Insights into cephalochordate genome and gene evolution from the early-diverging amphioxus Asymmetron lucayanum

Significance

Cephalochordates (amphioxus or lancelets) are invertebrate chordates that diverged first in the chordate lineage leading to the vertebrates. Thus, their genomes can help reveal the genetic basis of the evolutionary transition from an invertebrate ancestor to vertebrates. For a comprehensive understanding of the cephalochordate genome architecture and gene repertoires in the context of chordate evolution, we generated a chromosome-scale genome assembly for Asymmetron lucayanum, representing the earliest diverging cephalochordate genus and performed a comparative genomics study at both chromosomal and local scales with species representing other cephalochordate and vertebrate lineages, which provides insights into the genome biology and evolution of cephalochordates.

Abstract

Cephalochordates (amphioxus or lancelet) are considered as living proxies for ancestral chordates due to their key phylogenetic position and slow evolutionary rate. The genomes of living amphioxus thus can help to reveal the genetic basis shaping the evolutionary transition from nonvertebrate animals to vertebrates. To gain a comprehensive understanding of the genome architecture in amphioxus, we generated a chromosome-anchored genome assembly for Asymmetron lucayanum, representing the earliest diverging cephalochordate genus. We show that Asymmetron has an enlarged genome compared to those of the other four cephalochordate genomes decoded so far (all in the genus Branchiostoma), caused by pervasive expansions of intergenic transposable elements (TEs). Nevertheless, both macrosynteny and microsynteny remain highly conserved between Asymmetron and Branchiostoma, enabling reconstruction of the ancestral genomic architecture of the cephalochordate lineage for tracing genome evolutionary processes during deuterostome and chordate diversification. Integration of developmental transcriptomic data further reveals that selective constraints on cotranscriptional regulation underline the maintenance of the conserved microsynteny blocks among cephalochordate species. We also examine the evolutionary history of the Hox cluster in cephalochordates and vertebrates, and identify species-specific inversions and TE invasions at this locus in both Asymmetron and Branchiostoma. Finally, we survey key molecular building blocks underlying both innate and adaptive immunity (e.g., TLR, NLR, MHC, and RAG) and uncover their evolutionary dynamics and plausible ancestry in chordates. Taken together, our findings illuminate the genome and gene evolution of cephalochordates and provide valuable resources for understanding the early evolution of chordates and the origin of vertebrates.

The chordate subphylum Cephalochordata (amphioxus or lancelets) diverged from other chordates (vertebrates and urochordates) more than 550 Mya (1). Modern amphioxus strongly resemble Cambrian fossil chordates such as Pikaia and Haikouella, although the latter, unlike amphioxus, has paired eyes (2, 3). Compared to vertebrates, amphioxus has a similar body plan with a dorsal, hollow nerve cord, pharyngeal gill slits, and a postanal tail (4), suggesting a deep conservation of chordate body and organ architecture. Within the Cephalochordata, Asymmetron represents the earliest diverged lineage while Branchiostoma and Epigonichthys radiated more recently (5, 6). While all three genera display similar body plans, they differ primarily in gonadal arrangement: Asymmetron and Epigonichthys have a single row of gonads on the right side, whereas Branchiostoma features bilateral gonads. Surprisingly, Asymmetron and Branchiostoma can hybridize, although only some hybrids with Branchiostoma mothers complete metamorphosis (7).

To date, the genomes of four species of Branchiostoma (B. floridae, B. lanceolatum, B. japonicum, B. belcheri) have been sequenced. Their genome sizes range from ~383 Mb (B. japonicum) to 490 Mb (B. floridae) and their chromosome numbers vary from 18 (B. japonicum) to 20 (B. belcheri) (8–13). Unlike vertebrates, cephalochordates have not undergone the two rounds of whole-genome duplications (2R-WGDs) characteristic of vertebrates. Therefore, their genome architectures are highly streamlined with much less redundancy. For example, genomes of the Branchiostoma species have a single Hox cluster with 15 colinear genes, in contrast to the quadrupled Hox cluster configuration in most vertebrates (14, 15). Moreover, initial comparison of B. floridae and vertebrate genomes revealed conserved macrosynteny, suggesting 17 ancestral linkage groups (ALGs) in their chordate common ancestor (8, 11). Subsequent studies including additional deuterostome outgroups further suggested the ancestral chordate likely had 23 of 24 putative bilaterian ALGs (13, 16, 17). In addition, while cephalochordates, like other nonvertebrate animals, only have an innate immune system, they do possess the prototypical forms of some key molecular building blocks of adaptive immunity (e.g., the MHC locus and RAG genes) (12, 18–22). Together, these unique features make cephalochordates the best living proxy for the ancestral chordates before the rise of vertebrates. A more comprehensive understanding of cephalochordate genome biology and evolution has been hindered by the limited genomic data existing for Asymmetron and Epigonichthys. To bridge this gap, we chose Asymmetron given the discovery of an accessible population of A. lucayanum in Bahamas (23, 24). Our transcriptome analysis revealed very slow evolution of A. lucayanum protein-coding genes, indicating its suitability for studying early chordate evolution (6). Together with some initial genome sequencing data, we also examined conserved noncoding elements, primordial germ cell formation, and the evolution of genes coding for fluorescent proteins in Asymmetron (25–28).

In the present study, we used PacBio sequencing and Hi-C to generate a chromosome-level genome assembly for A. lucayanum. We showed that while the A. lucayanum genome is much larger than those of Branchiostoma species due to transposable element expansion, the evolutionary rate of its protein-coding genes remains characteristically slow. Its local and chromosomal genomic architectures are also highly conserved with those of Branchiostoma. By resolving lineage-specific interchromosomal fusions in these two genera, we reconstructed a cephalochordate ancestor with 20 ALGs and confirmed the loss of a previously defined bilaterian ALG (ALG R) in the chordate lineage (16). Furthermore, we detected coordinated expression of genes within microsynteny blocks and illuminated the evolutionary history of some key genomic loci and genes implicated in development and immunity. Notably, we found species-specific inversions and transposable element invasion in the Asymmetron Hox cluster. Collectively, our results provide resource and insights for understanding the evolution of both genome architecture and gene repertoire in cephalochordates.

Results

Chromosome-Level Genome Assembly of A. lucayanum.

Using PacBio and HiC sequencing data, we generated a 721.3-Mb chromosome-level reference assembly for A. lucayanum (Alu) with haplotype reconciliation. This reference assembly distributed across 18 chromosomes, with the chromosome-anchored assembly size being 677.1 Mb (Fig. 1A and SI Appendix, Tables S1–S3). This agrees with the genome size estimates based on both read k-mers (Fig. 1B) and our initial diploid assemblies (~1.3 to 1.4 Gb) (SI Appendix, Table S2). This substantially larger genome size of Asymmetron than that of Branchiostoma species is in part due to the expansion of transposable elements (TEs) in both intronic and intergenic regions (Fig. 1 C and D and SI Appendix, Table S4), especially the inverted terminal repeat TEs (TIR-TEs) (SI Appendix, Fig. S1). While their expansion time is tricky to estimate, we found the expansion of long-terminal-repeat TEs (LTR-TEs) in A. lucayanum occurred relatively recently (<1 Mya) (SI Appendix, Fig. S2). By integrating Iso-Seq and RNA-Seq data, we annotated 35,203 protein-coding genes, 94.3% of which are located on assembled chromosomes (SI Appendix, Fig. S3 and Table S3). The sizes of coding regions are comparable between the two genera, i.e., 46.9 Mb in Asymmetron and 40.1 to 47.3 Mb in Branchiostoma. Our A. lucayanum assembly and annotation exhibit comparable quality to the existing chromosome-level Branchiostoma assemblies (SI Appendix, Table S6).

Fig. 1.

Chromosome-level genome assembly and annotation of A. lucayanum. (A) Hi-C contact heatmap of the A. lucayanum reference haplotype assembly. (B) Correlations among the assembled genome size, k-mer-estimated genome size, and assembled repetitive sequence size. (C) Genome-wide distribution of different genomic features along the A. lucayanum genome. (D) Relative and absolute abundance of different genomic features across the five cephalochordate genomes. (E) Phylogeny and segmental duplication content across chordate genomes. Alu: A. lucayanum; Bbe: B. belcheri; Bfl: B. floridae; Bja: B. japonicum; Bla: B. lanceolatum; Eat: Eptatretus atami; Gga: Gallus gallus; Hsa: Homo sapiens; Loc: Lepisosteus oculatus; Lva: Lytechinus variegatus; Mmu: Mus musculus; Oan: Ornithorhynchus anatinus; Pfl: Ptychodera flava; Pye: Patinopecten yessoensis; Xtr: Xenopus tropicalis. MRCA: most recent common ancestor.

Universally Slow Evolution of Protein Coding Genes in Cephalochordates.

We reannotated the chromosome-level assemblies of the four Branchiostoma species (B. belcheri [Bbe], B. floridae [Bfl], B. japonicum [Bja], B. lanceolatum [Bla]) with their respective transcriptomes in the same way as we did for Asymmetron (SI Appendix, Tables S3–S8). For each Branchiostoma species, approximately 20,000 protein-coding genes in our annotation correspond to those in the original annotation (SI Appendix, Fig. S4 and Dataset S1). Those that lack matches in the original sets tend to have shorter lengths and higher dN/dS values (SI Appendix, Figs. S5 and S6). We incorporated three more nonvertebrate bilaterians (scallop, sea urchin, acorn worm) and seven vertebrates (human, mouse, platypus, chicken, frog, spotted gar, brown hagfish) for a phylogenetic analysis based on 947 shared single-copy orthologs (Fig. 1E and SI Appendix, Table S9). We estimated a divergence time between Asymmetron and Branchiostoma as about 85.5 Mya, with the speciation of the four Branchiostoma species occurring more recently ~24.2 to 41.1 Mya, in contrast to the deep pre-Cambrian divergence between cephalochordates and vertebrates (SI Appendix, Fig. S7 and Table S10). We compared the divergence-time-calibrated amino acid substitution rate between cephalochordates and vertebrates relative to their shared most recent common ancestor (i.e., chordate MRCA) and confirmed a significantly slower rate of protein evolution in cephalochordates (SI Appendix, Fig. S8; phylogenetic generalized least squares regression: P = 7.29 × 10⁻⁹). In the same context, A. lucayanum and B. japonicum showed the slowest protein evolution rates. It is tempting to conjecture that the 2R-WGDs and subsequent rediploidization in vertebrates likely contributed to such molecular evolution rate difference between cephalochordates and vertebrates. In contrast, both Asymmetron and Branchiostoma cephalochordates showed a higher level of segmental duplications compared with vertebrates (Fig. 1E). This agrees with previous reports for Branchiostoma species (12, 13) and suggests segmental duplication as a shared shaping force for cephalochordate genome evolution. Based on intracephalochordate divergence times and corresponding synonymous substitution rates summarized from the 947 single-copy orthologs (SI Appendix, Fig. S9), we further estimated a cephalochordate-specific mutation rate of 4.37 × 10⁻⁹ substitution/site/year, concordant with a recent estimate by pedigree sequencing in B. floridae (29). This estimated mutation rate for cephalochordates falls within the typical range for metazoans and is therefore unlikely to explain their slow protein evolution.

Karyotype Evolution of Cephalochordates.

The chromosome-level genome assemblies of both Asymmetron and Branchiostoma species revealed a general one-to-one chromosomal correspondence between the two genera despite their 85.5-Mya divergence. The main difference is two additional Asymmetron–specific interchromosomal fusions in chromosomes 1 and 3 (Fig. 2A). By matching with the 24 Bilaterian ancestral linkage groups (ALGs) as well as the 29 Bilaterian, Cnidarian, and Sponge (BCnS) ALGs previously defined (16), we reconstructed the history of karyotype evolution in cephalochordates. These results extended the previously proposed 20 ALGs for the ancestral Branchiostoma to the common cephalochordate ancestor. These 20 ALGs correspond to 23 of 24 Bilaterian ALGs and 28 of 29 BCnS ALGs (13). Starting from the 24 Bilaterian ALGs, our analysis confirmed the loss of the R ALG (one of the 24 Bilaterian ALGs) in chordates (16). In addition, we identified three ancestral interchromosomal fusion events (A1⊗A2, C1●J2 and I●O1) that occurred before the Asymmetron–Branchiostoma divergence, among which the J2/C1 ALG fusion showed distinct patterns in Asymmetron and Branchiostoma (Fig. 2B). Specifically, Asymmetron may have undergone additional inversion events resulting in a “J2-C1...C1-J2” pattern, or alternatively Branchiostoma might have undergone J2●C1 (J2 insert into C1) while Asymmetron could have experienced C1●J2 (C1 insert into J2). Both scenarios could explain the observed patterns based on the model of metazoan chromosome dynamics (16). Conservation of synteny around the J2/C1 fusion site compared to the prefusion condition in the hemichordate Ptychodera flava (17) revealed additional reshuffling between the J2 and C1 ALGs in A. lucayanum compared to Branchiostoma species (SI Appendix, Fig. S10).

Fig. 2.

Evolution of the karyotypes of the five amphioxus species. (A) Conserved macrosynteny across five cephalochordate species. The chromosomal karyotype of each species is shown horizontally, with lines linking orthologous genes shared among them (colored by A. lucayanum chromosomes). (B) The reconstructed history of cephalochordate karyotype evolution. Each colored block represents one or more (when they still stay together on the same chromosome in cephalochordates) previously defined Bilaterian (n = 24) and BCnS (n = 29) ALGs. Different types of chromosomal fusion events (●: fusion without mixing, ⊗: fusion-with-mixing) are further indicated.

Except for B. belcheri, which lacks unique interchromosomal fusions, the other cephalochordate species each had 1 to 2 such fusions (SI Appendix, Table S11). Peaks of segmental duplications were found at all seven fusion sites examined (SI Appendix, Fig. S11), suggesting a possible intrinsic link between these two types of chromosomal rearrangements. In comparison, repetitive sequences are more evenly distributed across the fusion sites. At the global scale, we observed a modest tendency for more segmental duplications and repetitive sequences (but a lower proportion of CDSs) in fusion sites spanning over larger genomic regions (SI Appendix, Fig. S12).

Signatures of Macro- and Microsynteny Conservation in Chordate Evolution.

We detected a high degree of macrosynteny (i.e., chromosomal gene colocalization) between Asymmetron and species of multiple bilaterian lineages (SI Appendix, Fig. S13). Combined with similar findings made with Branchiostoma (8, 11, 13), this suggests universally slow evolution of macrosynteny in cephalochordates. Regarding microsynteny (i.e., the conservation of local gene orders), we detected 1,214 pan-cephalochordate microsynteny blocks, each of 3 to 64 genes, encompassing a total of 7,972 orthologous gene groups (SI Appendix, Fig. S14 and Dataset S2). The distribution of block sizes fits best to a log-normal distribution, suggesting additional processes such as selection modulating the exponential decay of gene order conservation (SI Appendix, Figs. S15 and S16). Among one-to-one orthologs shared across the five cephalochordate species, we found genes within microsynteny blocks show slower rates of molecular evolution (dN and dS) but similar strengths of selection constraints (dN/dS) compared with genes outside these blocks (Fig. 3A). Interestingly, the in-block correlation of gene-specific measurements of both selection (dN/dS) and expression were significantly higher than calculations based on reshuffled blocks (Fig. 3 B and C), suggesting that genes contained within the same microsynteny block are inherently more alike regarding these measurements. By further requiring strict in-block gene collinearity with the human genome, we detected 43 such highly conserved cephalochordate-human microsynteny blocks, capturing some well-known gene clusters that are important in development (e.g., Hox and pharyngeal genes) (30, 31) and cancer (the human 3p21.3 tumor suppressor cluster) (32) (SI Appendix, Fig. S17 and Dataset S2). This indicates a strong selection constraint on local gene orders over chordate evolution, potentially linking to the regulation constraint of spatiotemporal transcription.

Fig. 3.

Macrosynteny and microsynteny conservation. (A) Comparison of molecular evolution metrics (dN, dS, and dN/dS) between one-to-one orthologous genes within and outside the cephalochordate microsynteny blocks. (B) Correlations of molecular evolution metrics of genes within the observed cephalochordate microsynteny blocks versus those of genes within randomized blocks. (C) Correlation of multistage gene expressions (measured by transcripts per million, TPM) of genes within the observed cephalochordate microsynteny blocks versus those of genes within randomized blocks. (D and E) Negative correlation between synteny conservation (D: macro-, E: micro-) and divergence time. The two-sided Wilcoxon rank-sum test was used for the comparison in panels A–C.

To quantify the degree of conservation of macro- and microsynteny between different lineages, we performed all possible pairwise comparisons among the seven vertebrates and eight nonvertebrate bilaterians (SI Appendix, Tables S12 and S13). Not surprisingly, conservation of both macro- and microsynteny decreases with increasing divergence time between species, with a more rapid decay of the latter (Fig. 3 D and E). Notably, the hagfish (Eptatretus atami) genome showed substantially less conservation in both macro- and microsynteny compared with cephalochordates and other vertebrates (SI Appendix, Tables S12 and S13). In accordance, we found only eight pan-cephalochordate microsynteny blocks further conserved with the hagfish, which contrasts to the 43 microsynteny blocks shared across all five cephalochordates as well as human (Dataset S2). Such low level of synteny conservation of hagfish relative to other chordate species is likely due to the cyclostome-specific WGD and the subsequent rediploidization (33, 34). Likewise, a generally faster decay of macrosynteny relative to other vertebrate lineages was detected for the mouse within vertebrates (SI Appendix, Table S12), reflecting its accelerated chromosome evolution.

Evolutionary History of Hox Genes in Chordates.

Like Branchiostoma (4, 13), Asymmetron has a single Hox cluster of 15 genes, suggesting that this is the ancestral cephalochordate Hox configuration. Designating hemichordates as the outgroup (17), we analyzed the orthologous correspondence of Hox genes among both Asymmetron and Branchiostoma cephalochordates and several representative vertebrates. This analysis confirmed the recently proposed one-to-one orthology between the cephalochordate and vertebrate Hox1–Hox5 genes (13), while revealing potential supports for the one-to-one orthology for their Hox6 and Hox9 genes (Fig. 4 A and B, SI Appendix, Table S14, and Dataset S3). Using Hox1 as the anchoring phylogenetic root, we found the remaining Hox genes clustering into four additional paralog groups: Hox2–Hox3, Hox4–Hox5, Hox6–Hox8, and Hox9–Hox15. Cephalochordate Hox 10 to 12 genes collectively correspond to the vertebrate Hox10, suggesting that they are derived from cephalochordate-specific duplication events. Cephalochordate Hox13–15 genes are closely related to the vertebrate Hox11–14 genes, but their precise correspondence is obscured, presumably a reflection of so-called “posterior flexibility” (35). Specifically, cephalochordate Hox13 and Hox14 genes likely originated from a cephalochordate-specific gene duplication and together correspond to the vertebrate Hox11. Cephalochordate Hox15 is orthologous to vertebrate Hox13. We detected no cephalochordate ortholog for vertebrate Hox14, which only exists in early-diverged vertebrate lineages such as lamprey, hagfish, shark, and coelacanth (36–39). Our phylogenetic analysis suggests vertebrate Hox14 originated from a tandem duplication of vertebrate Hox13.

Fig. 4.

The Hox cluster evolution in cephalochordates and vertebrates. (A) Phylogeny of Hox genes from cephalochordates, vertebrates, and hemichordates. (B) Proposed model of Hox gene evolution in chordates. Solid lines: genes with unambiguous orthology; dashed lines with question marks: genes with uncertain orthology. Identified paralog groups are color-coded. (C) The dN, dS, and dN/dS of Hox genes based on the pairwise comparison among the five cephalochordate species. Dashed lines: algorithmic means for the anterior, central, and posterior Hox genes respectively. (D and E) Spatiotemporal expression patterns of cephalochordate Hox genes across developmental stages (D) and tissue types (E). (F–H) The local landscapes of 3D genome interactions (HiC) and repetitive elements (RE%, 10-kb nonoverlapped window) of the Hox cluster in B. belcheri (F), B. floridae (G), and A. lucayanum (H), with the B. belcheri and A. lucayanum Hox inversion breakpoints (relative to B. floridae) marked using solid (if breakpoints fall within the plotted region) or dashed (if breakpoints fall outside the plotted region) arrows.

In contrast to the posterior and central Hox genes of cephalochordates, the anterior ones appear to have evolved under more relaxed constraints, as shown by their generally higher dN/dS values (Fig. 4C). The same trend holds for the human-chimpanzee Hox comparison except for a few outliers, suggesting that this trend may be more general (SI Appendix, Fig. S18). Such characteristics of anterior Hox genes are potentially due to a relaxation of purifying selection or site-specific diversifying selection in directing their developmental stage- and tissue-specific expression (Fig. 4 D and E and SI Appendix, Fig. S19). Based on multiple transcriptomic datasets (SI Appendix, Table S5), we observed a prevalent pattern in cephalochordates in which more anterior Hox genes are expressed during early embryonic development, whereas the posterior Hox genes are expressed only at larval or juvenile/adult stages. Notably in all amphioxus species examined, Hox6 orthologs exhibit a more prominent and much earlier onset of expression compared to their immediately neighboring Hox genes, consistent with a previous in situ hybridization survey in B. lanceolatum (40). In addition, we found that anterior Hox genes (except Hox1 and Hox2), as well as central Hox genes, are generally expressed across a broader range of tissue types compared to posterior Hox genes (Fig. 4E).

We also found the A. lucayanum Hox cluster, like that of B. belcheri (13), is inverted relative to those in B. floridae, B. japonicum, and B. lanceolatum, but exhibits distinct breakpoints (Fig. 4 F–H). Interestingly, the inferred breakpoints suggest at least two distinct inversion events around the Hox clusters of both A. lucayanum and B. belcheri, highlighting the surprising structural dynamics of the Hox cluster in relation to its flanking genomic regions (SI Appendix, Fig. S20). In B. belcheri, the two Evx genes are located adjacent to Hox1, whereas in A. lucayanum and other Branchiostoma species examined here, the Evx genes are positioned adjacent to Hox15, consistent with the scenario of independent inversion events (Fig. 4 F–H and SI Appendix, Fig. S21). Hi-C contact maps further reveal that Evx and Hox genes reside within the same topologically associating domain (TAD) in B. floridae (Fig. 4G), but not in B. belcheri (Fig. 4F), likely reflecting structural rearrangements resulting from the Hox cluster inversion. Notably, the temporal expression profiles of Evx1 and Evx2 genes remain largely conserved despite this rearrangement (SI Appendix, Fig. S21), suggesting a likely independence of the Hox TAD architecture. Moreover, although Hox clusters typically contain relatively few repetitive elements (41, 42), we found substantial intra-Hox TE expansions between Hox9 and Hox10 in A. lucayanum (Mutator TE) as well as between Hox13 and Hox14 in B. belcheri (Helitron TE) (Fig. 4 F–H and SI Appendix, Fig. S22). Such co-occurrence of Hox cluster inversion and intra-Hox TE invasion in both A. lucayanum and B. belcheri may be nonrandom. The functional impact, if any, of such co-occurrence remains to be determined.

Ultra-Conserved Regions Shared between Cephalochordates and Vertebrates.

In addition to conserved synteny, conserved genomic regions in evolutionarily diverged lineages also reflect strong purifying selection. Whole-genome alignments revealed 49,178,410 bases conserved among the five cephalochordate species examined and 9,553,475 bases conserved among six vertebrates (spotted gar, frog, chicken, platypus, mouse, human) as well as 2,047,162 bases conserved among all eleven chordates (Fig. 5A). Throughout these comparisons, we found the second codon position of the CDSs consistently overrepresented compared to the first and third positions among the conserved bases, possibly driven by its crucial impact on the physicochemical properties of codons. In addition, there are substantially more conserved bases in the 3’ untranslated regions (UTRs) than in the 5’ UTRs, indicating strong evolutionary constraints on 3’ UTRs.

Fig. 5.

Genomic regions under strong selective constraints. (A) Breakdown bar plots of evolutionarily conserved genomic regions in cephalochordates (Left), vertebrates (center), and across all chordates (Right). (B) Enriched vertebrate transcription factor binding motifs in conserved noncoding regions identified for cephalochordates. (C) Enrichment of human disease-associated variants in the identified chordate UCRs.

Of the conserved bases among the five cephalochordates, 31.52% correspond to CDSs, covering at least 50% of the CDS length of ~70% of the A. lucayanum protein-coding genes. The remaining 68.48% of conserved bases correspond to noncoding regions such as introns, UTRs, and intergenic regions, suggesting their conserved regulatory functions. We merged these conserved bases into 808,417 pan-cephalochordate conserved regions, covering 62.6 to 70.4% of open chromatin peaks previously identified in B. lanceolatum across multiple developmental stages (10) and 70.2 to 83.6% micro-RNAs curated for B. belcheri and B. floridae (43). Of 52 cis-regulatory elements previously determined for Branchiostoma by transgenic reporter assays, 47 (90%) matched our identified pan-cephalochordate conserved regions (Dataset S4). By further projecting these regions to the A. lucayanum genome coordinates with coding and short regions filtered out, we obtained 518,979 nonredundant pan-cephalochordate noncoding regions, showing 579 significant matches to the binding motifs of vertebrate transcription factors (44). Many of these correspond to zinc finger genes, basic helix–loop–helix (bHLH) genes, and homeobox genes (Fig. 5B).

Not surprisingly, there are fewer regions further conserved between cephalochordates and vertebrates, likely due to their >550-Mya divergence and the rewiring of gene regulatory networks during and after the vertebrate 2R-WGDs (45). We merged these conserved regions into 18,584 human-genome-projected ultra-conserved regions (UCRs) that are shared among all 11 chordates. For the 33 chordate UCRs matched with human noncoding regions, 16 of them are strictly noncoding in all compared chordate species, with three such regions harboring known micro-RNA genes (Dataset S4). These 16 noncoding UCRs included known bilaterian/chordate ultra-conserved noncoding elements associated with MSX1, EBF3, HOX4, and ZNF503 (Bicore2) (25, 46). Our analysis filtered out one such known element associated with ID1 (Bicore1) (46) as it was lost in the frog (Xenopus tropicalis) genome. Beyond these known ones, we also identified other noncoding UCRs shared between cephalochordates and vertebrates (Dataset S4). These included four that are closely spaced along the 3’ UTRs of cephalochordate MEX3 and vertebrate MEX3B (SI Appendix, Fig. S23). Interestingly, among the four paralogs of MEX3 in vertebrates (i.e., MEX3A, MEX3B, MEX3C, MEX3D), only MEX3B exhibits such striking conservation in its 3’ UTR. In contrast, the ancestral chordate copy of the vertebrate PTPRN/PTPRN2 UCR was retained in both post-WGD paralogs, likely constrained by the highly conserved microRNA gene miR153 that it harbors. Other noncoding UCRs that we identified are associated with genes including SMAD6, MIB1, SIX1, TNPO1, and TPM1 (Dataset S4).

Although the specific functions of most of these chordate UCRs remain to be determined, given their deep sequence conservation, mutations in them are likely to cause genetic diseases. Indeed, we found that human pathogenic mutations curated in ClinVar (47) are considerably enriched in our identified chordate UCRs (Fig. 5C). In addition, many chordate UCRs that substantially overlapped (≥50% overlap) with human coding regions, are associated with genes involved in Wnt signaling and protein ubiquitination (SI Appendix, Fig. S24 and Dataset S4). Finally, there is a significant overlap with previously defined clinically important genes (48, 49) (1,000 permutations, P = 0.001). Taken together, the wide conservation of these UCRs likely indicates targets of deep functional constraints in chordates.

Pan-Cephalochordate Survey of the Repertoire and Genomic Organization of Immune Genes.

Innate immune systems serve as front-line defenses against pathogens in animals and their chief components are pattern recognition receptors, most notably the trans-membrane Toll-like receptors (TLRs) and the intracellular NOD-like receptors (NLRs) (50). Beyond innate immunity, vertebrates also have adaptive immunity to provide enhanced pathogen defense (51). Despite lacking adaptive immunity, cephalochordates do possess ancestral genomic regions and elements homologous to key components of adaptive immunity (e.g., the MHC locus and RAG genes) (4, 21, 52). Here, we performed a comprehensive genome-wide survey in diverse metazoan phyla to trace their evolutionary history in the context of chordate evolution.

Toll-Like Receptors (TLRs).

In each of the five cephalochordates, we identified 26 to 39 TLR genes, more than those in all examined vertebrates (e.g., 10 in humans and 9 in chicken) (Fig. 6A and Datasets S5 and S6). About one-quarter (e.g., 24.2% in B. japonicum) to one-third (e.g., 34.3% in B. lanceolatum) of these cephalochordate TLRs evidently resulted from tandem duplications (Fig. 6B and SI Appendix, Fig. S25). Notably, cephalochordate TLRs fall into two major groups at the upper and lower halves of the TLR gene tree (Fig. 6A). The upper group is closely clustered with the protostome and nonvertebrate deuterostome TLR genes, suggesting their deep evolutionary ancestry (Dataset S6). All four urochordate TLRs that we identified (from two species) belong to this upper group (Fig. 6A). The lower group is more closely related to hemichordate and vertebrate TLRs, representing evolutionarily more derived TLR genes (Dataset S6). Historically, TLR genes have been classified into two subclasses: single-cysteine cluster (SCC) TLRs and multi-cysteine-cluster (MCC) TLRs, with the latter considered as the characteristic feature for TLRs in protostomes (53). While our analysis confirmed the prevalence of MCC-TLRs in protostomes, we also detected some MCC-TLRs in deuterostome lineages such as hemichordates and cephalochordates (but none in vertebrates) (Fig. 6A and Dataset S6). Of the previously characterized vertebrate TLR genes, closely-related cephalochordate homologs were found for the bacterial sensing TLR5 as well as the viral sensing TLR3 and TLR7/8/9 (Dataset S6). Notably, the divergence among TLR7, TLR8, and TLR9 occurred after the rise of jawed vertebrates, with their pretriplicated ortholog (TLR7/8/9) maintained as a single copy in jawless vertebrates (e.g., lamprey and hagfish). The human-absent TLR11/12/13/21/22 are close related to TLR7/8/9. No closely related cephalochordate homolog was found for the bacterial sensing TLR1/2/6/10/15.

Fig. 6.

Evolution of immune-related genes and gene families in cephalochordates. (A) Phylogenetic tree of metazoan TLR genes. C: Cephalochordata, U: Urochordata, V: Vertebrata, H: Hemichordata, E: Echinodermata. (B) Chromosome-wide distribution of cephalochordate TLR and NLR genes. (C) Macrosynteny conservation between Asymmetron chromosome 7 and human chromosomes 1, 6, 9, and 19. Red and gray lines indicate Asymmetron-human orthology, with red highlighting matches with the human MHC genes or their close paralogs. Respective gene counts are indicated. The leftmost and rightmost anchor genes for the four human MHC-homologous blocks are denoted.

NOD-Like Receptors (NLRs).

NLR genes encode intracellular receptors that sense diverse pathogenic triggers and activate various innate immune responses. Their encoded proteins are characterized by a tripartite domain architecture, including a highly conserved central NACHT domain fused with variable N-terminal (e.g., CARD, Death, DED, HEPN_DZIP3) and C-terminal domains (e.g., LRR, TPR). Cephalochordates have 55 to 76 NACHT-encoding genes—a substantial expansion compared to most vertebrates (e.g., 25 in humans, 8 in chicken) (SI Appendix, Fig. S26 and Datasets S5 and S7). Like TLRs, tandem duplication accounts for 25.5 to 43.4% of the NLR gene repertoire across cephalochordate species (Fig. 6B and SI Appendix, Fig. S25). Notable examples include clusters of NLRs on chromosome 5 in both B. belcheri (containing 19 NLR genes within a 452.14-kb region) and B. floridae (containing 18 NLR genes within a 307.52-kb region). Our phylogenetic analysis of NLRs from diverse metazoan lineages suggests a strong pattern of lineage-specific expansion, with NLRs from species of the same (sub-)phylum predominantly clustered together (SI Appendix, Fig. S26 and Dataset S7). Notably, echinoderms and cnidarians have extensively expanded NLR genes (Dataset S5). In some protostomes and all cephalochordates, additional diversity of NLR genes was generated by fusions between some N-terminal domains (e.g., Death, DED, HEPN_DZIP3) and the central NACHT domain. The absence of such fusions in urochordates and vertebrates suggests lineage-specific loss of these NLR genes in both groups after their split from cephalochordates. In contrast, the domain combinations of BIR-NACHT, PYRIN-NACHT, and FISNA-NACHT are unique to vertebrates, representing vertebrate-specific innovations.

The Proto-MHC Locus.

The MHC locus in jawed vertebrates harbors genes encoding a wide variety of cell surface molecules that mediate antigen recognition and adaptive immune response (54). Its prototypical form (i.e., the proto-MHC) has an ancient metazoan origin, while the canonical MHC class I/II genes and rearranging antigen receptor (AgR) genes are jawed-vertebrate-specific innovations (55). Previous studies characterized some MHC-specific anchor genes in B. floridae based on fragmented genomic information and proposed a proto-MHC locus homologous to four human MHC paralogous blocks (18, 52). Here, we revisited the proto-MHC evolution by examining orthologs of human MHC-locus-containing genes in all five chromosome-level cephalochordate genome assemblies. A cephalochordate-human genome comparison revealed strong gene orthology correspondence between a single cephalochordate chromosome (e.g., the A. lucayanum chromosome 7) and four human chromosomes (chromosomes 1, 6, 9, 19), with hits on human chromosome 6 corresponding to the human MHC locus (SI Appendix, Figs. S27–S31). In the A. lucayanum-human comparison, we found a total of 432 genes on A. lucayanum chromosome 7 displaying orthologous correspondence with genes on human chromosomes 1, 6, 9, and 19. Among them, 55 genes are orthologous to genes in the human MHC locus on human chromosome 6 (n = 47) or to their close paralogs on human chromosomes 1 (n = 11), 9 (n = 11), and 19 (n = 5) (Fig. 6C). These 55 genes are scattered across the entire chromosome 7 in A. lucayanum, in sharp contrast to the much more concentrated distribution of their orthologs on human chromosomes: 1q23.3–1q25.3 [HSPA6, RNF2], 6p21.33–6p21.31 [GNL1, FKBP5], 9q33.2–9q34.3 [DAB2IP, EHMT1], and 19p13.2–19p13.11 [SLC44A2, PBX4]. Also, rather than being strongly colocalized within a narrow genomic locus like MHCs in jawed vertebrates, the cephalochordate orthologs of the 55 human genes in the MHC locus underwent substantial species-specific duplication and order reshuffling (e.g., SKIC2-like genes in A. lucayanum and PSMB7/10-like genes in B. floridae) (SI Appendix, Fig. S32). No orthologs of the canonical human MHC Class I and Class II genes such as HLA-A/B/C/DR/DP/DQ, or local gene linkages corresponding to human MHC ClassI-III-II partitions could be detected in the cephalochordate chromosomes carrying the proto-MHC. Many genes in the proto-MHC show only tenuous or indirect associations with immune function.

Proto-RAGs.

The extreme diversity of immunoglobulins and T cell–antigen receptors involved in adaptive immunity of jawed vertebrates is achieved by the RAG1 and RAG2 proteins via V(D)J somatic recombination (56). Previous studies suggested RAG1 and RAG2 apparently evolved from an ancestral transposable element containing both sequences (57). Such a proto-RAG element and its encoded RAG1-like (RAG1L) and RAG2-like (RAG2L) genes bear strong sequence similarity with the vertebrate RAG1/2 genes but do not show recombinase activity in vivo nor carry any known immune-related functions in nonvertebrate animals. The recent discoveries of intact proto-RAG transposons in B. belcheri and B. lanceolatum have suggested cephalochordate RAG1/2-like genes as the closest common ancestors for vertebrate RAG1/2 genes (21); however, there is phylogenetic evidence indicating that hemichordate RAG1Ls are seemingly more closely related to vertebrate RAG1s (58). To settle this debate, we made a genome-wide scan of RAG1L in all five cephalochordate species based on both genome assemblies and annotations. We found relatively complete proto-RAGs (RAG1L+RAG2L) in B. belcheri (three copies) and B. lanceolatum [three copies; consistent with a previous report (12)] as well as an additional RAG1L fragment in B. japonicum (SI Appendix, Table S15 and S16). Neither RAG1L nor RAG2L was found in A. lucayanum or B. floridae. We also extended our survey to several representative species of vertebrates, urochordates, hemichordates, and echinoderms. As expected, we found unambiguous copies of RAG1 and RAG2 in all surveyed jawed vertebrates (human, chicken, spotted gar and coelacanth, and elephant shark) but none in jawless vertebrates (lamprey and hagfish). In hemichordates and echinoderms, more RAG1L and RAG2L were found but with substantially varied species-specific distributions. By combining all RAG1(L) and RAG2(L) genes identified in our survey and a recent study (58), we reevaluated their evolutionary history in deuterostomes (SI Appendix, Fig. S33). As previously noted (58), we found that orthologs of both RAG1(L) and RAG2(L) in cephalochordates are tightly clustered together with greater similarity to the RAGL-B family members of hemichordates and echinoderms. The vertebrate RAGs on the other hand are most closely related to the hemichordate and echinoderm RAGL-A family members. The RAGL-A family members from the hemichordate P. flava appear as the most closely related nonvertebrate ortholog to vertebrate RAGs. Taken together, our study with an expanded set of cephalochordate RAG1/2Ls provides additional support for a nonvertical transmission origin (potentially from hemichordates) of the RAGL-A gene in the ancient common ancestor of jawed vertebrates (58), although the mechanistic details on how this process could have occurred awaits future investigation.

Discussion

In the present study, we constructed a chromosome-level genome assembly for an early-diverging cephalochordate species A. lucayanum, which reveals a notable genome size expansion primarily driven by TEs, in comparison to the genomes of cephalochordate species from the Branchiostoma genus. This observation mirrors similar findings in other animals with massively expanded genomes (e.g., Mexican axolotl, lungfish, and Antarctic krill) (59–61). Interestingly, while “copy-and-paste” retrotransposons are more prevalent in those expanded vertebrate genomes (e.g., for axolotl and lungfish), the “cut-and-paste” DNA transposons are major contributors to the expanded nonvertebrate genomes, such as those of Asymmetron and Antarctic krill.

Although the expansion of repetitive sequences (such as TEs) frequently promotes chromosome rearrangements, both macrosynteny and microsynteny remain highly conserved between Asymmetron and Branchiostoma. This stability enabled us to reconstruct the ancestral cephalochordate genome architecture as 20 ALGs, a configuration similar to that of the extant B. belcheri genome (13). From this ancestral state, we infer interchromosomal fusion events as the major force in shaping the current cephalochordate karyotypes. These interchromosomal fusion sites are characterized with a high abundance of segmental duplications, often coupled with a higher background of repetitive sequences. These features echo the previous observation around the human chromosome 2 fusion site (2q13–2q14.1) (62), suggesting a potentially common mechanism underlying interchromosome fusions. The presence of many microsynteny blocks across cephalochordate species raises the question of the evolutionary constraints on the conservation of gene order. By combining comparative genomic and transcriptomic analyses, we showed that, as in other metazoan lineages (63, 64), constraints on cotranscriptional regulation likely contributed to the maintenance of conserved microsynteny blocks. In addition, we observed widespread genome sequence conservation across all five cephalochordates species not only in CDSs and UTRs but also in many intronic and intergenic regions, which suggests a universally slow rate of genomic evolution in cephalochordates. Some of these pan-cephalochordate conserved genomic blocks are also highly conserved with vertebrate genomes at both synteny and sequence levels, indicating strong constraints shaping chordate genome architecture.

In contrast, genomes of the third chordate subphylum (Urochordata/Tunicata) are rapidly evolving (65), and thus omitted from our analysis. Unlike the regulative development of cephalochordates and vertebrates, where cell–cell communication is vital and early mutations are often lethal (66), urochordates have evolved mosaic development with very early determination of cell fates (67) and high tolerance for early gene knockouts (68). Such early decision of cell fates potentially lowered the evolutionary constraints of genes expressed later in urochordate development, resulting in a highly derived genome content and architecture.

Our analysis of the Hox cluster suggests that a single, colinear cluster of 15 Hox genes is ancestral to cephalochordates. Interestingly, the Hox clusters of both A. lucayanum and B. belcheri are inverted relative to those of B. floridae, B. japonicum, and B. lanceolatum. Both A. lucayanum and B. belcheri exhibit TE accumulations within and flanking their Hox clusters, hinting that TE expansion may be associated with these inversions. The distinct inversion breakpoints and TE types associated with their Hox clusters collectively suggest that the Hox inversions in these two species likely occurred independently. While the current Hi-C data for A. lucayanum lack sufficient resolution to define local 3D genome interactions around its Hox cluster, Hi-C data from B. belcheri show that the boundaries of the two TADs in the Hox cluster coincide with the Hox inversion breakpoints. In vertebrates (e.g., mouse), the TADs around the Hox gene clusters function in segregating regulatory elements to ensure proper embryonic gene expression (69). We propose that the TADs around the cephalochordate Hox cluster may similarly contribute to the regulation of coordinated expressions of Hox genes, imposing strong evolutionary constraints maintaining the “en-bloc” configuration despite TE invasion. Also, the invasion of TEs into Hox clusters has been previously associated with species radiation and morphological diversification in teleost fish and reptiles (70, 71). Therefore, the observed TE invasion in A. lucayanum and B. belcheri may have subtly altered the expression pattern of nearby Hox genes and thereby contributed to morphological differences such as the variation in myotome numbers among different Branchiostoma species and the distinct caudal fin shapes of Asymmetron and Branchiostoma (72).

As the early-branching chordate lineage that lacks adaptive immunity, cephalochordates are important for reconstructing the origins of the vertebrate immune system. In this context, we traced the evolutionary history of key immune-related genes (i.e., TLR, NLR, and RAGs) and the MHC locus in cephalochordates and beyond. Similar to Branchiostoma species (4, 73), A. lucayanum exhibits lineage-specific expansions of innate immune receptor genes (e.g., TLRs and NLRs), likely reflecting an adaptation to the microorganism-rich environment of shallow coastal waters. Additionally, we thoroughly examined the genomic and evolutionary patterns of the prototypical forms of the vertebrate MHC locus and RAG genes in cephalochordates. In contrast to the previously proposed en-bloc proto-MHC configuration for cephalochordates (18, 52), we found that proto-MHC-defining genes in both Asymmetron and Branchiostoma cephalochordates are widely distributed across the MHC-corresponding chromosomes. Therefore, the compaction of the MHC locus must have occurred after the separation of cephalochordates and vertebrates. The recent characterization of the compact MHC locus and nonrearranging AgR genes in cartilaginous fish provides an additional reference point for the emergence of adaptive immunity (74, 75). The precise selection scheme shaping this specialized genomic condensation remains to be discovered.

In conclusion, by decoding the genome sequence of the earliest branching cephalochordate, Asymmetron, and comparing it to those of four Branchiostoma species, we have confirmed that the subphylum Cephalochordata is evolving exceptionally slowly. Therefore, modern cephalochordates are ideal models for tracing the chordate ancestor and investigating the origin of vertebrates. Moreover, not only are macro- and micro-synteny highly conserved among cephalochordates from different genera, but synteny also remains comparable between cephalochordates and vertebrates, despite the 2R-WGDs that vertebrates have undergone. Notably, the Asymmetron genome is larger than those of Branchiostoma species, a difference chiefly driven by the accumulation of TEs. Interestingly, extensive TE invasion within and flanking the Hox cluster of A. lucayanum and B. belcheri likely contributed to the independent inversions of their Hox clusters relative to other Branchiostoma species. Taken together, these genomic features of cephalochordates provide a clearer picture of how complex vertebrates evolved from much simpler ancestral chordates, and offer valuable resources for investigating the genomic modifications underlying morphological variation across cephalochordates.

Materials and Methods

Genome Sequencing and Assembly.

Specimens of A. lucayanum were collected in Bimini, Bahamas as previously described (24). The sperm of a single male animal was used for DNA extraction and PacBio sequencing. De novo genome assembly was primarily performed by Canu (76) with additional haplotype decoupling and chromosome-level elevation. See SI Appendix, SI Materials and Methods for details.

Transcriptome Sequencing and Data Analysis.

A. lucayanum’s total RNA samples isolated from multiple developmental stages were mixed for PacBio’s Iso-Seq. In addition, multistage total RNA of A. lucayanum, B. belcheri, and B. japonicum was isolated separately for Illumina-based bulk RNA-seq. The naming of these developmental stages follows an updated staging system (77). The Iso-Seq data were used for gene annotation, while the Illumina-based bulk RNA-seq data were used for gene expression quantification together with similar datasets from published studies (SI Appendix, Tables S1 and S5 and Dataset S8). See SI Appendix, SI Materials and Methods for details.

Repeat Element and Gene Annotation.

We employed EDTA (78) and FunAnnotate (https://github.com/nextgenusfs/funannotate) to perform the annotation of repeat element, protein-coding genes, and tRNA genes for all examined amphioxus species (SI Appendix, Tables S3 and S9). See SI Appendix, SI Materials and Methods for details.

Phylogenetic Analysis and Molecular Dating for the Species Tree.

We used OrthoFinder (79) to identify orthologous groups shared by the five cephalochordates as well as other representative bilaterians (SI Appendix, Table S9). Species tree construction and molecular dating were performed using 947 strictly conserved one-to-one orthologs. This phylogenetic framework was then used to analyze lineage-specific protein evolution, cephalochordate-specific mutation rate, and LTR-TE insertion timing. See SI Appendix, SI Materials and Methods for details.

Macrosynteny and Microsynteny.

Macrosynteny and microsynteny conservation indices were defined as the proportion of orthologs located on homologous chromosomes and within microsynteny blocks, respectively. Using A. lucayanum as a reference, we identified a stringent set of conserved microsynteny blocks shared across all five amphioxus species (i.e., pan-cephalochordate microsynteny blocks). See SI Appendix, SI Materials and Methods for details.

Hox Cluster Analysis.

The Hox clusters of the five cephalochordate species were identified based on the flanking anchor genes (e.g., MEOX2, MTX2, and SMUG1). For each individual Hox gene, we construct its full-length CDS and protein alignments as described previously (80). Based on these alignments, molecular evolution and phylogenetic analysis were further performed. See SI Appendix, SI Materials and Methods for details.

Identification and Functional Characterization of Conserved Bases and Regions.

Multiway whole-genome alignments were generated using Cactus (81) for the five cephalochordates and six representative vertebrates respectively. Those fully aligned bases with phyloP scores >0 were defined as conserved bases for the corresponding group. The chordate conserved bases were further defined by intersecting the cephalochordate and vertebrate conserved bases based on the Asymmetron-human whole-genome alignment. The cephalochordate conserved bases were merged into pan-cephalochordate conserved regions and further compared with B. lanceolatum open chromatin peaks (10) and known B. belcheri and B. floridae microRNA genes (43). The chordate conserved bases were merged into chordate ultra-conserved regions (chordate UCRs) and subjected to functional enrichment analyses with: 1) Gene Ontology terms, 2) human pathogenic variants (47), 3) human genes with high clinical significance (48, 49). See SI Appendix, SI Materials and Methods for details.

Genome-Wide Survey of Immune-Related Genes.

A comprehensive genome-wide survey for TLR, NLR, and RAG(-like) genes as well as those genes defining the MHC locus was performed across multiple representative metazoan lineages (SI Appendix, Table S9). See SI Appendix, SI Materials and Methods for details.

Supplementary Material

Appendix 01 (PDF)

Dataset S01 (XLSX)

Dataset S02 (XLSX)

Dataset S03 (PDF)

Dataset S04 (XLSX)

Dataset S05 (XLSX)

Dataset S06 (PDF)

Dataset S07 (PDF)

Dataset S08 (XLSX)

Data, Materials, and Software Availability

Raw reads, assemblies, and analysis files and code data have been deposited in Raw reads: NCBI BioProject, Assemblies: National Genomics Data Center, Zenodo, and Analysis files and code: Zenodo [NCBI BioProject: PRJNA1130632 (82); National Genomics Data Center: GWHFWAS00000000.1 (83) and GWHFWAV00000000.1 (84); and Zenodo: 10.5281/zenodo.15280774 (85)]. Other data are included in the article and/or supporting information.