYs, two phylogenetically close gene families (Additional file 2: Fig. S10), photolyase PL (6–4), which repairs UV-induced pyrimidine–pyrimidone (6–4) DNA damage [37], and CRY-DASH, which repairs cyclobutane pyrimidine dimers (CPDs) [38] (Additional file 2: Fig. S14b), also exhibited interesting evolutionary patterns. D. pertusum retained PL (6–4), whereas Polymyces sp. lacked it (Additional file 2: Fig. S14a), reflecting divergent strategies for deep-sea darkness adaptations. In contrast, CRY-DASH was absent in all the azooxanthellate species (Additional file 2: Fig. S14a), which is consistent with their preferences for shaded habitats [4].

Discussion

The HiFi-based genome assembly of the deep-sea solitary coral Polymyces sp. constitutes a central contribution of this work by generating a representative deep-sea reference genome from the Complex clade and fills a major gap in available scleractinian genomes that have been dominated by shallow-water colonial taxa. Based on this valuable resource, several genome-enabled insights into the adaptations of azooxanthellate scleractinians were obtained. First, deep-sea azooxanthellate lineages are not genomically uniform. Despite sharing deep-sea habitats, Polymyces sp. shows substantial reductions in light-sensing and circadian clock complements, whereas the cosmopolitan deep-sea Desmophyllum pertusum retains a more complete toolkit (Fig. 6a), suggesting divergent evolutionary solutions in darkness. Second, gene repertoire reduction is a prominent route to heterotrophic adaptation. In particular, gene families involved in photosensitivity, lipid metabolism, and mitochondrial function are consistently reduced in azooxanthellate scleractinians (Fig. 5), suggesting an energy-saving strategy through eliminating gene complements with low functional demand. Third, convergent adaptations centered on vesicle trafficking and endosomal pathways are identified, implicating cellular transport remodeling as a recurrent feature of the lineages possessing a photosymbiosis-free lifestyle (Fig. 4). In the following sections, we will discuss genomic diversity patterns and demographic histories, synthesize shared adaptive signatures related to heterotrophy, and then consider the adaptive benefits and potential risks of gene loss.

Genomic diversity and historical demography

Both Desmophyllum pertusum and Madrepora oculata are recognized as important reef-building scleractinians in deep waters, forming complex three-dimensional structures and fulfilling ecological functions comparable to those of warm-water reef builders [23]. Although both species inhabit similar deep-sea niches, M. oculata exhibited relatively low genomic heterozygosity (Fig. 3a). The limited framework-building capacity and weaker competitive ability of M. oculata may constrain its spatial expansion and ecological dominance [23]. In a broader context, compared with the tropical common reef builders Porites australiensis and Stylophora pistillata, both deep-sea species displayed lower genomic diversity (Fig. 3a). These results likely reflect the evolutionary advantages conferred by photosymbiosis, which increase energy acquisition, promote growth, and facilitate reproduction. In contrast, the survival of deep-sea scleractinians in energy-limited environments relies solely on occasional heterotrophic feeding. These energy constraints increase their susceptibility to the impacts of environmental disturbances and anthropogenic stressors [39, 40].

Except the deep-sea solitary coral Polymyces sp., the species with higher levels of genomic heterozygosity (Fig. 3a) showed larger Ne sizes in recent times (Fig. 2b, c). The aberrantly higher genomic diversity observed in Polymyces sp. (Fig. 3a) should be attributed to the effects of its ancient demography, with genetic variations gradually accumulating in long-term stable deep-sea environments [41]. The effect of such environmental stability could have buffered the erosion of genome-wide heterozygosity. In contrast, the Ne of Polymyces sp. was dramatically reduced in recent times, consistent with the signal of recent inbreeding detected in its genome (Figs. 2b and 3c). The decrease in Ne may have been driven by external factors such as environmental stressors and habitat fragmentation [42], as well as internal genetic constraints (such as gene losses) that may have limited its adaptive potential [43].

Although reef-building corals are often considered highly sensitive to global climatic perturbations [9], their demographic trajectories revealed unexpected patterns during both the Pliocene Reef Gap (PRG) and the Last Glacial Maximum (LGM). PRG refers to a pronounced interval during the early Pliocene (5.33–7.25 Ma) when reef development in the central Indo-Pacific largely collapsed or drowned, resulting in a significant reduction or absence of coral reef frameworks [44]. In our PSMC results, however, all the species analyzed showed a steady increase in terms of Ne spanning that ancient period (Fig. 2b, c). These results suggested that the PRG phenomenon could be regionally restricted and improper to generalize. The LGM was approximately 20 kya, when great climate change (global cooling cycles) led to mass extinction among many kinds of marine faunas [22]. However, the fact that Ne fluctuations are not observed in the stony corals during this period (Fig. 2b and c) lends credence to the impacts of the LGM, which, although it fragmented the ranges of higher-latitude faunas into smaller refugia, likely had less impact on the tropical habitat.

Convergent evolution of azooxanthellate stony corals

The convergent genetic adaptations of azooxanthellate stony corals can be broadly categorized into two aspects based on our results (Figs. 4, 6, S13, and S14): the regulation of cellular processes for the handling of foreign cells and metabolic remodeling under conditions of nutrient limitation.

In symbiotic corals, strict control of vesicle-mediated endosome paths is essential for preventing intracellular attack on symbionts through the inhibition of phagosome–lysosome fusion and apoptosis [45–47]. In azooxanthellate species, PSGs associated with vesicular transport and apoptosis formed a noticeable component; moreover, the positively selected KIF28P that engaged in vesicle transport potentially evolved novel functions (Fig. 4). These results implied that the processes of vesicle transport and apoptosis had evolved to accommodate the azooxanthellate lifestyle without the requirements for recognizing and tolerating symbiotic algae. Because the vesicular trafficking pathway also plays a role in the intracellular handling of engulfed nutrients, such genetic modifications could also facilitate nutrient uptake in fully heterotrophic scleractinians. Consistent with this interpretation, transcriptomic analysis showed that nearly all the vesicular trafficking-related PSGs were expressed across azooxanthellate species, with median expression levels in the shallow-water azooxanthellate corals (Cladopsammia gracilis and Dendrophyllia cribrosa) slightly higher than those in the deep-sea species (Desmophyllum pertusum and Polymyces sp.) (Additional file 2: Fig. S15). This pattern is broadly compatible with higher food availability and accordingly greater nutrient assimilation demands in shallow-water habitats. Nevertheless, RNA integrity and expression profiles from deep-sea specimens can be affected by drastic pressure and temperature changes during sampling, which may introduce bias into comparative results. Additional biological replicates and improved in situ fixation approaches will therefore be important to more rigorously test these hypotheses.

The metabolic aspects of adaptations are reflected in the contractions of gene families shared among the four azooxanthellate scleractinians, which are associated with photosensitivity (RDH8), lipid metabolism (ELOVL, NRF6), and mitochondrial energy metabolism (ALXA) (Fig. 5). Retinol dehydrogenase 8 (RDH8) is a retinol dehydrogenase involved in phototransduction [48]. The decreases in RDH8 numbers in azooxanthellate scleractinians are consistent with their preferential occurrence in dark habitats with less necessity to sense light. Given that energy is typically stored first as glycogen before being converted to lipids, the reductions in lipid- and energy-related genes suggest metabolic streamlining under conditions with limited food resources.

Rethinking adaptation by gene loss

Gene loss has been recognized as a driving force in metazoan evolution [49]. However, in azooxanthellate stony corals, our findings indicate that gene loss has both adaptive advantages and potential risks.

On the advantageous side, gene loss may streamline cellular metabolism by removing redundant cellular functions in low demand, thereby conserving more energy to survive in nutrient-poor water. For instance, GFP-like chromoproteins (GFPLs), which protect symbiotic algae from excessive light [26], were lost in deep-sea Polymyces sp. and Desmophyllum pertusum (Fig. 5), reflecting their adaptations to deep-sea habitats without threats of UV damage. Similarly, the loss of carbonic anhydrase 1 (BCA1) in D. pertusum may have led to its reduced calcification rates [50]. This reduction aligned with the decline of symbiont-enhanced calcification in aposymbiotic corals [9].

On the other hand, the low demands for certain functions may be temporary. The complete loss of related genes could limit potential opportunities for adaptation. For example, although losses of light-sensing and circadian rhythm-regulating genes (Cnidopsin, ASOII, CRY, Clock, and Cycle) in Polymyces sp. may prune redundant biological processes under the darkness of the deep sea (Fig. 6 and Additional file 2: Fig. S14), such losses can also compromise its ability to exploit occasional bioluminescent cues to catch prey or synchronize feeding behaviors with zooplankton diel migration [51–53]. In contrast, the cosmopolitan deep-sea D. pertusum retains both light-sensing and circadian gene complements, conferring ecological advantages that could contribute to its broader geographic distribution. Similarly, the loss of autophagy-related genes (ATG9A, RICTOR, and NPRL3; Fig. 5) in the shallow-water azooxanthellate Dendrophyllia cribrosa eliminates a critical mechanism for recycling cellular components under nutrient-limited conditions, although D. cribrosa have eliminated the need for autophagy to integrate symbiont-derived nutrients (Fig. 5). Reductions in immune-related gene families (NWD1, NLRP1b, NLRC, CACTIN, etc., Fig. 5) in Polymyces sp. and D. cribrosa may weaken defenses against pathogens, although a trade-off of diminished symbiont recognition requirements for both species. These risks are particularly concerning given the evidence of recent inbreeding and low recent Ne detected in Polymyces sp. and D. cribrosa.

Taken together, these findings suggest that azooxanthellate stony corals exhibit divergent evolutionary trajectories, with the outcomes of gene loss ultimately shaped by the delicate balance between energy economy and functional necessity.

Conclusions

This study presents the first genome of a solitary azooxanthellate scleractinian coral, Polymyces sp., and provides comparative genomic insights by contrasting azooxanthellate corals with their zooxanthellate relatives. We found that azooxanthellate corals generally exhibit reduced genomic diversity and, in some cases, signals of recent inbreeding, suggesting that they are more vulnerable than shallow-water species. We detected positive selection in genes related to vesicle transport among all the studied azooxanthellate scleractinians, highlighting molecular components that may underpin adaptation to a fully heterotrophic lifestyle. In contrast, lineage-specific gene losses in Polymyces sp. or Dendrophyllia cribrosa—two species exhibiting recent inbreeding and reduced recent effective population sizes—include those involved in circadian rhythm, photosensitivity, autophagy and immunity, which reflects energy-saving trade-offs that may come at the cost of reduced functional flexibility. Notably, the cosmopolitan deep-sea species Desmophyllum pertusum retains complete light-sensing and circadian gene sets, whereas the deep-sea Polymyces sp. exhibits a substantial reduction in the canonical components of these systems. This contrast suggests that light perception and biological clock systems may remain important even in deep-sea environments, possibly contributing to their differing distributional ranges in the deep sea. Taken together, our results illustrate multiple genomic routes toward heterotrophy in scleractinians, provide new perspectives on coral evolutionary plasticity and inform future conservation efforts for azooxanthellate species.

Methods

Sample collection and genomic sequencing

Azooxanthellate scleractinian corals are mostly solitary [12]. However, no genomes of solitary corals have been reported to date. To make the comparative analysis more comprehensive, a solitary coral, Polymyces sp. (Additional file 3), was collected at a depth of 970 m from the seamount (10°04.69′ N, 140°12.19′ E) on the Caroline Ridge. The sampling was collected using a suction sampler mounted on the ROV Faxian carried by R/V Kexue in 2019. The sample was flash frozen in liquid nitrogen and then kept in a − 80 °C freezer. Most of the collected solitary coral was used for DNA extraction according to the standard phenol–chloroform extraction method. A paired-end DNA library for genome survey and genomic diversity analysis was constructed with an insert size of 300 bp using a Rapid Plus DNA Lib Prep Kit for Illumina (ABclonal, USA). The library was subsequently sequenced on an Illumina NovaSeq X plus sequencer (2 × 150 bp paired-end reads) (Illumina, USA). The mitochondrial genome was extracted from the Illumina short reads by MitoZ 3.6 [54] and annotated with MITOS 2 [55], followed by manual curation to refine gene boundaries. For HiFi-read sequencing, a 15-kb SMRTbell library was constructed with a molecular weight of genomic DNA larger than 15 kb following the instructions of the SMRTbell Template Prep Kit (PacBio, USA). The library was subsequently sequenced on a PacBio Sequel system with a Sequel Sequencing Kit 3.0 (PacBio, USA). A total of two SMRT cells were run. All sequencing services were provided by Novogene Bioinformatics Technology Co., Ltd. (Tian Jin, China). As our targeted sample in the deep sea was rare and solitary coral was not easily collected by the manipulator arm on the ROV, only one Polymyces sp. was obtained during the entire campaign. The remaining Polymyces sp. tissue after DNA extraction was not sufficient for Hi-C library construction; thus, chromosome-scale scaffolding was not realized in the present study.

Illumina paired-end reads are required for investigating population demography and genome-wide heterozygosity, yet in many published stony coral genomes, such data are either absent [56–58] or derived from sperm [59–61] that lack heterozygous information. To supplement these data, Illumina short reads were generated for two deep-sea frame-building species, Desmophyllum pertusum and Madrepora oculata, as well as one shallow-water common reef builder Porites australiensis. Specimens of D. pertusum and M. oculata were collected from the seamounts of the South China Sea (536 m depth, 15°03.53′N, 116°33.23′E; 2022) and the slope of the Mariana Trench (592 m depth, 11°17.11′N, 139°24.23′E; 2016), respectively, using the manipulator arm on the ROV. P. australiensis were collected on Weizhou Island (21°02.59′ N, 109°08.92′ E; 2025) in Guang Xi, China, by scuba diving. Library construction and Illumina sequencing followed the above procedures as described for the Polymyces sp. dataset.

Genome assembly and annotation

Data cleaning for the Illumina short reads was performed by Trimmomatic v. 0.36.3 [62]. The generated clean data used for the genome survey were analyzed with the k-mer method implemented in GenomeScope 2.0 and Jellyfish 2.2.3 [63]. The genome was assembled using Hifiasm 0.20.0 [13] with the PacBio HiFi reads under the parameters “-l 3 -s 0.5”. Sequencing contamination was screened by FCS-GX 0.5.5 [64], and no sequences were detected to originate from microbial sources.

Repeat masking and protein-coding gene prediction were performed by the GETA 2.4.13 pipeline (https://github.com/chenlianfu/geta), as described in our previous genomic study for deep-sea flytrap sea anemone Actinoscyphia liui [15]. Briefly, the repeated sequences were predicted and masked using RepeatModeler2 [65] and RepeatMasker [66]. Protein-coding genes were then predicted on the basis of the soft masked genome through the integration of homology-based analysis, transcript analysis, and ab initio gene prediction. Homologous sequences were combined from the protein sequences of other published stony coral genomes. The transcript data were derived from the tissues of Polymyces sp. left over after DNA extraction. Ab initio prediction was performed by the training software AUGUSTUS 3.3.3 [67] based on the obtained transcripts. Under the parameters in the conf_for_big_genome.txt file of the GETA pipeline for evidence integration, high-quality protein-coding genes were ultimately generated. The completeness of the assembled genome and predicted protein-coding genes was evaluated by BUSCO v. 4.1.4 [68] using the metazoa_odb10 database.

The translated amino acid sequences of all the protein-coding genes were blasted against the SwissProt [69], KEGG [70] and GenBank nr protein databases with the BLASTp program using an E-value threshold of 1e⁻⁵. Protein domains with known functions were scanned for the full set of predicted proteins using PfamScan v. 1.6

with an E-value threshold of 0.05 and cross-checked using InterPro [71].

Orthologous gene identification and genome synteny analysis

Orthologous gene families were identified among the genomes of 16 Hexacorallians, including six genomes in the Robust clade of scleractinians (Catalaphyllia jardinei, Desmophyllum pertusum, Orbicella faveolata, Pocillopora damicornis, Platygyra sinensis, and Stylophora pistillata) [11, 20, 56–58, 72]; eight scleractinian genomes of the Complex clade (Acropora digitifera, Astreopora myriophthalma, Cladopsammia gracilis, Dendrophyllia cribrosa, Montipora capricornis, Polymyces sp., Porites australiensis, P. lutea) [19, 59–61, 73]; and two genomes in Corallimorpharia (Discosoma sp. and Amplexidiscus fenestrafer) [74]. Except for the genome of Polymyces sp. that was generated in this study, all the gene models of the other 15 Hexacorallians were downloaded from public databases (Additional file 1: Table S15). The longest transcript variant from each gene was extracted by the AGAT toolkit [75] and used as the representative sequence of the gene. Orthologous genes among the 16 genomes were clustered by OrthoFinder 2.2.6 [76].

Ancestral linkage groups (ALGs) of cnidarians were retrieved from Zimmermann’s research [77]. These ALGs were projected onto the genome of the sea anemone Nematostella vectensis to construct a reference framework. To investigate the macrosyntenic relationships between the genomes of scleractinians and cnidarian ancestors, the PanSyn pipeline was used [78]. First, the macrosyn1 module was used to assign scleractinian genes to ancestral gene families. The macrosyn2 module was subsequently used to identify conserved macrosyntenic gene pairs between ancestral and extant genomes. Finally, the results of the Oxford plots were visualized using macrosyn3.

Phylogenetic tree construction and divergence time estimation

The CDSs of single-copy orthologs among the 16 Hexacorallians identified by OrthoFinder 2.2.6 were used to perform phylogenomic analysis. One-to-one CDSs in each orthologous group were aligned using MACSE 2.07 [79] separately and then trimmed and concatenated using Gblocks 0.91b [80]. After being translated to amino acid sequences, the concatenated alignments were used to infer the phylogenomic relationships with IQ-Tree 1.5.5 [81]. The species in Corallimorpharia were used as outgroups. The phylogeny based on mitochondrial genomes was reconstructed using the same pipeline, with the deep-sea coral Madrepora oculata (OR019701) additionally included in the dataset.

To estimate species divergence time, the MCMCTree module in the PAML package [82] was employed. Three calibration points were established according to the fossil record database: a minimum of 165 Ma and a soft maximum of 225 Ma for Acropora and Pocillopora [83], a minimum of 37 Ma and a soft maximum of 67 Ma for Acropora and Montipora [84], and a minimum of 160 Ma and a soft maximum of 291 Ma for the first appearance of Caryophylliidae [85]. An additional calibration point for the divergence of Poritidae and Dendrophylliidae was set as 62.3–132.8 Ma based on the results from the TimeTree website [86]. The first 200,000 iterations were discarded as burn-in. One sample was collected every 50 cycles, yielding a total of 20,000 samples. The program summarized the means and corresponding 95% confidence intervals (CIs).

Historical demography

Illumina short reads of tissue-derived DNA were generated in this study for Polymyces sp., Desmophyllum pertusum, Madrepora oculata, and Porites australiensis. Sequencing data for Catalaphyllia jardinei, Dendrophyllia cribrosa, and Stylophora pistillata were downloaded from GenBank (Additional file 1: Table S4). BWA 0.7.17 [87] was used to align these short reads to their respective genomes or to those of closely related congeners (Additional file 1: Table S4).

To infer the historical changes in effective population sizes (Ne), a pairwise sequentially Markovian coalescent (PSMC) model was applied. Consensus genome sequences were generated using SAMtools 1.21 [88], BCFtools 1.21 [89], and vcfutils.pl based on the bam files obtained from BWA 0.7.17 alignment. Read depth filters were set at one-third to three times the average sequencing coverage, with a minimum mapping quality threshold of 20. PSMC 0.6.5 [90] was run with the following parameters: -N25 -t15 -r5 -p "4 + 25*2 + 4 + 6". The generation times for the branching shallow-water corals (D. cribrosa and S. pistillata) were set at 5 years [91], and those for the brain shallow-water corals (C. jardinei and P. australiensis) were set at 8 years (https://www.scuba.com/blog/life-cycle-of-coral/). Owing to the limited records on the age of sexual maturity in deep-sea scleractinians, a value of 10 years was adopted in our analysis, which was estimated based on the smallest mature sizes and growth rates of deep-sea scleractinians [92, 93]. The mutation rates for each species were calculated by the CODEML module in the PAML v4 package [82] under a free-ratio model. The mutation rates and estimated generation times were used to scale the values of effective population size. One hundred replicate bootstraps for each species were performed.

Activities of transposable elements

After the prediction of repeat sequences, the activities of transposons were further investigated. The dynamic landscape associated with four major categories of transposable elements (LINEs, SINEs, LTRs, and DNA transposons) was parsed with the Perl script parseRM.pl for each species [94]. The mutation rates used as inputs for the parseRM.pl were estimated with the same method as in the previous section. The results of the analysis were packed into 5 Ma bins.

Genomic diversity

Variant calling and hard filtering were conducted using GATK 4.0.12 [95]. Additional filtering was applied with BCFtools 1.21 [89] to exclude the sites with read depths outside 50–250% of the genome-wide average coverage or with uncalled genotypes (./.). Genome-wide heterozygosity was estimated as the proportion of heterozygous genotypes among all callable sites that passed filtering. To assess local heterozygosity, a sliding window with a 1 Mb window and 20 kb step was used to calculate the ratio of heterozygous genotypes to callable sites within each window. Windows with less than 80% callable sites were excluded.

Runs of homozygosity (ROH) were calculated using PLINK 1.9.0 [96], allowing up to three heterozygous sites per window to accommodate potential sequencing errors. A minimum ROH length of 500 kb was required for at least 50 variants. Default settings were applied for all remaining parameters.

Gene family expansion and contraction

Branch-specific expansions and contractions of orthologous gene families were inferred with CAFE 4.2.1 [97]. To account for potential artifacts caused by genome assembly and annotation errors, the Python script caferror.py (included in CAFE) was employed to estimate error distributions. Gene family expansions and contractions were identified under a conditional p value threshold of less than 0.01. A gene family was defined as expanded or contracted if its size in a certain lineage was larger or smaller than both its ancestral value and those in all the other compared lineages. Only families whose Viterbi p values were less than 0.05 were considered significantly changed. GO and KEGG enrichment analyses for the significantly altered gene families were performed with GOEAST [98] and KOBAS [99], respectively.

Gene losses can be misidentified because of annotation errors, where genes that are actually present in the genome fail to be predicted. To validate the gene families considered lost, the homologous protein sequences from the species retaining these families were used as queries to search the genomes where gene losses were inferred. The search was conducted using tBLASTn and Genewise searches to minimize false negatives [100].

Testing for signals of positive selection and relaxed selection

The azooxanthellate scleractinians, Polymyces sp., Desmophyllum pertusum, Cladopsammia gracilis, and Dendrophyllia cribrosa were set as foregrounds in four separate positive analyses. Background species were selected on the basis of phylogenetic relationships (Fig. 2a). For D. pertusum in the Robust clade, Orbicella faveolata, Catalaphyllia jardinei, Stylophora pistillata, and Pocillopora damicornis were assigned as backgrounds. For the other three azooxanthellate scleractinians in the Complex clade, Porites australiensis, P. lutea, Astreopora myriophthalma, and Montipora capricornis were selected. Positive selection was assessed on every CDS alignment using the branch-site model implemented in the CODEML program of PAML 4 [82]. Genes were considered under positive selection if the LRT p value after FDR adjustment was < 0.05 and if at least one codon exhibited a posterior probability > 0.9.

To test whether selection pressures on azooxanthellate scleractinians have been relaxed because of reduced functional demands, the RELAX model implemented in HyPhy 2.5.64 [24] was applied. The RELAX model evaluates trends and shifts in the intensity of natural selection along specified test branches, allowing the detection of relaxed (k < 1) or intensified (k > 1) selection on a given gene. The assignment of test/background branches followed the methodology used in the above positive selection analyses. Genes with a significant result of k < 1 (p < 0.05) were inferred to be under relaxed selection.

Targeted analysis of genes associated with photosensitivity and the circadian clock

For targeted gene homology searches, lists of candidate genes were compiled from previous studies on opsins [31, 101] and circadian rhythms [30, 36]. For each of the candidate gene lists, custom BLAST databases were generated with a minimum of two high-quality reference sequences selected from the curated Swiss-Prot protein database. The protein sequences from scleractinian genomes were used as query sequences against the custom databases in BLASTp homology searches (E value cutoff of 1 × 10⁻⁶). The BLAST results for opsins were aligned using MUSCLE 3.8.1551 [102] and further screened by excluding the sequences without lysine K296, which is essential and indicative of chromophore binding. For the phylogenetic analysis of opsins and CRY/PL, sequences of 68 genera previously used in [31] covered a wide phylogenetic spectrum ranging from Fungi to Vertebrate. For circadian clock genes, a phylogenetic tree was constructed based on the sequences used in [36], with the identified human and fly genes included.

Transcriptome analysis

Clean reads from each species were aligned to their corresponding genomes using HISAT2 2.2.1 [103]. The resulting alignments were processed with StringTie 2.1.5 [104] to quantify transcripts for each dataset. Transcript abundance was represented as TPM. Accession for the transcriptome datasets is provided in Additional file 1: Table S16.

Due to the lack of bulk RNA-seq data for Cladopsammia gracilis, gene expression of this species was quantified based on a public single-cell RNA-seq dataset (Additional file 1: Table S16). Raw reads were adapter-trimmed and quality-filtered using fastp v1.0.1 [105]. The UMIs were extracted from read 1 and appended to the read name of read 2 using umi_tools 1.1.6 [106]. The processed reads were aligned to the reference genome using STAR 2.7.11 [107], with PCR duplicates removed by umi_tools [106]. Gene-level counts were generated from the de-duplicated alignments with featureCounts 2.1.1 [108] and transformed to TPM for downstream comparisons.

Protein structure simulation

Structural predictions of the targeted proteins were performed using AlphaFold 2 [109] with the monomer preset and full databases. Local confidence (pLDDT) and predicted aligned error (PAE) were inspected to ensure reliability near the domain junction and positively selected sites. The structures with the highest confidence scores were selected for visualization in PyMOL [110].

Abbreviations

ALGs

Ancestral linkage groups

ASOI

Anthozoan-specific opsin I

ASOII

Anthozoan-specific opsin II

BUSCO

Benchmarking universal single-copy ortholog

CPDs

Cyclobutane pyrimidine dimers

IUCN

International Union for Conservation of Nature

kya

Kiloannum ago

LGM

Last Glacial Maximum

Ne

Effective population sizes

PRG

Pliocene Reef Gap

PSGs

Positively selected genes

RDH8

Retinol dehydrogenase 8

ROH

Runs of homozygosity

ROV

Remotely operated vehicle

TEs

Transposable elements

Authors’ contributions

JYL: Conceptualization, Data curation, Formal analysis, Methodology, Software, Validation, Visualization and Writing. YL: Conceptualization, Investigation, Project administration, Resources, Supervision, and Validation. ZFZ: Investigation, Resources and Validation. XML: Resources and Validation. MKS: Software. KDX: Conceptualization, Validation, Funding acquisition, Project administration, Resources, Supervision, Writing–review& editing.

Funding

This work was supported by the National Natural Science Foundation of China (No. 41930533) and the Senior User Project of R/V KEXUE of the Center for Ocean Mega-Science, Chinese Academy of Sciences (No. KEXUE2019GZ04 and KEXUE2020GZ02).

Data availability

The assembled genome sequences and the raw sequencing read data reported in this paper are deposited in the National Center for Biotechnology Information (NCBI) database (BioProject: PRJNA1321156 and SRA: SRP619318) [111, 112]. The annotation file of the Polymyces sp. can be obtained from the figshare database 10.6084/m9.figshare.30073336 [113]. The mitochondrial genome of Polymyces sp. is under accession number LC916300 [114]. The computer codes used in this study are supplied as supplementary materials (Additional file 4).

Declarations

Ethics approval and consent to participate

Not applicable. This study did not involve human participants, clinical materials, or vertebrate animals. The specimens analyzed are deep-sea scleractinian corals collected for scientific research in accordance with institutional and cruise sampling procedures.

Consent for publication

All authors read and approved the final manuscript.

Competing interests

The authors declare no competing interests.