Recurrent evolution of ligand-binding domain multiplicity fine-tunes TGFβ signaling in vertebrates

Jatzlau, Jerome; Trumpp, Michael; Kühlwein, Julia; Obendorf, Leon; Le, Yao; Kuhl, Heiner; Preussner, Marco; Mendez, Paul-Lennard; Burkert, Hendrik; Burdzinski, Wiktor; Mundlos, Stefan; Winkler, Christoph; Stöck, Matthias; Knaus, Petra

doi:10.1038/s41467-026-73340-5

Download PDF

Article
Open access
Published: 19 May 2026

Recurrent evolution of ligand-binding domain multiplicity fine-tunes TGFβ signaling in vertebrates

Nature Communications volume 17, Article number: 4458 (2026) Cite this article

144 Accesses
Metrics details

Subjects

Abstract

From sponges to mammals, TGFβ signalling is a central regulator of body plan, cell fate and tissue homeostasis, with receptor architecture highly conserved across metazoans. Here we identify unexpected evolutionary divergence within a key receptor structure: 12 independent ligand-binding domain (LBD) duplications across three receptor classes (ACVR1, BMPR2 and TGFBR2) in diverse vertebrate lineages, including fish, amphibians, birds, and mammals. These duplications reveal previously unrecognized structural and functional plasticity arising from domain-level innovation, including in established model organisms such as zebrafish, African clawed frog and chicken. Recently diverged lineages conserve the membrane-distal LBD and ligand-interacting residues, correlating with enhanced ligand binding, whereas more ancient lineages exhibit elevated evolutionary rates of the membrane-distal LBD associated with inhibitory function. Our findings reveal LBD multimerization as a recurring, lineage-independent mechanism that diversifies and fine-tunes TGFβ signalling, adding a regulatory dimension to one of the best-examined conserved and essential pathways in metazoan biology.

Development of a zebrafish model of Loeys–Dietz syndrome through tgfbr2b knockdown

Article 03 February 2026

Shaping the composition of the mitochondrial outer membrane

Article 16 June 2025

Comparative multi-omic analysis reveals conserved and derived mechanisms of fin and limb regeneration

Article Open access 22 January 2026

Introduction

Signaling through ligand-binding receptors has been a cornerstone of multicellular life since the early evolution of metazoans¹. Multiplication of single exon-coded domains shaped the extracellular architectures of many growth factor receptors. Some, such as EGF-, FGF-, and VEGF receptors, contain repeated ligand-binding domains (LBDs), yet the number and arrangement of these repeats have remained conserved from early metazoans to vertebrates^2,3,4,5. By contrast, other key ligand-binding receptors such as those of the TGFβ receptor family have been supposed to contain only a single LBD, conserved from invertebrates to mammals. Remarkably, we found an exception in this otherwise highly conserved pathway: in medaka (Oryzias latipes), a single exon encoding the ligand-binding domain in ACVR1 underwent triplication, producing a receptor with three tandem LBDs⁶.

The TGFβ-receptor family, comprising TGFβ and bone morphogenetic protein (BMP) receptors, represents a highly conserved signaling system, essential for regulating developmental and homeostatic processes across all vertebrates⁷. Their roles span from orchestrating early embryonic tissue formation to fine-tuning cellular identities across diverse cell lineages^8,9 These receptors share a characteristic organization, an extracellular LBD followed by a transmembrane (TM) and an intracellular kinase domain (KD)^10,11, that is preserved from invertebrates to vertebrates.

Following two rounds of whole-genome duplication (WGD) events, the number of TGFβ receptor genes expanded from five in invertebrates¹² to thirteen in vertebrates¹³, of which twelve are retained in mammals, including five type II and seven type I receptors¹⁴. This expanded receptor repertoire co-evolved with more than thirty TGFβ ligand family members, forming highly specific ligand–receptor pairs¹⁴. Upon ligand binding to their high-affinity receptors, a tetrameric complex of two type I and two type II receptors assembles, enabling transphosphorylation and activation of the type I receptor kinase by a constitutively active type II kinase. Activated type I receptors phosphorylate intracellular mediators, most commonly SMAD transcription factors, which regulate the expression of TGFβ/BMP target genes in a cell type- and context-dependent manner to control key developmental decisions⁸.

In this study we investigated whether LBD multiplications within the TGFβ receptor family are more widespread as previously recognized and how such structural innovations affect receptor signaling. Through a comprehensive analysis across vertebrate genomes, we identified 12 independent cases of convergent LBD multiplication in three TGFβ receptor genes, spanning a broad phylogenetic range from fish to mammals. These events were detected in multiple species, including elephant-nose fish, carps, clawed frogs, chicken and horses, highlighting that LBD multimerization is a recurrent evolutionary phenomenon across vertebrate lineages. Our findings reveal that these can substantially modulate receptor signaling properties. Thus, LBD multiplication represents a previously unknown mode of molecular innovation within the highly conserved TGFβ signaling system, adding a new layer of complexity and regulatory potential to this essential developmental pathway.

Results

WGD-independent LBD expansion in TGFβ receptors

Previously, we identified and verified that the medaka ACVR1 paralogs, derived from teleost-specific whole genome duplication, differ in their number of LBDs⁶. Whereas the medaka ACVR1 carries three LBDs, the ACVR1L gene encodes for a BMP type I receptor with only 1 LBD. In contrast, zebrafish (Danio rerio) has lost the ACVR1 gene and retained only ACVR1L with 1 LBD. To trace the origin of the ACVR1 LBD triplication, we compared available genome and transcriptome data across various ray-finned fish (Actinopterygii) lineages, supplementing missing data with RNA sequencing of heart tissues from selected representatives of individual families (Fig. 1). We found that an initial duplication of the ACVR1 LBD occurred in the early ancestors of Osteoglossocephalai, followed by a secondary triplication event common to all Clupeocephala. Interestingly, individual LBDs were subsequently lost within herrings, anchovies and shads (Clupeoidei), and the entire ACVR1 gene was lost in all carp-like fishes (Cyprinoidei). We next asked whether LBD multiplication is specific to the ACVR1 locus, or if it is a common feature across all teleost genomes following evolutionary adaptation after the teleost whole genome duplication event¹⁵. Surprisingly, we found an LBD triplication event in the ACVR1 paralog ACVR1L, which is unique to the bonytongues (Osteoglossidae). We uncovered six additional LBD multiplication events in the type II receptor genes BMPR2 and TGFBR2, indicating a broader pattern beyond the ACVR1 locus. Notably, the BMPR2 gene underwent LBD triplication twice, once within the bichirs (Polypteridae) and once the featherbacks and knifefishes (Notopteroidei). These events appear to be independent of BMPR2 paralogy, as Polypteridae diverged before the teleost whole genome duplication. The TGFBR2 locus underwent four independent LBD multiplication events in Clupeoidei, Cyprinoidei, pikes and mudminnows (Esociformes), and salmonids (Salmoninae). While the carp-specific TGFBR2 LBD duplication preceded the carp-specific whole genome duplication (WGD) (Supplementary Fig. 1a), the TGFBR2 LBD multimerization in Salmoninae occurred after the salmon-specific WGD. Finally, we extended our analysis to all jawed vertebrates (Gnathostomata) (Fig. 2) to test whether TGFβ-superfamily receptor LBD multiplication depends on preceding WGD. We identified four additional TGFBR2 LBD multiplication events: one in the TGFBR2 paralog TGFBR2L within the african lungfish (Protopteridae), and three in TGFBR2 within amphibians (Amphibia), chicken (Gallus)¹⁶, and odd-toed hoofed mammals (Perissodactyla). Collectively, we identified 12 convergent LBD multiplication events across three TGFβ-family receptor genes, ACVR1/ACVR1L, BMPR2, and TGFBR2/TGFBR2L, in species that had undergone WGD and retained a backup paralog (e.g., ACVR1 in all teleosts, or TGFBR2 in Salmoninae). However, we also observed LBD multiplication in species with only a single gene copy, such as BMPR2 in Polypteridae and TGFBR2 in Perissodactyla, suggesting that WGD is not a prerequisite for LBD multimerization. Interestingly, type I receptor LBDs are encoded by a single exon, whereas type II receptor LBDs are split across two exons, highlighting that, in the case of BMPR2 and TGFBR2, duplication of an entire exon pair was required (Supplementary Fig. 2a). Since the TGFBR2 locus exhibited the highest number of LBD duplication events, we next investigated whether repetitive elements flanking exons 2 and 3 of TGFBR2 are conserved across species, potentially facilitating repeated independent duplications. Whereas in the zebrafish Tgfbr2a locus, retroelement derived GT-rich sequences are framing the duplicated exon pair, these or other repetitive sequences are absent in the chicken (Gallus gallus) and horse (Equus caballus) TGFBR2 loci at duplication sites (Supplementary Fig. 2b). The duplication sizes range widely, from ≥2.1 kb in zebrafish to ≥18.7 kb in horse, indicating that these exon-pair duplications arose through separate genetic events rather than from a shared ancestral duplication (Supplementary Fig. 2b).

**Fig. 1: Convergent evolution of TGFβ & BMP-Receptor LBD multiplication in ray-finned fish.**

**Fig. 2: Convergent LBD multiplication in TGFBR2 orthologs of jawed vertebrates.**

Membrane-proximal LBD drives gnBmpr2a signaling

To assess the molecular function of convergently multimerized LBDs, we first analysed the sequence conservation among 3 LBDs within individual BMPR2 proteins of the Notopteroidei and Polypteridae lineages (Fig. 3a, and Supplementary Fig. 3). In both groups, the LBDs exhibited intermediate levels of sequence identity (50-70%) relative to the membrane-proximal LBD (LBDⁱⁿ), suggesting partial functional divergence within the repeats (Fig. 3a). BMPR2 functions as a type II receptor for a wide range of BMP and Activin ligands^17,18. Whereas BMP binding to BMPR2-containing receptor complexes requires the presence of a high affinity type I receptor, BMPR2 can directly bind Activin A through a specific set of hydrogen bonds and hydrophobic interactions primarily involving residues located between fingers 2 and 3 (Fig. 3b, c)¹⁹.

**Fig. 3: Inner LBD conveys signaling competence of LBD multimerized BMPR2 ortholog.**

To determine the conservation of these critical interaction sites, we compared the BMPR2 LBD residues involved in ligand binding across the 3 domains. In both lineages, the inner LBD showed the highest degree of sequence conservation with human BMPR2, particularly at residues forming the Activin-specific binding interface (Fig. 3c). Interestingly, the inner LBDs, of Notopteroidei species retained 85% of the key residues involved in A-loop contacts in finger 2, while Polypteridae retained 69% (Fig. 3c, d). Consistent with this, members of the Notopteroidei also exhibited minimal loss of conserved interaction-mediating residues on finger 3, consistent with a maintained or partially preserved capacity of Activin binding in LBDⁱⁿ (Fig. 3d, and Supplementary Fig. 4a).

To explore the functional impact of LBD triplication in BMPR2 receptors from Notopteroidei, we performed computational protein-protein docking using the Rosetta Commons suite²⁰. Docking simulations were conducted using the triplicated LBD of BMPR2a from elephant-nose fish (Gnathonemus petersii; member of the Notopteroidei), and human Activin A as the ligand (Fig. 3e, and Supplementary Fig. 4b). Additionally, lineage-specific docking analyses were performed using the appropriate Activin A orthologs for both Notopteroidei and Polypteridae species (Supplementary Fig. 4e–g). Consistent with the LBD conservation analysis (Fig. 3a–c), the inner LBD showed the lowest predicted interface energy, indicating the highest ligand-binding affinity (Fig. 3e). To experimentally test the functionality of the individual LBDs, we generated expression constructs of elephant-nose fish BMPR2a, including full-length receptors, deletion mutants containing only one or two LBD (LBD^mid-in). Binding of fluorescently-labelled human Activin A-Cy5 was assed using ligand surface binding assay (LSBA, Fig. 3f; and Supplementary Figs. 4c, 5)²¹, and downstream signaling was quantified via a CAGA₁₂-driven dual luciferase assay, which reports SMAD2/3 activity (Fig. 3g; and Supplementary Fig. 4d).

In agreement with the docking predictions, only the inner, membrane-proximal LBD bound Activin A with high affinity. Neither the outer, nor the middle LBD alone showed detectable binding, while LBDⁱⁿ and LBD^mid-in exhibited comparable binding to the full-length receptor (Fig. 3f). This binding pattern was mirrored in the functional signaling output: constructs containing only the outer or middle LBD showed minimal SMAD2/3 activation, whereas LBDⁱⁿ displayed the highest signaling competence (Fig. 3g). Notably, deletion of the outer and middle LBDs enhanced signaling efficiency, indicating that these LBDs are functional but act repressively, whereas only the membrane-proximal inner LBD contributes positively to Activin A responsiveness in the LBD triplicated receptor (Fig. 3g). Interestingly, the BMPR2 ortholog from elephant fish enhanced both Activin A binding and SMAD2/3 signaling, whereas the human BMPR2 control increased ligand binding but failed to efficiently promote SMAD2/3 activation. This functional divergence suggests that the elephant fish BMPR2 exhibits ACVR2-like receptor behavior²².

Divergent evolution of TGFBR2 LBD multimers tunes signaling

Next, we examined the TGFBR2 gene, which independently acquired LBD multiplications in seven distinct lineages, followed by additional expansions in salmonids and amphibians, as seen in atlantic salmon (Salmo salar; 3 LBDs) and lake trout (Salvelinus namaycush; 5 LBDs), gaboon caecilian (Geotrypetes seraphini; 3 LBDs), and african clawed frog (Xenopus laevis; 3 LBDs) (Figs. 1, 2). To predict the molecular function of the multimerized LBD receptors, we first compared the LBD sequence homology within one protein (Fig. 4a, and Supplementary Fig. 6). We observed that the degree of conservation varied among the seven families. While the carp (Cyprinus carpio) and clawed frog TGFBR2 LBDs showed the lowest conservation (25–50%) relative to the membrane-proximal LBD (LBDⁱⁿ), pike (Esox lucius) and herring (Clupea harengus) exhibited intermediate conservation (50–90%), and the salmons, chicken, and horse displayed high conservation levels (90–100%). (Fig. 4a). TGFBR2 is known for its high affinity binding to TGFβ-ligands via polar and hydrophobic interactions of residues mainly residing in finger 1 of the LBD (Fig. 4b, c)²³. When comparing the amino acids, which form the interaction epitope for human TGFβ1 within TGFBR2, these are highly conserved in the LBDⁱⁿ domain of all tested species, whereas the middle or outer LBD domains show a lower degree of conservation (Fig. 4c, and Supplementary Fig. 7). Interestingly, consistent with its overall low degree of conservation, the zebrafish Tgfbr2a LBD^out shows no conserved TGFβ1-binding amino acid residues (Fig. 4c, d). In contrast, the chicken LBDs exhibit 98% overall similarity and share an identical TGFβ1-binding epitope (Fig. 4c, d). Lastly, the horse and tapir LBD^out show a high level of overall conservation (93.1%, 96%), with three or one amino acid substitutions within the TGFβ1-binding epitope, respectively (Fig. 4c, d; and Supplementary Fig. 7). In contrast, to the varying degree of conservation on the side of the LBD interface, all TGFβ1 ligand orthologs show a complete conservation of amino acids forming the receptor interacting interface (Supplementary Fig. 7). Finally, to evaluate whether the extent of duplicated LBD sequence divergence corresponds to the evolutionary time separating each lineage, we calculated molecular clocks using orthologous LBD domains. The resulting lineage-specific clocks indicated that the LBD duplications in horse, chicken, and salmon evolved at rates consistent with their respective orthologous LBDs, indicating that these domains remain highly conserved under a clock-like model of evolution. However, the zebrafish LBD duplication strongly deviated from this pattern, exhibiting an exceptionally high substitution rate, far exceeding that of the clade-specific (otomorpha) LBD molecular clock, indicative of accelerated evolution (Supplementary Fig. 8).

**Fig. 4: Multimerized LBDs of TGFBR2 orthologs display varying degree of TGFβ1-interface conservation.**

To functionally predict the effect of TGFBR2 LBD duplication on ligand binding capabilities we again performed protein-protein docking studies (Fig. 5a) using human TGFβ1 and the respective animal ortholog TGFβ1, which exhibit a high degree of conservation (Fig. 5, and Supplementary Figs. 9,10,11). Furthermore, we cloned zebrafish, carp, clawed frog, chicken and horse TGFBR2 expression constructs, and generated single LBD mutants, to separately test the functionality of individual LBDs. Using fluorescently labeled human TGFβ1-SiRd12, we performed LSBA²¹ to verify the predicted docking data (Fig. 5b) and tested the downstream signaling response using the SMAD2/3-sensitive CAGA₁₂-dual luciferase assay (Fig. 5c). In line with the low degree of LBD conservation in both zebrafish and carp Tgfbr2a, the LBD^out exhibits almost no TGFβ1 binding, whereas the LBDⁱⁿ domain is capable of TGFβ1 binding and transmitting signals (Fig. 5d–f, Supplementary Figs. 9a–c, 12, 13). Strikingly, only in the zebrafish

**Fig. 5: TGFBR2 LBD duplications differently finetune signaling competence.**

(Figure 5e, Supplementary Fig. 12), and reduced SMAD2/3 activity as seen in the reporter gene assay (Fig. 5f).

Similarly, in horse and frog TGFBR2 orthologs, the membrane-proximal domain LBDⁱⁿ shows the highest predicted and experimentally confirmed binding to TGFβ1, while the outer domain (LBD^out) exhibits reduced or no binding and lacks signaling competence (Fig. 5g–i, Supplementary Figs. 9d–f, 14, 15). Notably, although most key TGFBR2 human TGFβ1-contact residues are conserved within the LBDⁱⁿ domains, the frog LBDⁱⁿ contains a non-conservative substitution at position 119 (E119L) within the C-terminal segment previously implicated in ligand interaction (Fig. 4c)²³. Despite this substitution, the LBDⁱⁿ domain facilitates ligand binding, suggesting that the overall binding interface remains functionally preserved, potentially through compensatory interactions or structural tolerance within the receptor–ligand complex. Importantly, the presence of the LBD^out does not impair binding or signaling function of the full length TGFBR2 for either species (Figs. 5h, i, and Supplementary Figs. 9e, f, 14, 15).

In contrast, chicken TGFBR2 shows a different pattern. Both LBDⁱⁿ and LBD^out share a conserved TGFβ1 binding epitope and 98% sequence identity, resulting in similar predicted and measured TGFβ1 binding. Each domain alone is sufficient to support efficient ligand binding and downstream signaling (Fig. 5j–l, Supplementary Fig. 16). Interestingly, the native chicken receptor containing both LBDs binds TGFβ1 more strongly than either single domain receptor although signaling output remains comparable regardless of the number of LBDs (Fig. 5k–l).

After characterizing the differential ligand-binding capacity of LBD-duplicated TGFBR2 variants, we sought to predict the impact of linker length between the two LBDs. While two identical LBDs may enable inter-ligand binding, thereby increasing total ligand engagement, a flexible linker of sufficient length could permit intra-ligand bivalency. In this scenario, two LBDs within a single receptor would occupy both type II receptor-binding sites, potentially interfering with the formation of a functional tetrameric receptor complex (Fig. 6a, and Supplementary Fig. 17a, b). To estimate the linker length required for intra-ligand bivalency, we modeled human TGFBR2 variants containing two identical LBDs connected by linkers of increasing length, starting from the 20-residue linker encoded at the E3–E2′ junction. Linkers of ≥36 residues were predicted to represent the minimal requirement for intra-ligand bivalency (Fig. 6b). Comparison with naturally occurring linker lengths across TGFBR2 variants revealed that most species possess linkers shorter than 24 residues, with the exception of C. harengus (44 residues) (Fig. 6c, and Supplementary Fig. 17c). While short linkers may permit inter-ligand binding, intra-ligand bivalency is therefore unlikely in most species. Notably, the membrane-distal LBD in C. harengus lacks key residues of the TGFβ-binding interface, consistent with reduced binding affinity (Supplementary Figs. 7, 11a), further arguing against intra-ligand bivalency.

**Fig. 6: Human TGFBR2 LBD duplication enhances ligand binding without affecting downstream signaling.**

Next, we engineered a synthetic human TGFBR2 variant with two identical LBDs which would result from E2-E3 duplication and assessed its ability to bind human TGFβ1. Consistent with the chicken receptor, LBD duplication significantly enhanced ligand binding, confirming the inter-ligand binding mode (Fig. 6d, e, and Supplementary Fig. 18). Whereas CAGA₁₂-luc activity after 24 h was comparable for both TGFBR2 variants (Fig. 6f), short term activation of SMAD2 was reduced in the presence of TGFBR2 2 LBD compared to TGFBR2 expression (Fig. 6g). Collectively, this highlights that 2 LBD TGFBR2 receptors remain functional but limit the downstream SMAD activation through excessive ligand binding (Fig. 6h).

These findings suggest that LBD duplication in TGFBR2 is a recurrent evolutionary mechanism capable of fine-tuning ligand binding without compromising signaling output, thereby offering a potential selective advantage under specific physiological or developmental contexts.

LBD duplication generates low-signaling dTgfbr2a in zebrafish

Notably, TGFBR2 two-LBD variants from different species exhibited different signaling capabilities, which raises the question how these integrate into the critical fine-tuning of balanced BMP and TGFβ signaling in the respective species. To answer this, we analyzed publicly available RNA-seq datasets to screen for tissue-specific expression patterns in chicken, horse and zebrafish, which were characterized by different TGFBR2 splice variants (Supplementary Fig. 19a). In both chicken and horse, the TGFBR2 two-LBD variant was highly expressed in the cardiovascular and reproductive systems (Supplementary Fig. 19b–d). Interestingly, the high degree of amino acid homology between both LBDs is equally reflected in the coding sequence of exon 3 and exon 3’ for both TGFBR2 ortholog variants (Supplementary Fig. 19a). Exon duplication events have been associated with alternative splicing in multiple gene families, e.g., immunoglobulin superfamily members and tropomyosin²⁴. Indeed, in both chicken and horse, we observed single LBD variants due to alternative splicing but at a marginal rate, indicating that the TGFBR2 two-LBD variant is the predominant receptor (Supplementary Fig. 19b–d).

Similarly, zebrafish express only a two-LBD variant of dTgfbr2a and do not exhibit alternative splicing at this locus. However, unlike chicken and horse, zebrafish possess a second paralog, Tgfbr2b, which encodes a receptor variant containing a single LBD (Fig. 7a). To compare the functional properties of these paralogs, we assessed their ligand-binding affinity and signaling capacity. The LBD of dTgfbr2b displayed a modestly higher affinity for TGFβ1 relative to the inner LBD of dTgfbr2a (Fig. 7b). Notably, dTgfbr2b also showed markedly increased ligand sequestration at the cell surface, likely attributable to the inhibitory effect of the outer LBD in dTgfbr2a (Fig. 7c). This was reflected in reduced downstream signaling competence of dTgfbr2a vs. dTgfbr2b, suggesting potential for fine tuning TGFβ signaling by dTgfbr2a compared to dTgfbr2b, indicating that the two paralogs may enable nuanced regulation of TGFβ signaling in zebrafish (Fig. 7d).

**Fig. 7: Duplicated LBD tgfbr2 paralog exhibits limited signaling competence and cell type specific expression pattern in zebrafish.**

To explore potential tissue-specific functions of the two zebrafish paralogs, we analyzed publicly available single-cell RNA-sequencing data from developing zebrafish embryos. In contrast to chicken and horse, where only the two-LBD variant is present, the single-LBD dTgfbr2b paralog was broadly expressed across all tissues (Fig. 7e). By contrast, the two-LBD dTgfbr2a paralog displayed selective enrichment in the intermediate mesoderm and hematopoietic lineages, with its expression increasing relative to dTgfbr2b during development (Fig. 7f, g).

Comparative transcriptomic profiling of cells expressing either Tgfbr2 paralog revealed largely non-overlapping gene expression signatures, suggesting distinct biological roles (Fig. 7h). Gene Ontology (GO) enrichment analysis revealed that genes specifically upregulated in dTgfbr2a⁺ cells were predominantly associated with cytoskeletal remodeling, small GTPase signaling and migration, particularly within immune and cardiovascular contexts (Fig. 7i). Given the well-established role of TGFβ-signaling in driving cellular plasticity through epithelial-to-mesenchymal-transition (EMT)^25,26,27,28, we next interrogated the expression of canonical EMT transcription factors. Intriguingly, EMT drivers such as the orthologs of Zeb1, Snail, Slug and Twist were markedly enriched in dTgfbr2b⁺ cells (Fig. 7j), consistent with its higher signaling competence relative to dTgfbr2a (Fig. 7d).

To validate these differences in vivo, we transiently expressed each receptor in developing zebrafish embryos. Expression of dTgfbr2b robustly induced phenotypes associated with imbalanced TGFβ/BMP-signaling^29,30, including increased incidence of cyclopia and ventralization compared to dTgfbr2a (Fig. 7k, and Supplementary Fig. 20). Notably, deletion of the outer LBD in dTgfbr2a restored signaling activity, phenocopying the higher incidences of ventralization and cyclopia observed with dTgfbr2b, and indicating a structural basis for reduced signaling capacity of dTgfbr2a (Fig. 7k). Conversely, the isolated expression of dTgfbr2a LBD^out enhanced cyclopia incidence relative to full-length dTgfbr2a, but not ventralization. Overall, the elevated ventralization observed in vivo is consistent with the higher signaling output measured in vitro for dTgfbr2a LBDⁱⁿ and dTgfbr2b (Fig. 5e, f; and Supplementary Fig. 12; Fig. 7c, d), suggesting that increased dTgfbr2 signaling capability contributes to the ventralizing effect (Fig. 7k).

Together, these findings reveal that LBD duplication in paralogous TGFBR2 receptors can drive neofunctionalization, yielding a receptor variant with attenuated signaling capacity. This structural innovation likely enables tissue-specific modulation of TGFβ-signaling output, offering a previously unrecognized mechanism by which vertebrates can diversify and fine-tune key developmental and homeostatic pathways.

Discussion

In this study, we identify a striking evolutionary innovation: independent duplication of ligand-binding domains across three receptor classes (ACVR1, BMPR2, TGFBR2) in at least 12 vertebrate lineages. Our findings reveal that LBD multiplication serves as an evolutionary platform for modulating TGFβ-signaling at two distinct levels, by influencing ligand-binding capacity (e.g., chicken) or signal transmission efficacy (e.g., zebrafish). Moreover, we uncover recurrent instances of secondary LBD trimerization, notably within amphibians and salmonids, suggesting a sustained selective pressure for structural innovations in receptor design. These repeated, lineage specific adaptations point to a broader strategy by which vertebrates fine-tune pathway responsiveness through domain architecture. Given the association of TGFβ receptor mutations with numerous human developmental disorders^31,32,33, our findings underscore the evolutionary constraints on receptor core architecture, while simultaneously revealing unexpected structural plasticity through domain-level innovation.

The evolutionary roots of the TGFβ-receptors trace back to early metazoans such as sponges and placozoans, which encoded a minimal complement of two type I and one type II receptor genes¹⁴. With the emergence of bilateral symmetry, the TGFβ ligand repertoire expanded in parallel, reflecting the increasing demand for finely tuned BMP and TGFβ signaling during axis specification and early morphogenesis³⁴. By contrast, vertebrates exhibit a markedly expanded TGFβ receptor repertoire, reflecting two rounds of whole-genome duplications, with additional lineage-specific duplications in teleosts, Xenopus laevis^35,36,37 and more recently in carps and salmonids^14,38,39. This expansion supports the complex and context-dependent roles of TGFβ family ligands in vertebrate development, where they direct tissue specification, cell movement, and patterning along dorsal–ventral and anterior–posterior axes⁷.

Despite this diversification, vertebrate genomes encode more ligands than receptors, making signaling specificity heavily reliant on selective ligand expression and combinatorial assembly of tetrameric receptor complexes^14,34. Tight regulatory control is achieved through multiple mechanisms, including soluble antagonists, co-receptors, transcriptional modulators, and both post-translational and epigenetic modifications⁴⁰. The evolutionary significance of maintaining these precise interactions is underscored by the strong sequence conservation observed in co-evolved ligand-receptor pairs⁴¹.

Recurrent vertebrate WGDs generated additional TGFβ receptor paralogs, facilitating the evolutionary refinement of TGFβ signaling. Retention and divergence of paralogous receptor genes allowed one copy to evolve neofunctionalized or modular domain, while the other maintained a canonical receptor form^41,42. This is exemplified by multiple teleost species, where Y-linked AMHR2 paralogs evolved truncated LBDs and ligand-independent activity, co-opting as sex-determining factors across diverse lineages, including sharks, catfishes, percids, the alligator pipefishes and the common seadragon^43,44,45,46.

Here we showed that twelve independent LBD expansions across ACVR1, BMPR2, and TGFBR2 arose at different time points in vertebrate evolution. These events demonstrate that ligand-binding domain multimerization has repeatedly evolved as a novel layer of refining TGFβ receptor function.

The initial functional advantage of LBD duplication likely relates to the context-specific assembly of tetrameric TGFβ-receptor complexes⁴⁷. While a second LBD does not alter receptor stoichiometry per se, it could enhance inter-ligand capture, especially for high affinity receptors such as TGFBR2, in tissues with low ligand availability or restricted diffusion. For instance, duplicated LBDs in chicken, tapir and salmon TGFBR2s are nearly identical, potentially facilitating enhanced ligand avidity or interaction with ECM-bound latent ligands, analogous to the role of the co-receptor endoglin in BMP9 signaling^48,49. In these species the duplicated LBDs show low amino-acid substitution rates, in line with the clade-specific molecular clock, indicating that they are under strong evolutionary constraint and remain highly conserved. This is reflected by enhanced TGFβ1 binding of a synthetic human 2 LBD TGFBR2. Whereas, this receptor exhibits limited SMAD activation, likely due to high ligand sequestration, it remains an open question how 2 LBD receptors integrate into the native receptor complex compositions, including tetrameric complex stoichiometries and modulatory co-receptors. Notably, in lineages such as chicken and horse, alternative splicing can generate receptor variants that selectively include or exclude individual LBDs, suggesting that both duplicated domains remain functionally relevant and may be differentially deployed depending on cellular context.

This contrasts with zebrafish, where the duplicated outer LBD shows markedly higher amino-acid substitution rate, indicating relaxed constraint and suggesting that the domain is drifting or diversifying rather than being maintained for interchangeable function. Such relaxed constraint is likely enabled by the presence of the paralogous receptor (Tgfbr2b) that provides functional redundancy and permits neofunctionalization of the duplicated domains. Consistent with this, we show that the outer LBD of Tgfbr2a attenuates signaling, acting as an inhibitory module and illustrating domain-level neofunctionalization.

A remaining question is why LBD duplication has occurred repeatedly in only a subset of receptors while remaining rare across the broader TGFβ receptor family. Intriguingly, domain multimerization is not limited to vertebrates. Independent extracellular domain duplications in ancestral TGFBR2 orthologs have been described in invertebrates such as sponges and oysters^50,51, indicating that LBD duplication is an evolutionarily widespread strategy, independent of WGDs. However, within vertebrates, sporadic occurrence of LBD duplication in only a subset of receptors, suggests strong structural constraints on evolvability, with the affected receptors uniquely tolerant of domain multiplication while preserving initial ligand binding/signaling integrity in certain organisms. Tolerance of LBD duplication in some receptors may be facilitated by the co-expression of functionally conserved paralogous partners, such as ACVR1L alongside ACVR1 or ACVR2A/B alongside BMPR2. In contrast, TGFBR2 functions as the sole high-affinity type II receptor transmitting canonical TGFβ signaling. At the same time, the apparent restriction of LBD duplications to these receptors should be interpreted cautiously. Current genome assemblies and transcriptomic datasets remain uneven across the vertebrate phylogeny, and structural variants affecting ECDs may remain undetected in poorly sampled or fragmented genomes. Thus, additional examples of receptor domain multiplication may exist but fall below the current limits of detection.

Together, these findings reveal that LBD duplication acts as a versatile mechanism for modulating TGFβ-receptor function, enabling the evolution of context-sensitive signaling outputs. Given the pivotal role of TGFβ-signaling in regulating cell fate, plasticity, and migration, such domain-level adaptations likely contributed to the evolutionary diversification of vertebrate body plans. More broadly, these results establish LBD multimerization as a general mechanism to tune ligand engagement and signaling specificity within the TGFβ pathway.

Methods

Ethics and animal welfare statement

Our research complies with all relevant ethical regulations; the relevant animal protection committee of the Leibniz-IGB and the city of Berlin (LaGeSo) approved the animal use protocol where relevant. Anguilla anguilla and Cyprinus carpio were kept under animal husbandry permit ZH114 (LaGeSo, Berlin) at IGB; three other fish species (Gnathonemus petersii, Pantodon buchholzi, Erpetoichthys calabaricus) were obtained from commercial dealers and humanely euthanized using an overdose of buffered Tricaine PHARMAQ 1000 MG/G (MS222; concentration: 500 mg/L), for commercial fish at the day of arrival at IGB. This ensured that no pain, suffering, distress or lasting harm was inflicted on the animals. RNA samples of Xenopus laevis were obtained from control groups of an animal experiment approved by the German State of Health and Social Affairs (LaGeSo, Berlin, Germany; G0359/12).

Zebrafish experiments were performed in accordance with protocol BR22-1497 approved by the Institutional Animal Care and Use Committee (IACUC) of the National University of Singapore. Adult zebrafish were housed in recirculating aquaria systems at 28 °C under a 14 h/10 h light/dark cycle in the fish facility of the Department of Biological Sciences (DBS) at the National University of Singapore. Strain DBSWT zebrafish, source Department of Biological Sciences (DBS), National University of Singapore. Wild type zebrafish embryos were obtained by crossing corresponding adult male and female fish.

RNA isolation and cDNA synthesis

Taxon sampling includes selection from Danio rerio (fin–adult, whole larvae), Xenopus laevis (brain, testis, liver and gonad mixture (adult), larvae), Cyprinus carpio (gonad, muscle, brain, heart, fin – adult), Gnathonemus petersii (gonad, muscle, brain, heart, fin – adult), Pantodon buchholzi (gonad, muscle, brain, heart, fin–adult), Erpetoichthys calabaricus (gonad, muscle, brain, heart, fin–adult), Anguilla anguilla (gonad, muscle, brain, heart, fin – adult), and Gallus gallus (DF1 cell line). Tissues of Anguilla anguilla, Cyprinus carpio Gnathonemus petersii, Pantodon buchholzi, Erpetoichthys calabaricus were stored in RNAlater (Thermo Fisher Scientific) for 24 h at 8 °C, then drained and stored at −80 °C until further processing. RNAs were extracted using TRIzol Reagent (Thermo Fisher Scientific, Waltham, USA) according to the supplier’s recommendation and cleaned using the RNeasy Mini Kit (Qiagen). For synthesis of Danio rerio and Gallus gallus cDNA, 1 μg of total RNA was reversely transcribed by incubation with random primers (100 pmol μL⁻¹, Invitrogen) and M-MuLV reverse transcriptase enzyme (New England Biolabs) following manufactures recommendations. Synthesis of cDNA of Xenopus laevis and Cyprinus carpio was performed using iScript RT Supermix for RT-qPCR kit (Biorad); 10 μL of RNA (concentration of 80 ng/μL) were randomly reverse-transcribed following the manufactures recommendations.

RNA sequencing and de novo transcriptome assembly

For transcriptomics of Anguilla anguilla, Gnathonemus petersii, Pantodon buchholzi and Erpetoichthys calabaricus muscle tissue, RNA and library processing and RNA-seq were carried out by NOVOGENE (Cambridge, UK) on a NovaSeq 6000 PE 150, generating 9.3 – 10.0 Gb of raw data per sample. Raw data of each sample was processed using the fq2fa –merge command from IDBA v1.1.1⁵² package to convert fastq to interleaved fasta format. The corresponding files were assembled using idba_tran v 1.1.1 with default parameters. The highest kmer transcript assemblies (kmer 60) were aligned to a custom protein database (consisting of sequences for ACVR1, ACVRL1, BMPR1A, BMPR1B, ACVR1B, ACVR1C, TGFBR1, ACVR2A, ACVR2B, TGFBR2, BMPR2, AMHR2 and INHBA from Siniperca chuatsi⁵³ by blat⁵⁴ using parameters (−t=dnax −q=prot). Blat results were sorted by matching score and the best hit transcripts were written to a bed6 formatted table and extracted from the transcriptomes in 5’ to 3’ orientation using bedtools getfasta (-name -s)⁵⁵. The extracted sequences were further processed as described below.

Phylogenetic analysis

Sequences of TGFβ receptor family orthologs were screened for LBD multimerization using Ensembl gene tree function⁵⁶ and NCBI blast (https://blast.ncbi.nlm.nih.gov/Blast.cgi) against available vertebrate transcriptomes and/or genomes. RNA-seq was performed for selected species (Gnathonemus petersii, Pantodon buchholzi, Erpetoichthys calabaricus, Anguilla anguilla) to complement missing information in respective taxa. In the absence of available transcriptome data, ORFs were predicted in regions identified by nucleotide Blast of closely related species using Augustus web interface (https://bioinf.uni-greifswald.de/augustus/)⁵⁷. LBDs were predicted from amino acid sequences using NCBI Conserved Domain Search web interface (https://www.ncbi.nlm.nih.gov/Structure/cdd/wrpsb.cgi)⁵⁸. Phylogenetic trees were extracted from NCBI taxonomy CommonTree (https://www.ncbi.nlm.nih.gov/Taxonomy/CommonTree/wwwcmt.cgi) and represented using Interactive Tree of Life (iTOL) v6⁵⁹. Species images were extracted from Adobe Stock (ID: #301044215, #327193716, #370603116, #61858943, #115805136, #490808539, #125573996, #730691035, #552043292, #227759813, #574310862, #35897212, #370612679) under the license of the “Max Planck Institut für Molekulare Genetik, Berlin” and edited in Adobe Photoshop Version 26.8.0.

RNA splicing analysis

To analyse TGFBR2 alternative splicing in horse and chicken, RNA-seq data from various tissues (horse: PRJNA1017964; chicken: PRJEB26695) were aligned to the respective reference genomes (chicken: GCA_016699485.1; horse: EquCab3.0) using STAR v2.7.9a. Unique and multi-mapping reads for TGFBR2 exons 1–5 were extracted from the SJ.out.tab STAR output files using standard Python scripts. For sashimi plots, merged BAM files were visualized in IGV, and coverage tracks were exported. Junction reads were manually corrected by replacing them with unique and multi-mapping counts from the SJ.out.tab files. Reported values correspond to the sum across all analysed replicates. In chicken samples, no reads corresponding to the E1–E2’ junction were detected. In horse, E2 inclusion relative to E2’ was calculated using uniquely mapping reads only. For the final table, only samples with at least 40 unique reads for the constitutive E4–E5 junction and tissues with at least two biological replicates were included. For horse, selected tissues known to be regulated by TGFβ-signaling are represented.

RNA single cell analysis in zebrafish

Single-cell RNA-seq data from the zebrahub atlas were downloaded and processed in R using Seurat (v4)⁶⁰. Cells with <500 detected features or genes expressed in <30 cells were excluded. Data were normalized, variable features identified, and expression values scaled prior to PCA (50 components) and UMAP embedding (first 30 PCs). Metadata annotations from the zebrafish anatomy ontology available from the zebrahub atlas were used for cell-type assignments. Gene expression was visualized on UMAPs and dot plots. For the intermediate mesoderm and hematopoietic system, cells were stratified into four groups based on tgfbr2a/tgfbr2b expression (double-positive, tgfbr2a + , tgfbr2b + , or double-negative). Differential expression relative to double-negative cells was assessed using Seurat’s FindMarkers function, and pseudobulk-like expression profiles were generated by using Seurat’s AverageExpression function. Differentially expressed genes were filtered by a log2 FC ≥ 0.58 and an adjusted p value ≤ 0.05. Significantly diff. exp. genes relative to double-negative condition were compared using DeepVenn⁶¹ (https://www.deepvenn.com/) and visualized in Photoshop. gProfiler g:GOSt (https://biit.cs.ut.ee/gprofiler/gost) was used for functional enrichment analysis of genes specifically enriched in dTgfbr2a+ or dTgfbr2b+ cells compared to double-negative cells on indicated time points.

Molecular cloning

To functionally characterize TGFBR2 or BMPR2 receptors of Danio rerio, Cyprinus carpio, Xenopus laevis, Gnathonemus petersii, Gallus gallus and Equus caballus, we generated an N-terminally Halo-tagged receptor expression plasmid library (Supplementary Table 1). For this, we used various cloning methods including restriction-, Gibson-, blunt end- and deletion mutagenesis cloning, specified for each animal below. All primers used for PCR are listed in (Supplementary Table 2). Cloning PCRs were carried out on a Peltier Thermal Cycler PTC-200. The elongation time was adapted according to the product size (Phusion Pol. = 1 kb/min). PCR products were resolved by agarose gel electrophoresis and purified using NucleoSpin Gel and a PCR Clean-up kit (Macherey-Nagel), according to the manufacturer’s guidelines. Gibson cloning using NEBuilder HiFi DNA Kit (New England BioLabs) was performed according to the manufacturer’s protocol. Final products were transformed in chemically competent DH5α E. coli bacteria cells. All constructs were confirmed via Sanger sequencing followed by whole plasmid sequencing. Expression is confirmed via Western blot (Supplementary Fig. 21).

Danio rerio and cyprinus carpio

drTgfbr2a, drTgfbr2b and cTgfbr2a were amplified from respective cDNA and subcloned into a previously generated pcDNA3.1. hTGFBR2-halo plasmid²¹ in between EcoRI and NotI sites, replacing the original receptor ORF resulting in drTgfbr2a-Halo, drTgfbr2b-Halo and cTgfbr2a-Halo expression plasmids. Further, to generate drTgfbr2a-LBDⁱⁿ-Halo, drTgfbr2a-LBD^out-Halo, cTgfbr2a-LBDⁱⁿ-Halo and cTgfbr2a-LBD^out-Halo constructs, deletion mutagenesis was performed on drTgfbr2a-Halo and cTgfbr2a-Halo templates. Deletion primers were used to flank either the inner or outer ligand binding domain sequences. Subsequently, PCR constructs were ligated via blunt end ligation.

Gnathonemus petersii

Following the RNA-seq analysis of Gnathonemus petersii heart tissue, we ordered the synthesis of gnBmpr2a-pTwist-CMV plasmid at “Twist Bioscience” and amplified gnBmpr2a, gnBmpr2a-LBD^mid-in and gnBmpr2a-LBDⁱⁿ. We then subcloned these constructs into hTGFBR2-Halo plasmid in between EcoRI and NotI sites, replacing the original receptor ORF resulting in gnBmpr2a-Halo, gnBmpr2a-LBD^mid-in-Halo and gnBmpr2a-LBDⁱⁿ-Halo expression plasmids. Further, to generate gnBmpr2a-LBD^out-Halo and gnBmpr2a-LBD^mid-Halo constructs, Gibson cloning was performed on gnBmpr2a-Halo template.

Xenopus laevis

XlTgfbr2L was amplified from respective cDNA and subcloned using Zero Blunt™ PCR Cloning Kit (Thermo Fisher Scientific) according to the manufacturer’s protocol into a pCR™-Blunt-vector. The full length xlTgfbr2.L-Halo construct was synthesized by “Genewiz”. Consecutively, Gibson-cloning was performed on this vector. By this, we obtained xlTgfbr2.L-LBDⁱⁿ-Halo, xlTgfbr2.L-LBD^out-Halo, xlTgfbr2.L-LBD^mid-in-Halo.

Equus caballus

Due to high sequence identity between the inner and outer ligand binding domains, we obtained a codon optimized eTGFBR2-pTwist-CMV plasmid from “Twist Bioscience” and amplified eTGFBR2. We then subcloned this construct into hTGFBR2-Halo plasmid in between EcoRI and NotI sites, replacing the original receptor ORF resulting in eTGFBR2-Halo. Further, to generate eTGFBR2-LBDⁱⁿ-Halo and eTGFBR2-LBD^out-Halo constructs, Gibson cloning was performed on eTGFBR2-Halo template.

Gallus gallus

gTGFBR2 (ENSGALT00000018657.6) and gTGFBR2-LBD^out (ENSGALT00000037691.5) were amplified from respective cDNA and subcloned into hTGFBR2-halo plasmid in between EcoRI and NotI sites, replacing the original receptor ORF resulting in gTGFBR2-Halo and gTGFBR2-LBD^out-Halo. The subcloned gTGFBR2 ORF contained an additional exon in between the kinase coding exon 6 and exon 7, which was removed by Gibson cloning. Thereby, both gTGFBR2 and gTGFBR2-LBD^out variants encoded for the same kinase, making a direct comparison of the duplicated LBD signaling competence possible. Further, to generate gTGFBR2-LBDⁱⁿ-Halo, deletion mutagenesis was performed on gTGFBR2-Halo template. Deletion primers were used to flank the outer ligand binding domain. Subsequently, PCR constructs were ligated via blunt end ligation.

Capped mRNA synthesis and microinjection in zebrafish

To synthesize capped mRNA, the mMESSAGE mMACHINE T7 Transcription Kit (ThermoFisher) was used with DraIII-linearized dTgfbr2a-Halo, dTgfbr2a LBDⁱⁿ-Halo, dTgfbr2a LBD^out-Halo, and BfuI-linearized dTgfbr2b-Halo plasmids as templates, respectively. To synthesize capped PMT-mEGFP (membrane bound EGFP) mRNA, the mMESSAGE mMACHINE SP6 Transcription Kit (ThermoFisher) was used with NotI-linearized pcs2 + -PMT-mEGFP ^62,63 plasmids as templates. The final concentrations of synthesized mRNAs were measured by Qubit 4 Fluorometer (ThermoFisher) using Qubit™ RNA HS Assay Kit (ThermoFisher). For microinjection, wild type (WT) embryos were collected immediately after spawning and transferred to 0.3X Danieau’s solution (17.4 mM NaCl, 0.21 mM KCl, 0.12 mM MgSO₄, 0.18 mM Ca(NO₃)₂, 1.5 mM HEPES, pH = 7.2). Injected embryos were raised in a 28 °C incubator until 22 hpf for screening and imaging. Embryonic stages were defined by hours post fertilization (hpf) at 28 °C and morphological features⁶⁴. To compare the respective effects of dTgfbr2a-Halo, dTgfbr2a LBDⁱⁿ-Halo, dTgfbr2a LBD^out-Halo and Tgfbr2b-Halo in WT embryos, 170 pg dTgfbr2a-Halo, 148 pg dTgfbr2a LBDⁱⁿ-Halo, 148 pg dTgfbr2a LBD^out-Halo, or 148 pg of dTgfbr2b-Halo capped mRNAs were co-injected with 10 pg PMT-mEGFP at the border between cytoplasm and yolk of 1-cell stage WT embryos, respectively.

Fluorescence screening and brightfield imaging of dTgfbr2-Halo injected zebrafish embryos

Injected embryos were incubated in a 28 °C incubator and screened for mEGFP fluorescence at 22 hpf. mEGFP-positive embryos were characterized into three categories (normal, cyclopia and ventralized) based on eye, head and trunk morphologies for quantification. Representative embryos were imaged under a stereomicroscope (Nikon, SMZ18).

Cell culture

COS-7 and HEK293T cells were obtained from the German Collection of Microorganisms and Cell Cultures (DSMZ) and cultured in Dulbecco’s Modified Eagle’s Medium (DMEM) supplemented with 10% FCS, 2 mM L-glutamine and penicillin (100 units/mL) / streptomycin (100 µg/mL) (DMEM full medium) in a humidified atmosphere at 37 °C and 5% CO₂ (v/v). COS-7 and HEK293T cells were maintained in T175 flasks and split 1:5 or 1:10, depending on need and were kept sub-confluent. For passaging, cells were washed once with PBS before being removed from the flasks surface with trypsin/EDTA (0.05/0.02% in PBS).

SDS-PAGE & Western-blotting

For sodium dodecyl sulphate polyacrylamide gel-electrophoresis (SDS-PAGE), cells transiently expressing Halo-receptors were lysed in 150 μL Laemmli buffer and frozen at −20 °C. To ensure a homogeneous loading, cell lysate was pulled through a 1 ml syringe and boiled for 10 min at 95 °C before loading onto 10% polyacrylamide gels. After performed gel-electrophoresis, proteins were transferred onto Methanol-activated PVDF membranes by Western-blotting. Next, membranes were blocked for 1 h in a solution containing 0.1% TBS-T and 3% w/v bovine serum albumin (BSA), then washed three times in 0.1% TBS-T and incubated with indicated primary antibodies overnight at 4 °C. Primary antibodies: anti-Halo (ProMega; #G9211; monoclonal mouse antibody), anti-pSMAD2 Ser465/467 (Cell Signaling; #3108; monoclonal rabbit antibody), and anti-GAPDH (Cell Signaling; #2118; monoclonal rabbit antibody) were used at a 1:1000 dilution in 3% w/v BSA/ TBS-T solution. For HRP-based detection, membranes were incubated with secondary goat-α-mouse or goat-α-rabbit IgG HRP conjugates ( ± 0.8 mg/ml, Dianova; #111-035-144, #115-035-068) for 1 h at a dilution of 1:10000. Chemiluminescent reactions were processed using WesternBright Quantum HRP substrate (Advansta Inc.) and documented on a FUSION FX7 digital imaging system.

Confocal microscopy & ligand surface binding assay (LSBA)

LSBA assay and consecutive image analysis and quantification was previously described in detail for SiR-d12-hTGFβ1 and Cy5-hActivin A²¹. In brief, for visualization of SiR-d12-hTGFβ1 and Cy5-hActivin A ligand binding, 200.000 COS-7 cells / well were seeded in 1 mL DMEM full medium on glass cover slips in a 12 well plate. On the following day, cells were transfected with desired constructs (500 ng DNA per well) using Lipofectamine2000 according to the manufacturer’s instructions. The day after, cells were washed with PBS and, additionally to 0.5 mM fluorescent HaloTag-ligand CA-Alexa488 (Alexa Fluor 488, Promega, #G1002), simultaneously incubated with saturating conc. of SiR-d12-hTGFβ1 or Cy5-hActivin A for 30 min at 4 °C. Subsequently, cells were washed once with ice-cold PBS before fixation with 100% methanol for 5 minutes. Cells were washed again with PBS and mounted with Fluoromount G (Invitrogen, 00-4958-02). Confocal images were acquired on a commercial expert line Abberior STED microscope, using the Imspector software from Abberior Instruments (Version 16.3) in line-scanning confocal mode; objective lens: 100X NA1.4 (oil) [UPLS], pinhole: 1.0 AU, range: 75 µm × 75 µm, confocal pixel size: 60 nm, pixel dwell time: 5/10 µs. Images were acquired using 485 nm (20% laser power) and 640 nm excitation (20% laser power). The detection windows were set to 500–540 nm and 650–750 nm. Confocal raw data were post-processed and adjusted for color and contrast (linear adjustments maintained for confocal datasets represented within one figure) using Fiji (ImageJ) software and Adobe Photoshop (Adobe Systems). Surface binding quantification of SiR-d12-hTGFβ1 and Cy5-hActivin A on COS-7 cells transiently expressing Halo-Receptor constructs was performed with Fiji. Per cell, four regions of interest (ROI) (100 µm²) were chosen, and the raw integrated intensity (RawIntDen) of each ROI was measured both in receptor and ligand channels. Per condition, 30 cells were quantified in total in 3 independent experiments (N = 3). Where fewer than three datasets are shown for clarity, the images are representative; all replicates exhibited consistent trends despite variability in signal intensity. Receptor-ligand binding was calculated relative to intensity values of untransfected COS-7 cells, representing endogenous ligand-receptor binding. Normalized RawIntDen values of receptors and ligands were plotted in GraphPad Prism 10.5 (GraphPad Software Inc.).

Dual luciferase reporter gene assay

For dual luciferase reporter gene assay, 50.000 HEK293T cells / well were seeded in 200 µL DMEM full medium in a 96 well plate. On the following day, cells were transfected with 50 ng of corresponding Halo-tagged receptors constructs together with 50 ng of SMAD2/3 sensitive (CAGA)₁₂ luciferase reporter. A constitutively expressing construct encoding renilla luciferase (RL-TK; Promega) was co-transfected (30 ng) as internal control. The next day, cells were starved in serum-free medium for 3 h before stimulation with 0.2 nM rhTGFβ1 (PeproTech, Hamburg, Germany) or 0.02 nM rhActivin A (Gift from Marko Hyvönen) for 24 h. Cell lysis was performed using passive lysis buffer (Promega, # E1910) and measurement of luciferase activity was carried out according to manufacturer’s instructions using a TECAN Spark plate reader. Experiments testing the TGFBR2 ortholog signaling competence were normalized against the rhTGFβ1 response of hTGFBR2-Halo. All experiments were performed at least three independent times.

Structure prediction and interface energy calculations

The structure of the receptor and ligand complexes were first predicted using AlphaFold2 multimer^65,66 using the single receptor ECDs and two ligand chains (either hActivin A, hTGFβ1 or the respective animal variants) for the ligand homodimers. In order to obtain a physically more validated structure while simultaneously scoring the interface, first coordinate constrained relaxation protocol was performed to optimize H-bonds based on the Rosetta Energy Function 2015 (ref2015_cst) within RosettaScripts. Local Monte-Carlo-based protein-protein docking using RosettaScripts²⁰ was performed in 2500 replicas. Here, first, the “Docking” mover was used with low resolution (backbone plus centroid) flags before a high resolution (full atom) docking (using ref2015 scoring function). Other parameters as distance perturbation and angle perturbation were left on default. The Interface RMSD (iRMSD) was plotted against the score of the interface residues (i_SC) of the top 500 structures (determined by i_SC) and mean binding energies were calculated. Lower interface score correlates with favourable interface energies. Lower iRMSD correlates to docking structures closer to the starting AlphaFold2-multimer conformation. For structural representation, the top 10 hits showing the lowest i_SC were structurally investigated for the final homology model.

To evaluate whether linker length permits intra-ligand bivalency between two LBDs of hTGFBR2 2LBD, complexes with hTGFβ1 dimer and varying linker lengths were generated using AF2 as described above. In the predicted structures, both receptor domains are placed at their canonical binding positions on opposing sides of the TGFβ dimer, with the interdomain linker spanning between them. To assess whether the linker is long enough to wrap around the dimer surface, or whether it is too short and forced to thread through the dimer interior, we developed a spatial embedding metric. For each linker residue, identified by matching the known core sequence (SESVNNDMIVTDNNGAVKFP) and its flanking GS extensions, the surrounding space was partitioned into eight octants. A residue was classified as embedded within the dimer if heavy atoms of the ligand occupied five or more of eight octants within 5 Å of its Cα atom. AF2 generated structures with misplaced receptor domains were excluded. A linker length of 36 residues did not result in threading through the dimer interior and was therefore defined as the predicted cutoff for intra-ligand bivalency. This threshold was derived from the top three ranked predictions. Naturally occurring linker lengths in TGFBR2 variants were compared to this cutoff. All used scripts for structure predication and interface energy calculations are available under (https://github.com/agknaus/Jatzlau-Trumpp-et-al.-structural-computational-analysis/).

Molecular clock and evolutionary rate analysis of LBDs

Amino-acid sequences of LBD regions were aligned in MEGA X⁶⁷. Alignments were generated using the MUSCLE algorithm with default parameters. Pairwise evolutionary distances between LBD sequences were calculated in MEGA X using the Analyze → Compute Pairwise Distances function. Distances were estimated under the Jones–Taylor–Thornton (JTT) substitution model with gamma-distributed rate heterogeneity (Γ = 1.0) and a homogeneous substitution pattern among lineages. The substitution type was set to Amino acid, and gaps or missing data were handled using pairwise deletion. Variance estimation was performed using the bootstrap method with 1000 replicates. For each focal species (chicken, horse, salmon, zebrafish), distances were calculated between its inner LBD and the orthologous inner LBDs of other species within the same clade. Distances between the two LBD copies within the focal species were computed separately to quantify intra-species divergence associated with the domain duplication event.

Species divergence times for all pairwise comparisons were obtained directly from TimeTree.org⁶⁸. These values were used as the independent variable in all molecular-clock analyses. For each clade, the amino-acid substitution rate (r, substitutions per site per million years) was estimated by performing a weighted linear regression of pairwise amino-acid distances against their corresponding TimeTree-derived species divergence times. The slope of the regression provided the clade-specific molecular-clock rate for the inner LBD domain. The observed amino-acid distance between the two LBD copies within each focal species was divided by the clade-specific molecular-clock rate to obtain the duplication-specific evolutionary rate (r₍dup₎). This value reflects the effective rate of amino-acid substitution between the two paralogous domains since their duplication.

Statistical analysis

All statistical tests were performed using GraphPad Prism version 10.5.0 software and are listed in the figure legends. Normal distribution of data sets was tested with the Shapiro–Wilk normality test. In cases of failure to reject the null hypothesis, the ANOVA and Tukey’s or Dunnett post hoc test were used to check for statistical significance under the normality assumption. For all experiments statistical significance was assigned, with an alpha-level of p < 0.05.

Graphical schemes and figures

Graphs were generated using GraphPad Prism 11.0.1. Molecular structure representations were generated using PyMOL Version 3.1.0. Selected schematic elements were created with BioRender.com under a publication license and incorporated into the final figure panels during assembly in Adobe® Photoshop (Adobe Systems, San José, USA). BioRender-created schematic elements are included in the graphical legend elements of Figs. 1 and 2; Fig. 3a, e; Fig. 4a; Fig. 5a–c, d, g, and j (left panel); Fig. 6a, h; Fig. 7a; Supplementary Fig. 1a, b; Supplementary Fig. 2a, b; Supplementary Fig. 4b–e; Supplementary Fig. 9a,d (left); Supplementary Fig. 10a–e (receptor scheme); Supplementary Fig. 11a–d (left); Supplementary Fig. 17a,b; and Supplementary Fig. 19a,c,d. These elements are covered by the BioRender publication license: Created in BioRender. Trumpp, M. (https://BioRender.com/8p64dhh).

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Data availability

The RNA raw sequence datasets for Gnathonemus petersii, Pantodon buchholzi, Erpetoichthys calabaricus and Anguilla anguilla generated in this study have been deposited in the NCBI/SRA under the BioProject database under accession code PRJNA1314022. Source data are provided with this paper.

Code availability

All used scripts for structure predication and interface energy calculations are available under https://github.com/agknaus/Jatzlau-Trumpp-et-al.-structural-computational-analysis/, https://doi.org/10.5281/zenodo.19291097⁶⁹.

References

Pires-daSilva, A. & Sommer, R. J. The evolution of signalling pathways in animal development. Nat. Rev. Genet 4, 39–49 (2003).
Article CAS PubMed Google Scholar
Barberan, S., Martin-Duran, J. M. & Cebria, F. Evolution of the EGFR pathway in Metazoa and its diversification in the planarian Schmidtea mediterranea. Sci. Rep. 6, 28071 (2016).
Article ADS CAS PubMed PubMed Central Google Scholar
Bogdan, S. & Klambt, C. Epidermal growth factor receptor signaling. Curr. Biol. 11, R292–R295 (2001).
Article CAS PubMed Google Scholar
Itoh, N. & Ornitz, D. M. Evolution of the Fgf and Fgfr gene families. Trends Genet 20, 563–569 (2004).
Article CAS PubMed Google Scholar
Kipryushina, Y. O., Yakovlev, K. V. & Odintsova, N. A. Vascular endothelial growth factors: A comparison between invertebrates and vertebrates. Cytokine Growth Factor Rev. 26, 687–695 (2015).
Article CAS PubMed Google Scholar
Trumpp, M. et al. Characterization of fibrodysplasia ossificans progessiva relevant Acvr1/Acvr2 activin receptors in medaka (Oryzias latipes). PLoS ONE https://doi.org/10.1371/journal.pone.0291379 (2023).
Zinski, J., Tajer, B. & Mullins, M. C. TGF-β family signaling in early vertebrate development. Cold Spring Harb Perspect Biol. https://doi.org/10.1101/cshperspect.a033274 (2018).
Hiepen, C., Yadin, D., Rikeit, P., Dorpholz, G. & Knaus, P. Actions from head to toe: an update on bone/body morphogenetic proteins in health and disease. Cytokine Growth Factor Rev. 27, 1–11 (2016).
Article CAS PubMed Google Scholar
Nickel, J. & Mueller, T. D. Specification of BMP signaling. Cells https://doi.org/10.3390/cells8121579 (2019).
Yadin, D., Knaus, P. & Mueller, T. D. Structural insights into BMP receptors: Specificity, activation and inhibition. Cytokine Growth Factor Rev. 27, 13–34 (2016).
Article CAS PubMed Google Scholar
Bragdon, B. et al. Bone morphogenetic proteins: a critical review. Cell Signal 23, 609–620 (2011).
Article CAS PubMed Google Scholar
Upadhyay, A., Moss-Taylor, L., Kim, M. J., Ghosh, A. C. & O’Connor, M. B. TGF-β family signaling in drosophila. Cold Spring Harb Perspect Biol. https://doi.org/10.1101/cshperspect.a022152 (2017).
Huminiecki, L. et al. Emergence, development and diversification of the TGF-beta signalling pathway within the animal kingdom. BMC Evol. Biol. 9, 28 (2009).
Article PubMed PubMed Central Google Scholar
Zheng, S., Long, J., Liu, Z., Tao, W. & Wang, D. Identification and evolution of TGF-β signaling pathway members in twenty-four animal species and expression in tilapia. Int. J. Mol. Sci. https://doi.org/10.3390/ijms19041154 (2018).
Hoegg, S., Brinkmann, H., Taylor, J. S. & Meyer, A. Phylogenetic timing of the fish-specific genome duplication correlates with the diversification of teleost fish. J. Mol. Evol. 59, 190–203 (2004).
Article ADS CAS PubMed Google Scholar
Ning, B. et al. Genomic organization, intragenic tandem duplication, and expression analysis of chicken TGFBR2 gene. Poult. Sci. 101, 102169 (2022).
Article CAS PubMed PubMed Central Google Scholar
Heinecke, K. et al. Receptor oligomerization and beyond: a case study in bone morphogenetic proteins. Bmc Biol. 7, 59 (2009).
Article PubMed PubMed Central Google Scholar
Sebald, W., Nickel, J., Zhang, J.-L. & Mueller, T. D. Molecular recognition in bone morphogenetic protein (BMP)/receptor interaction. Biol. Chem. 385, 697–710 (2004).
Article CAS PubMed Google Scholar
Chu, K. Y., Malik, A., Thamilselvan, V. & Martinez-Hackert, E. Type II BMP and activin receptors BMPR2 and ACVR2A share a conserved mode of growth factor recognition. J. Biol. Chem. 298, 102076 (2022).
Article CAS PubMed PubMed Central Google Scholar
Fleishman, S. J. et al. RosettaScripts: a scripting language interface to the Rosetta macromolecular modeling suite. Plos One 6, e20161 (2011).
Article ADS CAS PubMed PubMed Central Google Scholar
Jatzlau, J. et al. A versatile Halo- and SNAP-tagged BMP/TGFβ receptor library for quantification of cell surface ligand binding. Commun. Biol. https://doi.org/10.1038/s42003-022-04388-4 (2023).
Olsen, O. E. et al. Activin A inhibits BMP-signaling by binding ACVR2A and ACVR2B. Cell Commun. Signal 13, 27 (2015).
Article PubMed PubMed Central Google Scholar
Radaev, S. et al. Ternary complex of transforming growth factor-β1 reveals isoform-specific ligand recognition and receptor recruitment in the superfamily*. J. Biol. Chem. 285, 14806–14814 (2010).
Article CAS PubMed PubMed Central Google Scholar
Gunning, P. W., Schevzov, G., Kee, A. J. & Hardeman, E. C. Tropomyosin isoforms: divining rods for actin cytoskeleton function. Trends Cell Biol. 15, 333–341 (2005).
Article CAS PubMed Google Scholar
Piera-Velazquez, S. & Jimenez, S. A. Endothelial to mesenchymal transition: role in physiology and in the pathogenesis of human diseases. Physiol. Rev. 99, 1281–1324 (2019).
Article CAS PubMed PubMed Central Google Scholar
Goumans, M.-J., van Zonneveld, A. J. & ten Dijke, P. Transforming growth factor β–induced endothelial-to-mesenchymal transition: a switch to cardiac fibrosis?. Trends Cardiovascular Med. 18, 293–298 (2008).
Article CAS Google Scholar
Lamouille, S., Xu, J. & Derynck, R. Molecular mechanisms of epithelial–mesenchymal transition. Nat. Rev. Mol. Cell Biol. 15, 178–196 (2014).
Article CAS PubMed PubMed Central Google Scholar
Dooley, S. & ten Dijke, P. TGF-β in progression of liver disease. Cell Tissue Res 347, 245–256 (2012).
Article CAS PubMed Google Scholar
Preiß, H. et al. Regulation of Nodal signaling propagation by receptor interactions and positive feedback. Elife https://doi.org/10.7554/eLife.66397 (2022).
Pomreinke, A. P. et al. Dynamics of BMP signaling and distribution during zebrafish dorsal-ventral patterning. eLife 6, e25861 (2017).
Article PubMed PubMed Central Google Scholar
Verstraeten, A., Alaerts, M., Van Laer, L. & Loeys, B. Marfan syndrome and related disorders: 25 years of gene discovery. Hum. Mutat. 37, 524–531 (2016).
Article CAS PubMed Google Scholar
Ye, D. et al. Insights into bone morphogenetic proteins in cardiovascular diseases. Front Pharm. 14, 1125642 (2023).
Article CAS Google Scholar
Kaplan, F. S., Al Mukaddam, M., Stanley, A., Towler, O. W. & Shore, E. M. Fibrodysplasia ossificans progressiva (FOP): a disorder of osteochondrogenesis. Bone 140, 115539 (2020).
Article CAS PubMed PubMed Central Google Scholar
Hinck, A. P., Mueller, T. D. & Springer, T. A. Structural biology and evolution of the TGF-β family. Cold Spring Harb Perspect. Biol. https://doi.org/10.1101/cshperspect.a022103 (2016).
Jaillon, O. et al. Genome duplication in the teleost fish Tetraodon nigroviridis reveals the early vertebrate proto-karyotype. Nature 431, 946–957 (2004).
Article ADS PubMed Google Scholar
Nakatani, Y., Takeda, H., Kohara, Y. & Morishita, S. Reconstruction of the vertebrate ancestral genome reveals dynamic genome reorganization in early vertebrates. Genome Res 17, 1254–1265 (2007).
Article CAS PubMed PubMed Central Google Scholar
Session, A. M. et al. Genome evolution in the allotetraploid frog Xenopus laevis. Nature 538, 336–343 (2016).
Article ADS CAS PubMed PubMed Central Google Scholar
Wang, J.-T., Li, J.-T., Zhang, X.-F. & Sun, X.-W. Transcriptome analysis reveals the time of the fourth round of genome duplication in common carp (Cyprinus carpio). BMC Genom. 13, 96 (2012).
Article CAS Google Scholar
Gundappa, M. K. et al. Genome-wide reconstruction of rediploidization following autopolyploidization across one hundred million years of salmonid evolution. Mol. Biol. Evol.https://doi.org/10.1093/molbev/msab310 (2021).
Deng, Z. et al. TGF-β signaling in health, disease and therapeutics. Signal Transduct. Target. Ther. 9, 61 (2024).
Article PubMed PubMed Central Google Scholar
Herpin, A., Lelong, C. & Favrel, P. Transforming growth factor-β-related proteins: an ancestral and widespread superfamily of cytokines in metazoans. Dev. Comp. Immunol. 28, 461–485 (2004).
Article CAS PubMed Google Scholar
Rivera, A. M. & Swanson, W. J. The importance of gene duplication and domain repeat expansion for the function and evolution of fertilization proteins. Front Cell Dev. Biol. 10, 827454 (2022).
Article PubMed PubMed Central Google Scholar
Kuhl, H. et al. Multi-genome comparisons reveal gain-and-loss evolution of anti-Mullerian hormone receptor type 2 as a candidate master sex-determining gene in Percidae. Bmc Biol. 22, 141 (2024).
Article CAS PubMed PubMed Central Google Scholar
Wen, M. et al. An ancient truncated duplication of the anti-Müllerian hormone receptor type 2 gene is a potential conserved master sex determinant in the Pangasiidae catfish family. Mol. Ecol. Resour. 22, 2411–2428 (2022).
Article CAS PubMed PubMed Central Google Scholar
Qu, M. et al. Seadragon genome analysis provides insights into its phenotype and sex determination locus. Sci. Adv. https://doi.org/10.1126/sciadv.abg5196 (2021).
Feron, R. et al. Characterization of a Y-specific duplication/insertion of the anti-Mullerian hormone type II receptor gene based on a chromosome-scale genome assembly of yellow perch, Perca flavescens. Mol. Ecol. Resour. 20, 531–543 (2020).
Article CAS PubMed PubMed Central Google Scholar
Goebel, E. J., Hart, K. N., McCoy, J. C. & Thompson, T. B. Structural biology of the TGFβ family. Exp. Biol. Med (Maywood) 244, 1530–1546 (2019).
Article CAS PubMed PubMed Central Google Scholar
Saito, T. et al. Structural basis of the human endoglin-BMP9 interaction: insights into BMP signaling and HHT1. Cell Rep. 19, 1917–1928 (2017).
Article CAS PubMed PubMed Central Google Scholar
Kim, S. K., Henen, M. A. & Hinck, A. P. Structural biology of betaglycan and endoglin, membrane-bound co-receptors of the TGF-beta family. Exp. Biol. Med (Maywood) 244, 1547–1558 (2019).
Article CAS PubMed PubMed Central Google Scholar
Suga, H., Ono, K. & Miyata, T. Multiple TGF-β receptor related genes in sponge and ancient gene duplications before the parazoan–eumetazoan split 1. FEBS Lett. 453, 346–350 (1999).
Article CAS PubMed Google Scholar
Herpin, A. et al. Structural and functional evidence for a singular repertoire of BMP receptor signal transducing proteins in the lophotrochozoan Crassostrea gigas suggests a shared ancestral BMP/activin pathway. Febs j. 272, 3424–3440 (2005).
Article CAS PubMed Google Scholar
Peng, Y. et al. IDBA-tran: a more robust de novo de Bruijn graph assembler for transcriptomes with uneven expression levels. Bioinformatics 29, i326–i334 (2013).
Article CAS PubMed PubMed Central Google Scholar
He, S. et al. Mandarin fish (Sinipercidae) genomes provide insights into innate predatory feeding. Commun. Biol. 3, 361 (2020).
Article CAS PubMed Google Scholar
Kent, W. J. BLAT—The BLAST-like alignment tool. Genome Res. 12, 656–664 (2002).
CAS PubMed PubMed Central Google Scholar
Quinlan, A. R. & Hall, I. M. BEDTools: a flexible suite of utilities for comparing genomic features. Bioinformatics 26, 841–842 (2010).
Article CAS PubMed PubMed Central Google Scholar
Dyer, S. C. et al. Ensembl 2025. Nucleic Acids Res. 53, D948–D957 (2024).
Article Google Scholar
Keller, O., Kollmar, M., Stanke, M. & Waack, S. A novel hybrid gene prediction method employing protein multiple sequence alignments. Bioinformatics 27, 757–763 (2011).
Article CAS PubMed Google Scholar
Wang, J. et al. The conserved domain database in 2023. Nucleic Acids Res 51, D384–D388 (2023).
Article ADS CAS PubMed PubMed Central Google Scholar
Letunic, I. & Bork, P. Interactive tree of life (iTOL) v6: recent updates to the phylogenetic tree display and annotation tool. Nucleic Acids Res 52, W78–W82 (2024).
Article PubMed PubMed Central Google Scholar
Hao, Y. et al. Integrated analysis of multimodal single-cell data. Cell 184, 3573–3587.e3529 (2021).
Article CAS PubMed PubMed Central Google Scholar
Hulsen, T. DeepVenn--a web application for the creation of area-proportional Venn diagrams using the deep learning framework Tensorflow. js. arXiv preprint arXiv:2210.04597 (2022).
Balasubramanian, H., Sankaran, J., Pandey, S., Goh, C. J. H. & Wohland, T. The dependence of EGFR oligomerization on environment and structure: A camera-based N&B study. Biophys. J. 121, 4452–4466 (2022).
Article ADS CAS PubMed PubMed Central Google Scholar
Le, Y. et al. Midkine-a interacts with Ptprz1b to regulate neural plate convergence and midline formation in the developing zebrafish hindbrain. Dev. Biol. 521, 52–74 (2025).
Article CAS PubMed Google Scholar
Kimmel, C. B., Ballard, W. W., Kimmel, S. R., Ullmann, B. & Schilling, T. F. Stages of embryonic development of the zebrafish. Dev. Dyn. 203, 253–310 (1995).
Article CAS PubMed Google Scholar
Jumper, J. et al. Highly accurate protein structure prediction with AlphaFold. Nature 596, 583–589 (2021).
Article ADS CAS PubMed PubMed Central Google Scholar
Evans, R. et al. Protein complex prediction with AlphaFold-multimer. Preprint at bioRxiv, https://doi.org/10.1101/2021.10.04.463034 (2021).
Kumar, S., Stecher, G., Li, M., Knyaz, C. & Tamura, K. MEGA X: molecular evolutionary genetics analysis across computing platforms. Mol. Biol. Evol. 35, 1547–1549 (2018).
Article CAS PubMed PubMed Central Google Scholar
Kumar, S. et al. TimeTree 5: an expanded resource for species divergence times. Mol Biol Evol. https://doi.org/10.1093/molbev/msac174 (2022).
Obendorf, L. agknaus/Jatzlau-Trumpp-et-al.-structural-computational-analysis: publication_version_01. Zenodo https://doi.org/10.5281/zenodo.19291098 (2026).

Download references

Acknowledgements

The authors would like to thank Francesca Bottanelli (Freie Universität Berlin) for providing access to her commercial expert line Abberior STED microscope.

Funding

J.J. and P.K. acknowledge the support from Deutsche Forschungsgemeinschaft DFG (SFB1444) and the Einstein Center ECRT). M.T. and L.O. were supported by the Max Planck Research School IMPRS-BAC. Funding has been in part provided by the German Research Foundation (DFG) STO493/8-1 to MS. C.W. was funded by grants from the Singapore Ministry of Education (grant numbers MOE-2016-T3-1-005 and MOE-T2EP30221-0008). W.B. was supported by Berlin–Brandenburg School for Regenerative Therapies. Open Access funding enabled and organized by Projekt DEAL.

Author information

These authors contributed equally: Jerome Jatzlau, Michael Trumpp.

Authors and Affiliations

Freie Universität Berlin, Institute of Chemistry and Biochemistry - Biochemistry, Berlin, Germany
Jerome Jatzlau, Michael Trumpp, Julia Kühlwein, Leon Obendorf, Marco Preussner, Paul-Lennard Mendez, Hendrik Burkert, Wiktor Burdzinski & Petra Knaus
International Max Planck Research, School for Biology and Computation, Berlin, Germany
Michael Trumpp & Leon Obendorf
Department of Biological Sciences and Centre for Bioimaging Sciences, University of Singapore, Singapore, Singapore
Yao Le & Christoph Winkler
Department of Fish Biology, Fisheries and Aquaculture, Leibniz Institute of Freshwater Ecology and Inland Fisheries, Berlin, Germany
Heiner Kuhl & Matthias Stöck
Ecotoxicological Laboratory, German Environment Agency (UBA), Berlin, Germany
Heiner Kuhl
Berlin-Brandenburg School for Regenerative Therapies (BSRT), Berlin, Germany
Wiktor Burdzinski & Petra Knaus
Max Planck Institute for Molecular Genetics, Berlin, Germany
Stefan Mundlos

Authors

Jerome Jatzlau
View author publications
Search author on:PubMed Google Scholar
Michael Trumpp
View author publications
Search author on:PubMed Google Scholar
Julia Kühlwein
View author publications
Search author on:PubMed Google Scholar
Leon Obendorf
View author publications
Search author on:PubMed Google Scholar
Yao Le
View author publications
Search author on:PubMed Google Scholar
Heiner Kuhl
View author publications
Search author on:PubMed Google Scholar
Marco Preussner
View author publications
Search author on:PubMed Google Scholar
Paul-Lennard Mendez
View author publications
Search author on:PubMed Google Scholar
Hendrik Burkert
View author publications
Search author on:PubMed Google Scholar
Wiktor Burdzinski
View author publications
Search author on:PubMed Google Scholar
Stefan Mundlos
View author publications
Search author on:PubMed Google Scholar
Christoph Winkler
View author publications
Search author on:PubMed Google Scholar
Matthias Stöck
View author publications
Search author on:PubMed Google Scholar
Petra Knaus
View author publications
Search author on:PubMed Google Scholar

Contributions

J.J., M.T. designed the study, evaluated the data and generated all figures; M.T., J.J., J.K., Y.L., W.B., H.B. performed the experiments; J.J. and M.T. performed the phylogenetic analysis; L.O., J.K. performed receptor homology modeling & docking calculations; M.T. produced hTGFβ1-SiR-d12; Y.L. performed zebrafish experiments; M.S. collected the tissue samples of G. petersii, P. buchholzi, E. calabaricus, A. anguilla and X. laevis, and prepared RNA and cDNA and discussed ancient/recent genome duplications; C.W. provided D. rerio cDNA samples. M.T., J.J., J.K. and H.B. cloned all Halo-receptor constructs. H.K. processed the transcriptome data and analyzed the results. M.P. performed RNA splicing analysis, P.M. analyzed scRNA-seq data; J.J. W.B. analyzed LSBA data; P.K., M.S., C.W. and S.M. discussed the project and gave critical input. J.J., M.T. and P.K. wrote the manuscript; all authors commented on the manuscript.

Corresponding authors

Correspondence to Jerome Jatzlau or Petra Knaus.

Ethics declarations

Competing interests

The authors declare no competing interest.

Peer review

Peer review information

Nature Communications thanks the anonymous reviewers for their contribution to the peer review of this work. A peer review file is available.

Additional information

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary information

Supplementary Information (download PDF )

Peer Review file (download PDF )

Reporting Summary (download PDF )

Source data

Source Data (download XLSX )

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.

Reprints and permissions

About this article

Cite this article

Jatzlau, J., Trumpp, M., Kühlwein, J. et al. Recurrent evolution of ligand-binding domain multiplicity fine-tunes TGFβ signaling in vertebrates. Nat Commun 17, 4458 (2026). https://doi.org/10.1038/s41467-026-73340-5

Download citation

Received: 11 January 2026
Accepted: 08 May 2026
Published: 19 May 2026
Version of record: 19 May 2026
DOI: https://doi.org/10.1038/s41467-026-73340-5