Evolutionary trajectories, early diversification, and species-specific amplification of the metazoan inhibitor of apoptosis (IAP) repertoire

Abstract

IAP proteins play central roles in regulating apoptosis and diverse cellular processes, influencing cancer biology, or stress resistance in economically important species, highlighting their major biomedical and ecological relevance. Despite this importance, IAP evolutionary diversity and history remain largely unexplored. Here, we present a large-scale comparative analysis of 2,843 IAP proteins from 312 species spanning all major animal lineages, revealing striking variation in IAP repertoire size and exceptional architectural diversity. We show that IAP expansion has repeatedly emerged through lineage-specific duplication events, seemingly shaped by different duplication dynamics with more recent gene family expansions in arthropods compared with mollusks and chordates. Expression data show that IAP expansion supports distinct strategies. Bivalves and gastropods mobilize multiple IAP subfamilies in response to biotic and abiotic stress, whereas aphids exhibit differential IAP expression associated with polyphenism. These contrasting patterns indicate that heterogeneous selective pressures have recurrently reshaped IAP repertoires, promoting lineage- and species-specific functional diversification. Despite this diversification, phylogenetic analyses reveal the maintenance of a core set of three IAP types. Survivin/Deterin-like and Bruce-like IAPs emerged early in metazoans and remained structurally conserved, although Bruce-like IAPs were independently lost in chelicerates, most nematodes, and some hemipteran insects. RING-containing IAPs also originated early but followed more dynamic evolutionary trajectories, being lost in nematodes and platyhelminthes while expanding in other lineages through domain acquisition (eg gnathostomes) or repeated domain duplication and loss (eg mollusks and insects). Our findings establish a comprehensive evolutionary framework for metazoan IAPs, linking lineage-specific diversification to structural innovation and functional specialization.

Introduction

Inhibitors of apoptosis (IAPs), also known as baculovirus IAP repeat (BIR)-containing proteins (BIRCs), were discovered in baculoviruses and shown to inhibit apoptosis (Crook et al. 1993; Birnbaum et al. 1994). IAP homologs have since been identified in all the fungi and animals where they were studied (Uren et al. 1998; Berthelet and Dubrez 2013; Ribeiro Lopes et al. 2019; Gebreegziabher Amare et al. 2022; Witkop et al. 2022). Early studies showed that most IAPs can inhibit the activity of caspases, a family of cysteine proteases that form the effector arm of the apoptotic machinery, by sterically blocking their active site, sequestering them away from their substrates, or promoting their degradation by the proteasome (Tenev et al. 2005; Domingues and Ryoo 2012; Berthelet and Dubrez 2013). In addition, there is growing evidence that IAPs also regulate many important non-apoptotic cellular processes, such as cell division (Wheatley and Altieri 2019), cell migration (Kenneth and Duckett 2012), signal transduction (Kocab and Duckett 2016; Kumar et al. 2020), autophagy (Cheung et al. 2020), or innate immunity (Estornes and Bertrand 2015; Sharma et al. 2017). These functions make them important players in several types of cancer and promising oncotherapeutic targets (Cossu et al. 2019). Recent studies also suggest that IAPs may enhance stress resistance in economically important mollusk species (Zhang et al. 2012; Zhu et al. 2021; Witkop et al. 2022) and serve as RNAi-targets for pest control (Yoon et al. 2020). These findings highlight the broad significance of IAP research and the need to better understand the genetic and structural basis of their pleiotropy.

The extreme modularity of IAP architecture, with multiple protein domains that vary in number and type, may account for the wide range of cellular functions they can regulate. The IAP family is defined by the presence of at least one N-terminal BIR domain (Eckelman et al. 2008; Berthelet and Dubrez 2013), involved in interactions with apoptotic or signaling proteins. BIRs are ∼70-amino acids long and contain a conserved stretch of three cysteines and one histidine (G-X(2)-Y-X(5-8)-D-X(3)-C-X(2)-C-X(6)-W-X(9)-H-X(6)-C) coordinating a single zinc ion (Berthelet and Dubrez 2013; Cossu et al. 2019) (Fig. S1). They are classified as short or long, depending on their size, and as type I or II, based on the presence of a deep surface hydrophobic groove involved in the interaction with the IAP Binding Motif (IBM) of caspases and IAP antagonists (Berthelet and Dubrez 2013) (Fig. 1). Most IAPs also possess one or more additional domains, often involved in ubiquitination (Fig. 1). The most common is a C-terminal RING (really interesting new gene) zinc finger domain, whose ubiquitin ligase (E3) activity targets caspases, IAP antagonists, and IAPs themselves for ubiquitination, often leading to their degradation by the proteasome (Vaux and Silke 2005; Budhidarmo and Day 2015). Other domains include (i) a central caspase recruitment domain (CARD), which can inhibit the E3 ligase activity of IAPs by preventing RING domain dimerization, (ii) a ubiquitin associated (UBA) domain, which promotes the binding of IAPs to poly- ubiquitinated proteins from cell signaling complexes, and (iii) a ubiquitin conjugating (UBC) domain, which catalyzes the covalent attachment of ubiquitin to target proteins (Budhidarmo and Day 2015).

Figure 1.

Domain composition of IAP proteins from Drosophila melanogaster, Homo sapiens, and Caenorhabditis elegans. Functional domains are plotted according to their position within the proteins. BIR (Baculoviral IAP Repeat) domains, characteristic of the IAP protein family, are represented by red, orange, or yellow rectangles depending on their size (eg short vs. long BIR domains) and on the presence of a deep surface hydrophobic groove, which is only found in Type II BIR domains and allows the interaction with the IBM (IAP binding motif) of caspases and IAP antagonists. Type I BIR domains have a shallow pocket and interact mainly with proteins involved in cell signaling pathways (Berthelet and Dubrez 2013; Sharma et al. 2017). Additional domains include: RING domains (Really Interesting New Gene, green oval), UBA domains (Ubiquitin Associated domains, light blue squircle), CARD domains (CAspase Recruitment Domain, purple rectangle), UBC domains (Ubiquitin-Conjugating Enzymes, blue hexagon), NACHT domains (NAIP, CIITA, HET-E, and TP1, black rectangle), and LRR (leucine-rich repeat, blue circle). Architectures and domains are not represented to scale.

Since their discovery, in-depth studies of IAPs and their functions have largely focused on humans (8 IAPs) and model organisms such as the nematode Caenorhabditis elegans (2 IAPs) and the dipteran Drosophila melanogaster (4 IAPs). Although these studies have been essential in advancing our understanding of IAPs, they are insufficient to account for the extraordinary diversity and complexity of IAPs in metazoans. Indeed, sequencing efforts in recent decades have revealed unexpectedly large lineage-specific IAP expansions in many mollusks (from 40 IAPs in the Pacific oyster Crassostrea gigas to 159 in the hard clam Mercenaria mercenaria (Song et al. 2021; Witkop et al. 2022)) and in certain insects (28 IAPs in the pea aphid Acyrthosiphon pisum (Ribeiro Lopes et al. 2020)). The exploration of animal genomes has also revealed new domain architectures, such as IAPs with up to five BIRs (model species have a maximum of three, Fig. 1), or IAPs with two RING domains in aphids, with functions related to the regulation of mutualistic bacteria-containing symbiotic cells (Ribeiro Lopes et al. 2020).

Phylogenetic analyses of IAPs and BIR domains are rare and based on a limited number of species (mainly deuterostomes) (Verhagen et al. 2001; Robertson et al. 2006; Cao et al. 2008), which has not allowed a clear picture of their evolution in metazoans, especially regarding the early steps of their diversification. As a consequence, the evolutionary patterns and processes that have led to IAP diversity are only partially understood.

Here, taking advantage of the ever-increasing number of sequenced animal genomes (Hotaling et al. 2021), we investigate the evolution of IAP gene across metazoans. We established an ad hoc computational pipeline to systematically mine 312 genomes, representative of 41 metazoan classes, and reconstructed the diversification and evolutionary history of this complex and biologically important gene family. Our analyses reveal striking variation in IAP repertoire size and domain architectures, even among closely related species. Investigation of the mechanisms underlying IAP expansion and divergence among homologs suggests that this diversity is seemingly shaped by different duplication dynamics across lineages and evolutionary timescales. Some IAPs belong to ancient orthogroups that have been conserved over evolutionary time, yet exhibit marked differences in their copy numbers among species, indicative of independent and lineage-specific expansions. In contrast, other orthogroups are restricted to particular taxa or even single species, consistent with more recent and taxon-specific diversification. By integrating expression data from lineages with expanded repertoires, we further show that IAP diversification supports distinct functional strategies across lineages. Mollusks deploy multiple IAP subfamilies in response to biotic and abiotic stress, whereas decapods and insects show limited transcriptional responses to environmental or pathogen-induced stress, suggesting alternative roles for expanded IAP repertoires. A remarkable example is represented by aphids, in which differential expression of distinct IAPs is associated with different polyphenisms or symbiosis. Together, these contrasting patterns, from ancient duplicates in mollusks and chordates to recent, rapid radiations in arthropods, suggest that distinct evolutionary pressures have repeatedly reshaped IAP lineage- and species-specific functional diversification. To gain a better understanding of this diversification, we analyzed IAPs in early branching metazoan lineages to deduce the ancestral metazoan repertoire of IAPs and propose scenarios for the first stages of their diversification. We then focus on deuterostomes and on the otherwise overlooked protostome lineage, uncovering complex evolutionary trajectories involving multiple and recurrent gene and protein domain duplication, rearrangement, or loss. Collectively, our findings provide the first comprehensive and large-scale evolutionary framework for metazoan IAPs, highlighting how lineage-specific duplication dynamics have shaped their structural and functional diversification.

Results and discussion

Metazoans show extreme variation in IAP number

We used a standardized procedure (see Materials and Methods section) to search for IAPs in 312 metazoan proteomes (including 282 species for which IAPs had never searched before), belonging to 14 phyla and 41 classes representative of the major metazoan lineages (Table 1). We identified 2,843 putative IAP-encoding genes. We observed huge differences in the size of IAP repertoires with numbers ranging from one (eg in the placozoan Trichoplax adhaerens, the nematode Trichinella spiralis, or in ctenophores) to 169 (eg in the hard clam M. mercenaria) (Fig. 2a and Data S1). The median number of IAPs is 4 (6.5 on average), but 49 species (15.7%) have more than 10 IAPs and 12 (3.8%) have more than 50. Amplification of IAPs is not restricted to specific lineages, and variability is observed at all evolutionary scales (Fig. 2a and Data S1). For example, within the superphylum Lophotrochozoa, the annelid Helobdella robusta has 3 IAPs, while mollusks show strong variation, with cephalopods having 8 to 11 IAPs, gastropods 22 to 89 (average 48), and bivalves 54 to 169 (average 81). Within Arthropoda, decapods have 6 to 75 IAPs (22 on average) and hemipterans 3 to 60 (14 on average). Variability is also observed at lower taxonomic levels, eg at family level in Ixodidae (3 to 43 IAPs) and Formicidae (4 to 18 IAPs) or within genera in Bombus bumblebees (6 to 13 IAPs) (Fig. 2a). While there is a weak but significant correlation between the number of IAP proteins and the total number of predicted genes (Spearman correlation r = 0.38; P-value = 10⁻¹²; Fig. S2), there is a strong variability in IAP numbers even for similar gene numbers and vice versa, suggesting that, in some organisms, IAP expansion has been favored or constrained independently of the rest of the genome. While assembly or annotation artifacts may contribute to some overestimation of IAP numbers, they are insufficient to fully account for the extent of the observed variability, specifically the up to 40-fold increase in IAP number seen in certain species, and the consistent expansion seen in almost all species in some taxa (eg mollusks and aphids). Such substantial differences are more plausibly reflective of genuine biological divergence rather than technical artifacts alone. Taken together, this shows the dynamic nature of the IAP family and its propensity to lineage-specific expansion.

Table 1.

Distribution of the 312 metazoan species considered in this study into superphyla and phyla, and information about their number of predicted IAPs.

Figure 2.

Diversity of IAP repertoire in metazoans. a) Distribution of the 2,843 putative IAP-encoding genes identified in 312 metazoan species. Colors indicate the phylum to which a species belongs. b) Schematic representation of the 121 IAP architectures predicted in the 312 metazoan species. Architectures are sorted according to the number of BIR domains they contain. The background of each architecture is colored according to a gradient depending on the proportion of species containing said architecture among the 312 considered (<1%; 1% to 5%; 5% to 20%; 20% to 50%; 50% to 100%).

IAP duplication dynamics vary across lineages

The mechanisms driving IAP expansion have previously been examined in only three bivalve species, where tandem duplication was identified as a major contributor (Song et al. 2021; Zhu et al. 2021; Witkop et al. 2022). In the present study, we broadened these analyses by investigating the mechanisms underlying IAP expansion in 20 additional species: four chordates, one more bivalve, three gastropods, one arachnid, three decapods, and eight insects (three hemipterans, one coleopteran, three hymenopterans, and one dipteran). These species were selected because they possess more than 10 IAPs and chromosome-level genome assemblies are available. For each species, we determined the relative frequencies of tandem, proximal, transposition, and segmental duplication events contributing to IAP expansion (Fig. 3a and Data S2). The results reveal variable proportions of each duplication mode across species, even closely related ones.

Figure 3.

Duplication modes and pairwise distance patterns of IAP genes across 20 selected species possessing more than 10 IAPs and with chromosome-level genome assemblies. a) Distribution of duplication modes for IAP genes in 20 species, categorized as tandem duplications (TDs), proximal duplications (PDs), segmental duplications (SDs), and dispersed duplications (DDs). Species belong to eight distinct animal groups (Chordata, Mollusca, Chelicerata, Decapoda, Hemiptera, Coleoptera, Hymenoptera, and Diptera), and, for each species, pie charts indicate (i) the proportion of IAPs located on short unplaced scaffolds, (ii) the proportion of intronless IAPs among IAPs derived specifically from dispersed duplication events, and (iii) the proportion of intronless IAPs among all IAPs. b) Distribution of pairwise dissimilarity scores calculated for all IAP gene pairs within each species. The dissimilarity score reflects both sequence similarity and alignment coverage and is defined as follows: 1−(Identity × Coverage), ranging from 0 to 1. For each species, the percentage of gene pairs with scores below 0.5 is indicated in the upper-left corner of the corresponding distribution plot. Che., chelicerates; Col., Coleoptera.

Tandem duplications were detected in almost all organisms (18 out of 20), but they were the predominant type only in certain cases. For example, tandem duplications account for 60% of duplication events in the black soldier fly Hermetia illucens, compared with only 13% in the shrimp Penaeus chinensis and 0% in the tick Dermacentor silvarum, despite these species having similar numbers of IAPs (Fig. 3a). In the scallop Pecten maximus, tandem duplications represent only 26% of all duplication events. This is considerably lower than what has been previously reported in other bivalves, where tandem duplications account for 43%, 58.5%, and 78% of duplication events in M. mercenaria (Song et al. 2021), Mizuhopecten yessoensis (Zhu et al. 2021), and Crassostrea virginica (Witkop et al. 2022), respectively. These observations suggest that, unlike what was previously reported in bivalves, tandem duplications contribute only modestly to IAP expansion in some species, highlighting lineage-specific evolutionary mechanisms.

Our analysis of proximal duplication signatures (ie duplicates separated by a few intervening genes) further emphasizes the heterogeneity of duplication dynamics across species. As with tandem duplications, the proportion of proximal duplicates varies widely, ranging from 0% to 33% (Fig. 3a). Mechanisms behind proximal duplications mostly include unequal crossing-over and tandem expansion followed by local rearrangement, or transposon-mediated insertion of a copy nearby (Panchy et al. 2016; Lallemand et al. 2020). When tandem and proximal duplications are considered together, the combined signal supports the idea that local mechanisms, especially unequal crossing-over, may have contributed to IAP expansion in all 20 species examined. However, this contribution remains limited in most lineages: only four species (Bombus affinis, Bombus terrestris, H. illucens, and Diabrotica virgifera) show more than half of their duplicates arising from the combined action of tandem and proximal duplications. Thus, even when accounting for potential ancient tandem events that have since been obscured, tandem-derived duplication remains a modest driver of IAP diversification in many taxa.

It is important to note that short unplaced scaffolds in some genome assemblies may lead to an underestimation of proximal and tandem duplication events. In fact, 5 of the 20 species considered have more than half of their IAPs located on such scaffolds (Fig. 3a). Nevertheless, we find no significant negative correlation between the inferred number of IAPs originating from local duplication events and the number of unplaced scaffolds (Fig. S3). Moreover, several species still show substantial levels of tandem-derived IAPs despite the prevalence of unplaced scaffolds, as exemplified by Penaeus monodon, which shows 33% tandem-derived IAPs even though 83% of its IAPs reside on such scaffolds.

Interestingly, half of the species examined (Aphis gossypii, B. affinis, B. terrestris, H. illucens, Pomacea canaliculata, P. chinensis, P. maximus, P. monodon, Patella vulgata, and Xenopus laevis; Data S2) possess genomic clusters of at least three IAPs inferred to arise from tandem and/or proximal duplications, indicating that some genomic regions are hotspots for local IAP expansion. However, sequence divergence analyses reveal limited local synteny within these clusters: paralogs from different duplication events are interspersed, and gene orientations frequently differ. As an example, H. illucens has the largest cluster, containing 12 IAPs within a 384 kb region on chromosome 2 (Fig. S4a), interrupted by only two non-IAP genes. These 12 genes fall into three distinct paralog groups based on sequence similarity (Fig. S4b), showing that multiple independent duplication events contributed to the formation of the cluster. Consistent with this, even adjacent genes, such as LOC119650015 and LOC119650016, belong to different paralog groups despite their tandem arrangement. Moreover, the six genes in the third paralog group are split across the cluster: two are positioned upstream of members of the second paralog group, and four lie downstream, with one oriented in the same direction and three in the opposite direction. These findings demonstrate that IAP clusters do not reflect linear tandem amplification but instead result from recurrent duplication and rearrangement processes that continuously reorganize local genomic architecture.

Another contributor to IAP expansion that was not detected in previous studies is segmental duplication, which corresponds to the duplication of a large chromosomal region. Although this type of duplication is observed in only three species in our dataset (X. laevis, Gigantopelta aegis, and Aphidius gifuensis), it played a particularly important role in IAP expansion in the latter two species, accounting for 39% and 67% of the observed duplications, respectively (Fig. 3a).

Finally, dispersed duplicates, which are randomly distributed across chromosomes, were found in each of the 20 selected species. In 14 of them, they accounted for at least 53% of all duplication events (Fig. 3a). To estimate the contribution of retrotransposition to the origin of these dispersed duplicates, we quantified how many are intronless. Intronless duplicates were detected in 12 species, and in two of them, D. silvarum and A. pisum, they constituted more than half of all duplication events. This supports the hypothesis that retrotransposition contributes to IAP expansion to various extents across diverse taxa. Notably, intronless IAPs also occur in tandem in six species (H. illucens, A. pisum, D. silvarum, D. virgifera, G. aegis, and P. vulgata), suggesting that tandem duplication may sometimes follow retrotransposition. The remaining dispersed intron-containing IAP duplicates represent more than 50% of all duplication events in 11 species. Such dispersed duplicates often arise from DNA transposon activity (Wang et al. 2012; Tan et al. 2021; Ma et al. 2023). Enrichment of DNA transposons near IAP genes compared with other genomic regions, as was described in the hard clam M. mercenaria (Song et al. 2021), may have facilitated IAP expansion in some species.

Taken together, these results show that, contrary to what was suggested by earlier studies based on a limited set of species (Song et al. 2021; Zhu et al. 2021; Witkop et al. 2022), tandem duplication is not the primary driver of IAP expansion in most lineages. Instead, dispersed duplications often represent the largest fraction of duplication events, strongly suggesting that transposition may play a major role in shaping IAP repertoires across many species. Our results also reveal a strong but taxon-specific contribution of segmental duplication. These findings underscore the diversity and lineage-specificity of IAP duplication dynamics, and highlight the importance of comparative analyses that encompass a broad taxonomic range to accurately capture the evolutionary processes governing gene family expansion. They are moreover particularly relevant for understanding the functional consequences of IAP duplication, as dispersed duplicates, especially those generated via transposable element activity, are more likely than tandem or segmental duplicates to form chimeric transcripts or experience changes in regulatory context, thereby facilitating structural and functional diversification (Tan et al. 2021; Ma et al. 2023).

IAP duplication dynamics vary over evolutionary timescales

To better characterize the lineage-specific dynamics underlying the expansion of metazoan IAPs, we sought to estimate the relative evolutionary divergence of duplicated genes across the 20 species examined. Duplication age is commonly inferred using the number of synonymous substitutions per synonymous site (Ks) (Nei and Gojobori 1986; Qiao et al. 2019). However, the extreme diversity of IAP architectures, resulting from processes such as partial gene duplication as well as intra-IAP domain duplication and loss (see section Metazoans show extreme variation in IAP architecture), precludes the construction of reliable global sequence alignments. Such alignments would inevitably force non-homologous regions to align, leading to biased or uninterpretable Ks estimates. To circumvent this limitation, we adopted an alternative approach, based on pairwise sequence dissimilarity calculated from local alignments (see section Materials and methods). This metric was used as a proxy for the relative evolutionary divergence of duplicated genes, with lower dissimilarity score values interpreted as reflecting either more recent duplication events or reduced accumulation of substitutions since duplication. Importantly, variation in pairwise distances may reflect not only differences in duplication bursts but also lineage-specific evolutionary rates or differing levels of functional constraint acting on duplicated genes.

Using this framework, the inferred dissimilarity score profiles of IAP pairs within each species revealed distinct, lineage-specific patterns (Fig. 3b). Mollusks and chordates exhibit a predominance of highly dissimilar IAP pairs, consistent with either ancient duplication events and/or elevated rates of sequence divergence following duplication, with fewer than 1.3% of pairs showing low dissimilarity. Petromyzon marinus constitutes a notable exception, displaying an excess of pairs with low dissimilarity that may reflect a recent duplication burst or stronger evolutionary constraints limiting divergence. Arthropods also show a higher proportion of IAP pairs with low dissimilarity, indicating an enrichment of relatively recent duplications and/or reduced post-duplication divergence due to stronger functional constraints. Notably, duplication histories differ even among closely related species, as illustrated by the contrasting patterns observed in the three aphids A. gossypii, A. pisum, and Rhopalosiphum maidis (Fig. 3b). Collectively, these results point to substantial lineage- and species-specific variation in IAP duplication dynamics, likely shaped by a combination of duplication timelines, evolutionary rate heterogeneity, and selective constraints acting over distinct evolutionary timescales.

This is further supported by the search for orthologs among the 538 IAPs encoded by the 20 species selected for this analysis. This enabled us to place the duplicates within a phylogenetic context by inferring a shared evolutionary origin for IAPs from different species belonging to the same orthogroups. Indeed, the 538 IAPs can be sorted into 28 orthogroups out of the 26,354 identified by OrthoFinder when considering the total proteome of each species (Table 2 and Data S2). Two orthogroups (OG0000153 and OG000757) include representatives from all species, while two others (OG0000214 and OG0008528) contain members from 15 species out of 20 (Table 2). These four orthogroups contain known homologs of human and Drosophila IAPs: Drosophila DIAP2 and human XIAP in OG0000153, Drosophila DIAP1 in OG0000214, Drosophila Deterin and human Survivin/BIRC5 in OG0000757, and finally Drosophila Bruce and human BIRC6 in OG0008528. These orthogroups likely correspond to ancient IAPs that were already present in the common ancestor and conserved in most if not all species. Importantly, the number of IAPs within these four orthogroups varies greatly among species, indicating lineage-specific duplications. For instance, OG0000153 has expanded substantially in mollusks, chordates, and decapods, whereas OG0000214 shows marked expansion in insects. By contrast, OG000757 and OG0008528 generally retain a single copy per species, with only few notable exceptions. On the other hand, some orthogroups are restricted to particular taxa, and 17 contain IAPs from only a single species (Table 2), reflecting further lineage-specific diversification.

Table 2.

Distribution of IAP genes across all IAP-containing orthogroups identified in 20 selected species possessing more than 10 IAPs and with chromosome-level genome assemblies.

^*NA, not attributed.

Interestingly, we consistently identified a tandem pair of genes, one from OG0000153 and one from OG0000214, that is conserved across all examined insects apart from H. illucens (Data S2). Each gene in this pair shows clear orthology to its counterpart in other insects, suggesting that this arrangement originated from an ancestral tandem duplication and has been stably retained throughout insect evolution. Surrounding genomic regions, however, exhibit minimal conserved synteny, with at most two neighboring genes out of 20 belonging to the same orthogroups across species (excluding the two closely related bumblebees), indicating dynamic local rearrangements. The persistence of this tandem pair across ∼400 million years of insect evolution, despite extensive rearrangements around it, strongly suggests that the tandem genes themselves are under selective pressure to remain adjacent and that their physical linkage is biologically meaningful. Several studies have documented such conserved micro-synteny in the absence of broader synteny across diverse animal lineages, and interpreted them as the consequence of cis-regulatory constraints or functional linkage (Irimia et al. 2012; Moretti et al. 2017; Robert et al. 2022). For instance, Irimia et al. (2012) reported almost 600 pairs of genes that have remained tightly physically linked in diverse lineages across over 600 million years of evolution, and showed that conserved pairs are often transcriptionally coordinated, and can even form genomic regulatory blocks, where enhancers located in one gene affect the expression of the neighboring gene. The aforementioned IAP pair may therefore represent an ancestral functional module within the IAP family whose integrity has been maintained by selective pressure rather than neutrality.

Overall, these results indicate that distinct evolutionary pressures might act on IAPs and support lineage-specific expansion and functional diversification, as was documented for many other gene families such as the toll-like receptor (Leulier and Lemaitre 2008), the olfactory receptor (Policarpo et al. 2024), or the cytochrome P450 gene family (Kawashima and Satta 2014).

Metazoans show extreme variation in IAP architecture

In addition to changes in IAP number, we hypothesize that changes in IAP architecture would also be a factor contributing to functional diversification. To assess this, we performed a comprehensive survey of IAP domains across the 2,843 putative IAPs identified in our whole dataset, and identified 121 different domain architectures (Fig. 2b). Most IAPs contain additional domains in addition to BIR domains. RING domains are the most frequently observed, found in 50.3% of all IAPs (ie 81.4% of the IAPs that combine non-BIR and BIR domains). Additional frequently observed domains include UBC, UBA, death-like domains, WD40-repeat, and Quinoprotein alcohol dehydrogenase-like domains. Other domains are rare, even in species with an expanded IAP repertoire, and architectures that include them are often present in less than 1% of the species in our dataset (Fig. 2b). Importantly, CARD and NACHT domains, which had previously only been described in vertebrates (Cao et al. 2008), can also be found in some lophotrochozoans (eg in M. mercenaria or Lingula anatina). Similarly, other domains, like peptidase or reverse transcriptase domains, which had previously only been described in mollusks, are also found in several arthropods. Several domains were observed in a single IAP sequence in a single species (eg DAPIN, C2, Tropomyosin, SRCR, GPCR, MAB21, and MTA70 domains) (Fig. 2b and Data S3 and S4). Further studies are needed to determine whether these atypical architectures are real or artefactual (eg possible genome assembly and/or gene prediction artifacts due to lack of sensitivity (ie false negatives) or specificity (ie false positives) of the tools and/or HMM profiles used to identify domains). The same applies to proteins that appear to be the result of a fusion of two IAPs (eg BIR(s)-RING(s)-BIR(s)-RING(s)) and have been identified in several mollusks and insects, such as members of the Apis genus. While the genome of Apis mellifera contains two tandem IAPs with BIR-BIR-BIR-RING and BIR-BIR-RING architectures, respectively, Apis dorsata, Apis florea, and Apis laboriosa each possess a single, unique IAP with a BIR-BIR-BIR-RING-BIR-BIR-RING architecture (Data S4 and Fig. S5). Importantly, although this structure could reflect an annotation artifact, its consistent prediction in three distinct species would rather support the possibility of a gene fusion event.

Overall, the above results indicate that insertion of new domains is not a major driver of IAP diversification in Metazoa. Instead, among the 121 architectures identified, diversity results mostly from three elements: (i) the presence/absence of UBA and Death-like domains, (ii) the nature of the repeat surrounding the BIR in UBC-containing proteins (Fig. S6), and (iii) variation in the number of short BIRs (one to six) and RING (zero to two) (Fig. 2b). The first two elements allow the sorting of most architectures into groups of IAPs that have the same number of short BIR and RING domains and only differ in the presence of UBA and/or Death-like domains, or the nature of the repeats in UBC-containing proteins. These distinct architectures are not restricted to a specific lineage and can be found even in closely related species (eg all 37 Drosophila species in the dataset have a RING-containing IAP with three BIRs, but only 32 possess a UBA domain), highlighting either the lability of these domains or difficulties in their prediction. For this reason, groups of IAPs with the same number of short BIR and RING domains but differing in the aforementioned domains were considered as having the same architecture for the remainder of our analysis. This allowed us to focus on core architectural features that are broadly conserved, and less likely to be affected by domain prediction inconsistencies.

In line with the idea that gene duplication can facilitate architectural diversification, members of the expanded OG0000153 and OG0000214 orthogroups present a wide variety of architectures, whereas there is no variation between members of orthogroups OG00000757 and OG00008528 (Table 2 and Data S2). Moreover, even within each species, duplicated gene copies often have different domain architectures (Data S2 and Fig. S4b), with varying numbers of BIR and RING domains, an observation that was already made in M. mercenaria (Song et al. 2021) and that we now confirm in several other species. Although some of these variations may stem from gene prediction errors (eg inaccurate gene boundary annotations), domain-level events such as intra-IAP duplications and losses likely also contribute to this diversity. For instance, H. illucens and X. laevis possess duplicate IAPs with varying number of BIR domains (from one to five) (Data S2), while aphids possess IAPs with two tandem RING domains, a specificity of this insect group, consistent with repeated intra-IAP domain duplication events. In B. terrestris, duplication of an IAP gene with two BIRs and one RING domain, followed by the loss of a ∼300 bp region, generated a new architecture with just one BIR and one RING domain (Fig. S7). Both gene variants subsequently underwent tandem duplications, forming a cluster of six genes with two distinct architectures. Mutations have also contributed to architectural changes. In A. gifuensis, for instance, duplication of gene LOC122856375 (containing two BIRs and one RING domain) led to the formation of LOC122850491 (two BIRs) and LOC122856363 (one BIR). In both cases, point mutations introduced premature stop codons upstream of the second BIR or RING domain (Fig. S8). Given the variety of mechanisms (duplication, loss, and mutation) involved in IAP structural diversification, further studies are needed to fully understand how this diversity evolved in each species.

Previous studies have shown that, within an IAP, distinct BIR domains often have distinct substrate specificity and can independently mediate different functions (Berthelet and Dubrez 2013). Therefore, we hypothesize that the addition or loss of specific BIRs could directly alter the diversity of proteins an IAP can interact with and have important consequences regarding its role. Concerning the RING domain, previous studies have highlighted its key role in the regulation of apoptosis and IAPs that have lost this domain could prove unable to directly inhibit this process, as has been shown for truncated DIAP1 in D. melanogaster (Vaux and Silke 2005). Interestingly, only 16 species in our dataset have no RING-containing IAPs at all and, with the exception of the bee Ceratina calcarata and the ctenophores, they are all Placozoa, Nematoda, or Platyhelminthes. In C. calcarata, the absence of a RING-containing IAP may be an assembly artifact, as tblastn searches using Eufriesia mexicana sequences suggest that LOC108625468 (encoding an IAP with two BIRs) and LOC108622601 (encoding a RING-containing protein and isolated on a short scaffold) may actually represent a single gene split during genome assembly. The lack of RING-containing IAPs in the remaining 15 species suggests key differences in apoptosis control compared with other animals and a relaxation of selective pressure in those lineages. The two IAPs of C. elegans have not been reported to inhibit apoptosis, and instead to have functions in cytokinesis and immunity (Fraser et al. 1999; Lee et al. 2021), but a recent study provided strong evidence that the IAPs of Echinococcus granulosus (platyhelminthes) act as apoptosis inhibitors, at least in vitro (Zhan et al. 2020). Other worms in our dataset (H. robusta and Priapulus caudatus) also have a reduced IAP repertoire compared with other closely related species implying that this specific morphological organization could promote the loss of IAP genes.

Combined with our findings on lineage-specific expansion and duplication dynamics, these results indicate that metazoan IAPs have experienced a highly complex evolutionary history, which likely includes multiple rounds of gene duplication and loss, as well as domain duplication and rearrangement, occurring at both lineage and species levels. This extensive architectural plasticity suggests that metazoan IAP diversity has been shaped by lineage-specific selective pressures, potentially driving functional diversification of IAP repertoires across animals.

Expanded IAP repertoire supports lineage-specific functional diversification

The idea that distinct selective pressures independently promoted IAP expansion and architectural diversification in multiple lineages is further reinforced by the absence of any consistent relationship between IAP copy number/domain composition and either the evolutionary ancestry of the phylum to which an animal belongs (Fig. 2a), or a specific lifestyle (eg habitat, diet, symbiotic status…). However, of the 49 species that have more than 10 IAPs, 13 (27%) are arthropods known to be vectors of animal or plant pathogens, and 26 (53%) are found at shallow depths or at the air–water interface (Data S5). Among the latter group, six species are bivalves, the only taxon for which conditions potentially favoring IAP expansion have been previously investigated. In this group, the expansion of IAPs has been linked to their adaptation to a sessile lifestyle in the intertidal zone, a highly variable environment with fluctuating abiotic factors (eg temperature, salinity, air exposure, and heavy metal concentration). Supporting this hypothesis, several studies have shown that IAPs are strongly upregulated in bivalves in response to various environmental stressors. For example, air exposure led to up to a 67-fold increase in expression of five IAPs in the oyster C. gigas (Zhang et al. 2012), and the induction of up to 17 IAPs in the hard clam M. mercenaria (Song et al. 2021). In the same study, M. mercenaria also exhibited differential expression of over 84% of its expanded IAPs in response to changes in temperature, oxygen, and salinity. Interestingly, these IAPs, even those that were part of the same tandem array, displayed distinct expression patterns, both across different stressors and among tissues or time points, suggesting functional diversification. More specifically, different sets of IAPs were overexpressed in different tissues (especially hemolymph, gills, and mantle) or induced following air exposure, hypoosmolality, or thermal stress. Beyond abiotic stress, bivalves face constant exposure to microbial pathogens and accumulate significant levels of dinoflagellate-derived toxins due to their filter-feeding behavior. These biotic stressors may represent additional drivers of IAP expansion. In line with this, exposure to the toxic dinoflagellate Alexandrium catenella resulted in differential expression of 14 IAPs in the scallop M. yessoensis (Zhu et al. 2021) and 2 IAPs in C. gigas (Medhioub et al. 2013). Moreover, Witkop et al. (2022) found that 36% of C. virginica and 62.5% of C. gigas IAPs, regardless of their domain architecture, were differentially expressed in response to a variety of immune challenges, including viral (Ostreid Herpesvirus OsHV-1), bacterial (Vibrio spp., Aliiroseovarius crassostreae), and parasitic (Perkinsus marinus) infections. Notably, expression patterns varied by species and challenge type. C. virginica tended to activate distinct sets of IAPs for each challenge, whereas C. gigas often responded with overlapping sets of IAPs across different challenges. These studies suggest that, in bivalves, IAP diversity may be functionally relevant in enabling flexible and specific responses to a broad range of environmental and microbial challenges.

To determine whether the patterns observed in bivalves also occur in other species facing comparable biotic and abiotic stresses, we analyzed publicly available RNA-seq datasets for four previously unstudied species selected from the 26 shallow-water or air–water interface species with expanded IAP repertoires.

We first investigated IAP expression in gastropods, another group within the phylum Mollusca, selecting two species whose transcriptomic response to stresses similar to those studied in bivalves have been examined (Moreira et al. 2021; Lu et al. 2024). The common limpet P. vulgata shares the subtidal environment of bivalves, whereas the snail Biomphalaria glabrata lives at the air–water interface but in fresh water. We found that, in both species, IAP expression is modulated following stress exposure. In P. vulgata, 9 out of 32 IAPs were differentially expressed following heat stress (Fig. 4a and Data S6). Importantly, these genes are not concentrated within a single orthogroup but are distributed across six distinct orthogroups. These include one orthogroup that appears specific to gastropods (OG0009776, one upregulated gene) and another unique to P. vulgata (OG0015429, three downregulated genes), as well as two genes from the OG0000153 orthogroup that is expanded in mollusks (both upregulated) (Table 2). In addition to providing evidence for lineage-specific roles of IAPs, this distribution indicates that the heat-stress response involves multiple, evolutionarily distinct IAP subfamilies rather than relying on a single specialized orthogroup. Such a pattern is consistent with the idea that copy-number expansion has promoted functional diversification within the IAP family. We also observed differential expression of IAPs in B. glabrata exposed to parasitic infection. When comparing IAP expression 40 d after infection with the trematode Echinostoma paraensei versus the nematode Daubaylia potomaca, 29 of 89 IAPs show significant differential regulation between the two infection conditions (Fig. 4b and Data S6). This demonstrates that, in addition to being induced by bacterial and viral infection as was shown in bivalves, IAPs can also be regulated in response to parasites. On another hand, our findings confirm a trend observed in bivalves: IAP response is not only pathogen-induced but also pathogen-specific, further supporting the notion that diversified IAP repertoires are selectively recruited according to the nature of the stressor. This functional diversification is further reflected at the level of domain architecture. In P. vulgata, six of the differentially regulated IAPs lack RING domains (Data S4 and S6). Since this domain confers E3 ubiquitin ligase activity involved in ubiquitin-mediated protein degradation (Budhidarmo and Day 2015), this suggests that these IAPs contribute to the heat-stress response independently of this pathway. By contrast, although the differentially regulated IAPs in B. glabrata span six distinct domain architectures, the majority (17 out of 29) retain RING domains, pointing to a more prominent role for canonical E3 ligase activity in the response to parasitic infection. Together, these findings indicate that IAP diversification is not only widespread across molluscan lineages but also functionally mobilized in a stress-specific manner, involving distinct subsets of paralogous genes and domain architectures. This supports the view of the IAP family as an adaptable molecular toolkit enabling mollusks to cope with heterogeneous and fluctuating environmental challenges.

Figure 4.

Differential expression of IAP genes indicating functional diversification in selected metazoan species with an expanded IAP repertoire. a) Expression of IAP genes in the common limpet Patella vulgata maintained at control temperature (16 °C) or exposed to progressive warming from 16 to 43 °C over 3 d (data from Moreira et al. 2021). b) Expression of IAP genes in the freshwater snail Biomphalaria glabrata at 40-d post-infection with Schistosoma mansoni, Echinostoma paraensei (digenetic trematodes), or Daubaylia potomaca (a nematode parasite of planorbid snails) (data from Lu et al. 2024). c) Expression of IAP genes in the hepatopancreas of the Louisiana crawfish Procambarus clarkii under control conditions (7 mg ml⁻¹ dissolved oxygen) and severe hypoxia (1 mg ml⁻¹ dissolved oxygen) (data from Zhang et al. 2022). d) Expression of IAP genes in the wing primordia of pea aphids, Acyrthosiphon pisum, destined to become either winged or wingless morphs (data from Zhou et al. 2023). e) Expression of IAP genes in A. pisum across three sexual morphs: males, sexual oviparous females, and asexual parthenogenetic females (data from Jaquiéry et al. 2013). In all panels, heatmaps show Z-score–normalized expression levels of IAP genes, with blue indicating lower expression and red indicating higher expression. Normalization was performed per gene to reduce baseline expression differences and emphasize condition-dependent variation. Gene identifiers are color-coded according to their orthogroups for species included in the OrthoFinder analysis. An asterisk (*) denotes IAP genes whose expression was significantly different in the tested condition compared with the condition used as reference, which is framed in black.

We then expanded our analysis to more distant lineages, focusing on decapods, the only other organisms among the 26 shallow-water or air–water interface species whose response to stresses similar to those studied in bivalves have been examined (Li et al. 2022; Liu et al. 2022; Zhang et al. 2022). In this group, IAP regulation upon stress appears more restricted than what was observed in mollusks: only 4 of 75 IAPs were differentially regulated under hypoxia in the Louisiana crawfish Procambarus clarkii (two from the expanded OG0000153 orthogroup, and one each from the conserved OG000757 and OG0008528 orthogroups), whereas no IAP responded to different salinities in P. chinensis shrimps (Fig. 4c and Data S6). Likewise, no significant change in IAP expression was detected in this species following viral infection. The more limited or condition-dependent expression of IAPs in decapods suggests that the functional consequences of IAP expansion may vary substantially among lineages and that the stress-specific transcriptional response observed in mollusks is not universal across all aquatic invertebrates.

Although terrestrial arthropods occupy environments that differ markedly from the predominantly aquatic habitats of mollusks and decapods, they too experience substantial environmental stress, including temperature extremes, desiccation, or exposure to pathogens (Chown and Nicolson 2004). Despite this, analysis of available RNA-seq libraries (Zhang et al. 2019) suggests that they likewise do not exhibit the kind of stress-specific transcriptional mobilization of IAPs observed in mollusks. In the pea aphid A. pisum, eg, our results show that neither heat stress nor infection by the entomopathogenic fungus Beauveria bassiana induces differential regulation of IAP family members (Data S6). This lack of inducible IAP regulation under canonical abiotic and biotic stressors suggests that, in some terrestrial arthropods, expanded IAP repertoires may be maintained for functions other than acute stress responsiveness.

In the case of the 13 terrestrial arthropod species with more than 10 IAPs that also serve as vectors of animal or plant pathogens, we hypothesized a link between expansion of their IAPs and their vectoring ability. No prior study, to our knowledge, had investigated whether IAP expression changes in the presence of the microorganisms they transmit. By mining public RNA-seq repositories (Villar et al. 2015; Widana Gamage et al. 2018; Lu et al. 2019; Driscoll et al. 2020; Chesnay et al. 2022; Mahanta et al. 2022; Pandey et al. 2024), we identified relevant datasets for six of these species (the melon thrips Thrips palmi, the aphids A. pisum, A. gossypii, and Myzus persicae, the cat flea Ctenocephalides felis, and the tick Ixodes scapularis) examining their response to the presence of viruses or bacteria. Our analysis revealed clear but pathogen-dependent patterns. In C. felis and I. scapularis, bacterial infections did not alter IAP expression, suggesting that, in these species, IAPs may not be central to antibacterial responses under the conditions tested. By contrast, multiple IAPs were differentially expressed following viral exposure in T. palmi and all three aphid species (Data S6). Strikingly, both the identity and number of responsive IAPs depended not only on the viral agent but also on the host plant. For example, T. palmi differentially expressed four IAPs when fed on Chenopodium quinoa infected with groundnut bud necrosis virus (GBNV) but showed no IAP response to capsicum chlorosis virus (CaCV)-infected cowpea. Similarly, in A. gossypii, feeding on CLRDV-infected cotton did not elicit any change in IAP expression, whereas distinct sets of two to three IAPs were differentially expressed when the same aphids fed on CLRDV-infected hibiscus, okra, or prickly sida. Importantly, this context-dependence is not restricted to IAPs. Indeed, Mahanta et al. (2022) found 2,260 genes uniquely expressed in the GBNV-exposed T. palmi, while 347 genes were unique to the CaCV-exposed T. palmi. Likewise, among the 2,942 differentially expressed genes in aphids that acquired CLRDV from different host species, Pandey et al. (2024) found only four genes in common, despite the host plants belonging to the same family (Malvaceae). Such variation in transcriptional responses is likely influenced by virus-induced, host plant-specific phenotypic changes, such as alterations in nutrient composition, physiology, or defense pathways, which themselves depend on host susceptibility to infection. Consistent with this interpretation, CLRDV-infected cotton displays visible disease symptoms, whereas infected hibiscus, okra, and prickly sida do not (Pandey et al. 2024). Together, these results reveal a highly specific regulatory landscape shaped by the interaction between vector, pathogen, and host plant. This complexity suggests that IAP copy-number expansion may facilitate fine-tuned, context-dependent responses, even if the particular stressors eliciting such responses differ among species and ecological settings.

Beyond its role in stress response and host-plant interactions, IAP expansion may support more specialized physiological innovations. Aphids represent an ideal group to investigate this. They have extremely expanded IAP repertoires, with evidence of aphid-specific duplications and architectures. They are also characterized by numerous polyphenisms, such as transition between parthenogenetic and sexual reproduction or between winged and wingless morphs (Ogawa and Miura 2014; Calevro et al. 2019), in which aphid-specific gene duplicates appear to be involved. Indeed, gene families specifically expanded in aphids have been shown to be differentially expressed between different morphs (sexual vs asexual and winged vs wingless) or between mothers that produce different morphs (Ogawa and Miura 2014). Moreover, Zhou et al. (2023) showed that wing polyphenism involves programed cell death of wing primordia in wingless-destined individuals, suggesting a role for the regulation of apoptosis in this process. We can thus hypothesize a link between the remarkable diversification of the aphid IAP repertoire and the polyphenisms that characterize their life cycle.

To test this hypothesis, we analyzed RNA-seq datasets examining A. pisum transcriptomic response to environmental cues known to induce the switch from wingless to winged individuals (eg crowding; Zhou et al. 2023), as well as datasets comparing gene expression between different sexual morphs (Jaquiéry et al. 2013). We found that two IAPs, one from OG0000757 and another from OG0000214, are differentially expressed in the primordial wing discs of A. pisum individuals destined to become winged versus those that will develop into wingless morphs (Fig. 4d and Data S6). These two IAPs correspond to the previously annotated Ap-Deterin-1 and Ap-IAP-A1, respectively, with the latter belonging to a group characterized by the presence of two tandem RING domains, a characteristic of aphid IAPs (Ribeiro Lopes et al. 2020). Because the switch from wingless to winged morphs requires to block flight muscle degeneration, our results support a role for IAPs in maintaining tissue integrity during wing development. In addition, several IAPs exhibit morph- or sex-specific expression patterns, with distinct members upregulated in males, sexual females, and asexual females (Fig. 4e and Data S6). Notably, the four pea aphid members of OG0000214, corresponding to the Ap-IAP-A group defined by Ribeiro Lopes et al. (2020), show distinct expression biases: Ap-IAP-A1 (LOC103310098) is overexpressed in males, Ap-IAP-A2 (LOC100159652) and Ap-IAP-A3 (LOC100168361) in sexual females, and Ap-IAP-A4 (LOC100168556) in asexual females. Similarly, within OG0000306, which contains all members of the aphid-specific and highly expanded Ap-IAP-C group, two genes are overexpressed in males relative to both female morphs, while a third is upregulated only relative to sexual females. This suggests that morph-specific expression does not result from the exclusive co-option of a particular orthogroup and that functional diversification occurs within each expanded group. Such finely partitioned regulatory profiles suggest that expanded IAP families may contribute to the maintenance and modulation of developmental plasticity, highlighting their potential involvement in lineage-specific physiological innovations beyond canonical stress responses.

Another interesting hypothesis concerns the possible involvement of expanded IAPs in mutualistic symbiosis. Recent studies by Ribeiro Lopes et al. (2020) and Argandona et al. (2023) showed that in two aphid species, M. persicae and A. pisum, several IAPs, including many from the aphid-expanded groups, are significantly upregulated in the bacteriocytes compared with other tissues. Bacteriocytes are specialized cells found in insects that rely on nutritionally unbalanced diet; they host obligate mutualistic symbionts that supplement missing nutrients (Luan 2024). Importantly, IAP expression in these cells increases over time in a tissue-specific manner, which has led to the hypothesis that IAPs might help prevent apoptosis and thus ensure the long-term maintenance of both the bacteriocytes and the symbionts they contain until the death of the insect (Simonet et al. 2018; Ribeiro Lopes et al. 2020). Such symbiosis-related functions may have evolved independently multiple times. Indeed, Sun et al. (2017) also showed a high expression of many IAPs in the gills of the deep-sea vent mussel Bathymodiolus platifrons, which host chemoautotrophic endosymbionts capable of reducing compounds such as methane for the host's benefit.

Together with our finding that duplication dynamics vary markedly across lineages, including clear cases of lineage-specific expansions and diversifications, these results reinforce the conclusion that distinct selective pressures independently drove IAP expansion in multiple clades, likely promoting lineage-specific functional diversification. The canonical role of IAPs as docking platforms that promote substrate ubiquitination and degradation, may have facilitated their repeated co-option into new pathways and physiological processes. Understanding the temporal and contextual factors that led to the emergence of such complex IAP repertoires could provide insight into the molecular and cellular functions of IAPs and their role in non-model metazoans.

Metazoans have a core set of IAPs that has been conserved throughout evolution

To gain insight into the evolutionary history of metazoan IAPs, we first focused on a subset of 41 species (hereafter referred to as the “META” dataset, Fig. 5), each representative of one taxonomic class in the full dataset. This corresponds to 509 IAPs and 52 distinct domain architectures (Data S3 and Fig. S9a). Due to the variability in domain architecture, it is not possible to infer the phylogeny of IAPs from their entire sequence. To overcome this difficulty, we analyzed the only region common to all IAPs, the BIR domains, clustering sequences to limit redundancy and inferring their phylogeny (see Note S1 for details). Phylogenies obtained using FastTree allowed us to delineate five monophyletic subgroups of BIR sequences (labeled as A to E) (Fig. 6 and Data S7). Subgroups A, B, C, and E were also recovered in trees generated with IQ-TREE. There were few minor inconsistencies between the two phylogenies. Notably, subgroup D was split into two in the IQ-TREE phylogeny, with sequences from vertebrates forming a nested cluster within the E subgroup (Fig. S10). Membership of a subgroup was taken as an indication of a possible common origin.

Figure 5.

Domain architecture and distribution of IAPs across 14 phyla of metazoans. Distribution of IAP domain architectures in the 41 metazoan species belonging to the “META” subdataset. The species tree was rooted with non-bilaterian animals and, for each node, bootstrap values are indicated as round symbol whose size is proportional to the bootstrap value. Species are color-coded according to their taxonomic group (purple, non-bilaterian; blue, deuterostomes; orange, Lophotrochozoa; green, Ecdysozoa). Architectures indicated by an asterisk (*) correspond to representative architecture for groups framed in dotted line in Figure S9a. The black bars below the table indicate the percentage of species where a given architecture was found. Architectures were ranked hierarchically according to these percentages. Numbers within the table correspond to the number of IAP homologs with the above architecture, in each species. Black bars at the right of the table indicate the number of putative IAP homologs identified in each species. Amphimedon queenslandica has two identical survivin/Deterin-like IAPs, but one of them may be the result of an assembly error as it is partial and present on a very short scaffold.

Figure 6.

Phylogenetic relationships of BIR domains from metazoan IAPs. Phylogenetic tree of selected BIR domains obtained with FastTree, after clustering and filtering (see Methods section), from IAPs identified in the “META” subdataset. The tree is rooted using BIR sequences from survivin/Deterin-like IAPs (group A, in red) and, for each node, local support values are indicated as round symbols whose size is proportional to the value. Leaf labels are color-coded according to the phylum of the species to which the corresponding IAP belongs and BIRs are grouped into five distinct subgroups (A to E) indicated by different branch colors in the phylogenetic tree. From outer to inner circles: (i) BIR domain subgroups, (ii) presence of amino acids thought to prevent binding of caspases and IAP antagonists (Cossu et al. 2019) indicated by red checkmarks, and (iii) domain composition of the corresponding IAP (red squares, BIR; green triangles, RING; dark blue circles, UBC; orange stars, NACHT). Only domains commonly found in IAPs are shown. The position of the filled red square in relation to the other unfilled squares indicates the respective position of the BIR domain used in the phylogeny in relation to the other BIR domains that make up the IAP under consideration. For each subgroup, the main architecture and the position of the BIR domain (in red) are represented near to the subgroup label. Sequences from B. glabrata clustering closely in the tree are used as an example to illustrate lineage-specific intra-IAP BIR domain or IAP sequence duplications. These BIRs belong to IAPs with distinct domain architectures or different positions within the IAPs (eg C-terminal vs. N-terminal). C. int, Ciona intestinalis; Gnath., Gnathostomes.

Subgroup A contains long BIRs that share similar exon/intron organization (Fig. 7a). These domains belong to IAPs with similar motif composition (Fig. S11a) and, with few exceptions (two BIRs in C. elegans and an additional domain not found in any other IAP in Schistosoma mansoni and Beroe ovata), a similar domain architecture consisting of a single BIR domain. These IAPs include human survivin and Drosophila Deterin (Fig. 1) as well as members of the conserved OG0000757 orthogroup (Data S2). Accordingly, these IAPs were classified as survivin/Deterin-like. All the species in the “META” subdataset have a clearly identified survivin/Deterin-like IAP gene. The only exception is H. robusta, for which a putative survivin/Deterin-like IAP was nevertheless found using tblastn against its raw genomic sequence despite the absence of an annotated gene (Fig. 5, Fig. S12 and Data S8). Survivin/Deterin-like IAPs show a relatively high degree of sequence conservation given the large evolutionary distances between the organisms being compared (45% identity and 76% coverage over the full protein length, 66% and 96% on the BIR domains) and previous studies have highlighted a conserved regulatory role in both cell cycle and apoptosis from sponges (Luthringer et al. 2011) to mammals (Wheatley and Altieri 2019). However, substitutions of key amino acids involved in Zn²⁺ chelation or fold stability (Berthelet and Dubrez 2013; Cossu et al. 2019) (eg in Amphibalanus amphitrite), and the presence of residues that have been shown to interfere with IBM binding in the BIR domain (Eckelman et al. 2008) (eg in Daphnia magna), indicate that some survivin/Deterin-like IAPs should not be able to properly fold or interact with caspases and IAP antagonists (Fig. 7a). Furthermore, some of the phosphorylation sites identified in humans as playing an important role in IAP function (Wheatley and Altieri 2019) are not conserved across species (Fig. 7a). Taken together, this suggests that functional diversification of survivin/Deterin-like IAPs has occurred in some animal lineages.

Figure 7.

Residues and exon-intron junction conservation in metazoan BIR domains. Sequence alignment of the BIR domains from survivin/Deterin-like a), Bruce-like IAPs b), and the C-terminal BIR domains of RING-containing IAPs with either three c) or two BIRs d). Conserved motifs are highlighted in red when fully conserved and written in red when partially conserved. Exon/exon ligation sites are indicated by black triangles. Phosphorylation sites that, in humans, play an important role in protein function are indicated with an asterisk (*). Position at which the presence of specific residues (K, R, L, V) has been predicted to interfere with IBM binding are indicated by a filled black circle (•). Exons are schematically represented under each alignment as rectangles, with identical colors for exons that are predicted to have a common origin.

Subgroup B also contains long BIRs that share similar exon/intron organization (Fig. 7b) but, contrary to subgroup A, they are always associated with a UBC domain. The sequence from P. caudatus is the only exception, but the corresponding IAP is truncated in this species. IAPs with a BIR from the B subgroup include human BIRC6 and Drosophila Bruce (Fig. 1) as well as members of the OG0008528 orthogroup (Data S2). They were classified as Bruce-like IAPs and are present in only 28 of the 41 species included in the “META” subdataset (Fig. 5). While their large size (averaging 4.825 amino acids) might make them difficult to annotate, especially in the case of genomes that consist of multiple unplaced scaffolds, tblastn searches against raw genomic sequences allowed the identification of putative Bruce-like IAP genes in only three additional species: T. adhaerens, H. robusta, and Penaeus vannamei (Data S8). In each case, the query sequences partially aligned with genes containing a UBC domain. Notably, the species lacking Bruce-like IAPs are distributed across the entire metazoan phylogenetic tree, suggesting multiple and independent secondary losses (Fig. 5). In Drosophila and humans, Bruce/BIRC6 inhibits apoptosis through BIR-dependent interaction with the IBM motif of IAP antagonists (Domingues and Ryoo 2012; Dietz et al. 2023). BIR domains of Bruce-like IAPs are highly conserved (78% identity and 99% coverage, on average) (Fig. 7b), with no residues that could prevent interaction with the IBM, suggesting an ancient and shared function of Bruce-like IAPs in apoptosis inhibition across metazoans. In addition, despite their large size and varying exon numbers (33 to 78), Bruce-like IAPs share several conserved motifs (Fig. S11b), potentially serving as receptor sites for client proteins (Ehrmann et al. 2023). Investigating the conservation of these motifs in each Bruce-like IAP may help predict the proteins they can interact with. Interestingly, BIR from subgroups B and A share several features, ie they are longer and share several exon/intron junctions (Fig. 7a and b) and amino acid signatures, suggesting that they may be closely related.

Subgroups C to E contain short BIR domains, often associated with a RING domain, in IAPs that have one to six BIRs, three being the most common (29/41 species) (Fig. 5). When IAPs have several BIRs, the C-terminal domain is primarily found in the C subgroup, whereas more N-terminal BIRs are often part of the E subgroup (Fig. 6 and Data S7). Members of this subgroup can be divided into subgroups E1 and E2 based on the position within IAPs and amino acid composition: unlike E1 BIRs, E2 BIRs often have amino acids that have been shown to interfere with IBM binding at key position in the IBM binding groove and, in IAPs with three BIRs, are typically found at the N-terminal end (Fig. 6). Importantly, RING-containing IAPs can vary in composition. For instance, for IAP with three BIRs, the composition is E2-E1-C in most animals (such as human XIAP and Drosophila DIAP2) but E2-E2-E2 in Amphimedon queenslandica and E2-C-C in C. gigas or L. anatina (Data S7). This implies that convergent evolution sometimes produced IAPs with similar architectures (ie similar domain numbers and arrangements), but with potentially distinct functions (ie domain pairs are not phylogenetically close).

Overall, this analysis suggests that some IAPs architectures originated early in metazoan evolution and are conserved across species, possibly playing evolutionarily conserved physiological roles. Consistently, most animals have a core set of IAPs constituted of a survivin/Deterin-like, a Bruce-like, and at least one RING-containing IAP.

The core set of IAPs appeared before the bilaterian divergence

To better understand the origin of IAPs, we compared the BIR domain phylogeny (Fig. 6) with the species tree (Fig. 5), first focusing on early diverging metazoan lineages (ctenophores, poriferans, placozoans, and cnidarians). The six species in our dataset belonging to these phyla contain one to six IAPs. They all possess a survivin/Deterin-like IAP (Fig. 5), suggesting that it existed in the last common ancestor of Metazoa. Combined with the facts that yeast IAPs are survivin homologs (Cao et al. 2008; Gebreegziabher Amare et al. 2022) and survivin/Deterin-like proteins play an essential role in cell division (Wheatley and Altieri 2019), this suggests that the BIR domain from survivin/Deterin-like proteins is the ancestral BIR from which all others were derived. This was previously proposed by Cao and colleagues, although their study did not include any non-bilaterian animal (Cao et al. 2008).

Bruce-like IAPs were only found in Cnidaria and Porifera (Fig. 3). A tblastn search using known Bruce homologs against the T. adhaerens (Placozoa) genome also revealed a potential unannotated Bruce-like IAP (Data S8). Conversely, no Bruce-like IAPs were found in Ctenophora genomes, or reported in previous studies focusing on fungi. This suggests that Bruce-like IAPs are ancient in animals and may have evolved from survivin/Deterin-like IAPs, consistent with similarities between BIRs from subgroups A and B (Fig. 7). Bruce-like IAPs may have originated from duplication of a survivin/deterin-like gene, followed by recombination with a UBC-containing gene. In support of this, we identified a unique yeast-like IAP in the choanoflagellate Salpingoeca rosetta (PTSG_01224, Data S9), and found significant alignment with a long UBC-containing protein (PTSG_00516) using BIR domains from subgroup A as a query. Although it clearly occurred early in metazoan history, the exact origin of Bruce-like IAPs remains uncertain because there is no consensus on the order in which the first animal lineages diverged. Under the “ctenophore-sister” theory (Schultz et al. 2023), the hypothesis would be that these IAPs would have appeared after Ctenophora's, but before Porifera's, divergence. Alternatively, if Porifera diverged first (Redmond and McLysaght 2021), Bruce-like IAPs likely arose in the common animal ancestor and were secondarily lost in Ctenophora. Sequencing and annotation of a greater diversity of non-bilaterian animal genomes will be necessary to distinguish between these two hypotheses.

Poriferans and cnidarians also have at least one RING-containing IAP with either two (Nematostella vectensis) or three (Hydra vulgaris and A. queenslandica) BIRs, while none was found in ctenophores and placozoans (Fig. 5 and Note S2). This shows that, similar to Bruce-like IAPs, RING-containing IAPs are ancient in metazoans, and originated either in the common ancestor and were lost in ctenophores or, alternatively, emerged after ctenophore divergence. In both cases, they have been lost in T. adhaerens. The presence of IAPs with three BIRs in both poriferans and cnidarians would imply that this is the ancestral architecture, but further analyses, with more representatives of the Porifera phylum, are needed to confirm this hypothesis. This would also help determine whether these IAPs are similar to XIAP/DIAP2-like proteins of other animals, as our study highlights distinct domain composition that could indicate convergent evolution (Fig. 6 and Data S7). Interestingly, RING-containing IAPs in the jellyfish Aurelia coerula were recently found to activate, rather than inhibit, caspases, a trait also observed in hydra, coral, and sponges (Chen et al. 2025). These functional differences from bilaterian IAPs support the hypothesis of an independent evolutionary origin.

The BIRs from RING-containing IAPs of cnidarians and poriferans are all encoded by a single exon suggesting they originated via an initial event of retrotransposition. In contrast, in XIAP/DIAP2-like proteins of bilaterians from the “META” subdataset, the N-terminal and middle BIR domains (from subgroup E2 and E1, respectively) remain encoded by a single exon, while the C-terminal BIR (subgroup C) is split over multiple exons (three in chordates, two in other bilaterians) (Fig. 7c). The conserved exon/intron junction resembles that of Bruce-like BIRs. While independent emergence of XIAP/DIAP2-like proteins in bilaterians and non-bilaterians cannot be ruled out, it is also possible that the ancestral gene evolved via intron transfer after a recombination event (Roy and Gilbert 2006) predating the protostome-deuterostome split over 600 million years ago (Erwin and Davidson 2002).

Overall, this analysis suggests that the core set of metazoan IAPs emerged before the bilaterian divergence. However, bilaterians have a more extensive and structurally diverse IAP repertoire (Fig. 5), with deuterostomes and protostomes exhibiting distinct features indicative of group-specific diversification events.

Deuterostome IAPs have evolved through insertion of new domains

There is no sign of IAP duplication in the common ancestor of Deuterostomia, but we found evidence of it in specific lineages, in particular in Anneissia japonica (29 IAPs), X. laevis (24 IAPs), and Ciona intestinalis (18 IAPs). In A. japonica, the similarity among BIR domains of distinct IAPs (eg cluster 3, Data S10), coupled with the highly fragmented genome assembly (Data S11) and presence of IAP-encoding genes on short scaffolds, suggests that some amplifications may be the result of assembly artifacts. In X. laevis, genome tetraploidy may have facilitated the expansion (Session et al. 2016), but it cannot be the sole explanation, as its diploid relative Xenopus tropicalis already presents an amplification of its IAP repertoire compared with other chordates, with 13 IAPs (Data S9). Other amphibians like Bufo bufo and Rana temporaria do not show this expansion (eight and seven IAPs, respectively, similar to other chordates; Data S9) suggesting that IAP expansion in Xenopus species is lineage-specific. Another specificity of Xenopus is the presence of FIIND-containing IAPs (eight in X. laevis and four in X. tropicalis), in which BIR domains associate with FIIND and CARD domains (Fig. 5). FIIND-CARD associations (without BIR domains) have previously been reported in non-IAP proteins like CARD8, which facilitates pro-caspase-1 recruitment to the inflammasome (Chui et al. 2020), and CARD9, which suppresses pro-caspase-9 activation in cancers (Pathan et al. 2001). The function of the FIIND domain remains unknown. All BIR domains from FIIND-containing IAPs form a well-supported cluster within clade E2 (Fig. 6 and Fig. S13) and, in several IAPs, adjacent BIR domains also group together, eg, the five downstream BIR domains of LOC121398946, indicating that they are more closely related to one another than to any other BIR domain. This pattern is consistent with multiple intra-IAP tandem duplications. In the case of C. intestinalis, BIR sequences cluster into multiple independent groups in the tree (Fig. 6). Combined with the fact that C. intestinalis is the only deuterostome lacking a Bruce-like IAP, this suggests significant divergence in its IAP repertoire. This aligns with previous findings of a reorganization of the apoptosis regulatory machinery in ascidians, possibly linked to the loss of caspase-9 and Apaf-1, two critical members of the apoptosome, in Tunicata (Krasovec et al. 2024).

Unlike A. japonica, other Ambulacraria in our dataset show no significant changes in their IAP repertoire, with no ancestral duplications or group-specific architectures observed. Chordates, on the other hand, have at least six IAPs and several architectures are conserved in multiple species. For instance, 10 of the 13 chordates in our dataset possess a RING-containing IAP with two BIR domains, from the E1 and C subgroups (Data S7). This architecture is likely derived from duplication of the ancestral XIAP/DIAP2-like IAP, as suggested by significant alignments against the 2-BIR IAPs with tblastn search using XIAP/DIAP2-like sequences as query. The amino-terminal part of the query sequences did not align, suggesting either a duplication event followed by loss of the N-terminal domain or a partial duplication from the outset. This architecture is also observed in several protostomes. It may have either originated in the bilaterian ancestor and been lost in Ambulacraria, or evolved independently in chordates and protostomes. The observations that the C-terminal domain of these IAPs has an additional exon-intron junction in chordates compared with protostomes (Fig. 7d) and that this junction is common with one found specifically in XIAP/DIAP2-like IAPs of chordates support the latter hypothesis. Further evidence for independent origins comes from the observation that, at least for the species considered in our OrthoFinder analysis, IAPs with this architecture are assigned to different orthogroups in chordates (OG0000153) and insects (OG0000214) (Data S2). In humans, the BIR from the E1 subgroup was secondarily lost, giving rise to an IAP with a single BIR associated with a RING (BIRC7/ML-IAP).

Gnathostomes (jawed vertebrates) also display several features that set them apart from other chordates. First, except for humans, they have two survivin/Deterin-like homologs (four in X. laevis), suggesting duplication of the ancestral IAP. Humans and the mouse Mus musculus (Cao et al. 2008) only have one survivin/Deterin-like homolog suggesting a secondary loss in mammals. Gnathostomes also possess an additional XIAP/DIAP2-like IAP, with BIRs belonging to the same subgroups at the same relative position, but with one CARD domain inserted in-between the three BIRs and the RING domain (eg human cIAP1 and cIAP2 proteins) (Fig. 5). It probably arose through duplication of the ancestral XIAP/DIAP2-like IAP, followed by insertion of a CARD domain as a two-exon module upstream of the RING-encoding exon. While an earlier study based on 11 organisms suggested that this duplication occurred after invertebrates and vertebrates diverged (Cao et al. 2008), our data indicate a more recent origin within the gnathostome lineage. NAIP-like IAPs, characterized by a nucleotide-binding NACHT domain and a leucine-rich repeat (LRR) at the C-terminus, also appear specific to gnathostomes. The aforementioned study proposed that these IAPs are unique to mammals (Cao et al. 2008) but our findings reveal that they are also present in non-mammalian gnathostomes. We also found that they do not all have three BIRs (they have between one and five, Fig. 5), although this is the most common architecture, and may even contain two NACHT and LRR domains. The phylogeny of BIR domains reveals that N-terminal BIRs of NAIP-like IAPs mostly fall into subgroup D, while others belong to subgroup C, forming a distinct well-supported cluster (Fig. 6, Fig. S14 and Data S7). This indicates that the C-terminal and directly upstream BIR domains in NAIP-like IAPs are more closely related to each other than to any other domains. Thus, although it is uncertain if NAIP-like IAPs arose from an XIAP duplication, intra-IAP BIR tandem duplications likely contributed to their current diversity. In mammals, NAIP and cIAP proteins regulate immunity and inflammation through interactions with inflammatory caspases (Labbe et al. 2011; Maltez and Miao 2014) and these proteins likely evolved alongside the innate immune system's increased complexity in jawed vertebrates (Litman et al. 2010). Their absence in the lamprey P. marinus suggests that they emerged after cyclostomes diverged from gnathostomes. However, this hypothesis is tentative due to the incomplete P. marinus genome (78.5% completeness, using the vertebrata_odb10 BUSCO dataset) and the scarcity of annotated cyclostome genomes in public databases. Overall, these results indicate that the gnathostome common ancestor probably had an IAP repertoire composed of seven proteins: two survivin/Deterin-like, one Bruce-like, one XIAP-like, one cIAP-like, one NAIP-like, and one with two BIRs and one RING.

Protostome IAPs have evolved through the dynamic gain and loss of BIR and RING domains

The “META” subdataset includes a comparable number of deuterostome and protostome species, but the latter exhibit greater variation in IAP numbers, with notable expansions (eg bivalves) and reductions (eg nematodes and platyhelminthes) (Fig. 5). To refine insights into evolution of IAPs in this interesting animal clade, we extended our phylogenetic analysis to the 289 protostomes with a RefSeq annotation (“PROTO” subdataset; Fig. 8). This corresponds to 2,660 IAPs (Data S4) whose BIRs form four monophyletic subgroups (A, B, C, and E; Fig. 9 and Data S12) with characteristics similar to those with the same name in the metazoan BIR tree. Notably, in the tree obtained using IQ-TREE from the same dataset (Fig. S15), subgroup E was split in two subgroups that broadly correspond to subgroups E1 and E2, once again highlighting the differences between these two types of BIR domains. Importantly, despite the large number of BIR domains in the “PROTO” subdataset, the protostome BIR tree is congruent with the metazoan BIR tree, since 208/230 (90.4%) of the BIR sequences from the 18 protostomes present in both datasets were assigned to the same subgroup. Invertebrate sequences from subgroup D were reassigned to subgroup C or E in the second tree (Fig. S16). Four architectures, resembling the well-studied and functionally characterized IAPs of D. melanogaster (ie Deterin, Bruce, DIAP1 and DIAP2), are found in over 70% of protostome species (Fig. S9b). Investigating whether these structurally similar IAPs share a common origin could provide insights into their roles.

Figure 8.

Domain architecture and distribution of IAPs across 289 species of protostomes. Distribution of the seven most common IAP domain architectures in the 289 species belonging to the “PROTO” subdataset. The phylogenetic tree was rooted with lophotrochozoan animals and, for each node, bootstrap values are indicated as round symbol whose size is proportional to the bootstrap value. Branch and leaf labels are color-coded according to the taxonomic group of the corresponding species. Architectures indicated by an asterisk (*) correspond to representative architecture for groups of similar IAPs (architecture group framed in dotted line in Figure S9b). The black bars on the left indicate the percentage of species where a given architecture was found. Architectures were ranked hierarchically according to these percentages. Black bars in the outer circle of the tree indicate the number of putative IAP homologs identified in each species. Black squares indicate that the corresponding architecture has been found in the considered species.

Figure 9.

Phylogenetic relationships of BIR domains from protostome IAPs. Phylogenetic tree of selected BIR domains obtained with FastTree, after clustering and filtering (see Methods section), from IAPs identified in the “PROTO” dataset. The tree is rooted using BIR sequences from survivin/Deterin-like IAPs (group A, in red) and, for each node, local support values are indicated as round symbols whose size is proportional to the value. Leaf labels are color-coded according to the taxonomic class of the species to which the corresponding IAP belongs and BIRs are grouped into four distinct subgroups (A, B, C, and E) indicated by different branch colors in the phylogenetic tree. From outer to inner circle: (i) BIR domain subgroups, (ii) presence of amino acids thought to prevent binding of caspases and IAP antagonists (Cossu et al. 2019) indicated by red checkmarks, and (iii) domain composition of the corresponding IAP (red square, BIR; green triangle, RING; dark blue circle, UBC). Only domains commonly found in IAPs are shown. For each subgroup, the main architecture and the position of the BIR domain (in red) is represented near the subgroup label. Lophotro, Lophotrochozoa; H. ill, Hermetia illucens; Ap, Aphids.

Survivin/Deterin-like IAPs, with a unique BIR from subgroup A, are present in 95.5% of species in the “PROTO” subdataset (Fig. 8 and Data S13) and the 13 species lacking this IAP are scattered across the protostome phylogenetic tree (Fig. 8). This absence could stem from species-specific losses (unlikely given the essential role of survivin homologs in cytokinesis (Wheatley and Altieri 2019) or from genome annotation errors. Supporting the latter possibility, tblastn searches against raw genomic data identified putative survivin/Deterin-like IAPs in 12 of these 13 species (Data S8). For example, although no survivin/Deterin-like homolog was annotated in Pieris brassicae (Lepidoptera), a tblastn search yielded a significant alignment against a genomic region syntenic with the Survivin/Deterin-like IAP locus in Pieris rapae (Fig. S17), indicating that a survivin homolog is likely present but unannotated in P. brassicae. Temnothorax curvispinosus, the only species for which no classical survivin/Deterin-like IAP was detected, harbors an IAP with two subgroup A BIR domains, which likely result from a tandem duplication event involving a survivin/Deterin-like IAP. A similar architecture is also observed in Caenorhabditis species, but not other nematodes. In Colias croceus (Lepidoptera) and Pediculus humanus corporis (Phthiraptera), an atypical association between the BIR domain and, respectively, a reverse transcriptase domain or a MIT domain was also observed (Data S4). Overall, all the species from the “PROTO” dataset may have a survivin/Deterin-like IAP.

In contrast, 44 species from the “PROTO” subdataset lack Bruce-like IAPs (Fig. 8 and Data S13) and tblastn searches against raw genomic data identified putative Bruce-like homologs for only 6 of them. Two out of these six species (the nematodes Loa loa and Opisthorchis viverrine) have a protein that shares several similarities with Bruce-like IAPs (size, presence of an UBC domain) but have no BIR domain. Conversely, several species from the “PROTO” subdataset (ie Homalodisca vitripennis, Nilaparvata lugens, Diaphorina citri, C. felis, and P. caudatus) possess an IAP with a BIR from subgroup B but no UBC domain. Analysis of the corresponding genes revealed that they probably arose either by the split of an ancestral gene or are the result of annotation errors, likely facilitated by the large size of Bruce-like IAPs, leading to the identification of partial proteins. In line with this, the Bruce-like BIR of P. caudatus is clearly identified as a partial gene, found at the end of a very short scaffold. The Bruce-like BIRs of N. lugens and C. felis, on the other hand, are both encoded by a short gene that is found directly upstream of a long gene that contains a “Baculoviral IAP repeat-containing protein 6” domain. H. vitripennis and D. citri are also found upstream of partial genes that bear resemblance with Bruce-like IAPs, despite no functional domain being identified. In spite of that, it is striking to note that Bruce-like IAPs appear to have been lost or degraded in several hemipteran insects (eg Bemisia tabaci retains a UBC-containing IAP, while homologs of Bruce in H. vitripennis, N. lugens, and D. citri are split, and Cimex lectularius, Halyomorpha halys, and aphids lack them entirely), in most nematodes (except Brugia malayi and Strongyloides ratti), and in all chelicerates. This hints to a relaxed selection pressure in these species. This was unexpected given the high level of sequence conservation of Bruce-like IAPs and their BIR domain across evolutionary time (Fig. 7b and Fig. S11b), reflecting its critical function in regulating apoptosis and autophagy in mammals and D. melanogaster (Ebner et al. 2018; Ehrmann et al. 2023). While disruption of these processes has not been reported in species lacking Bruce-like IAPs, research beyond Drosophila is limited in protostomes, and other proteins may compensate for the loss of Bruce homologs. Interestingly, the loss of Bruce-like IAPs correlates with the degradation or absence of the IMD immune pathway (a broadly conserved arthropod NF-κB immune signaling pathway) in hemipterans (Nishide et al. 2019) and chelicerates (Palmer and Jiggins 2015; Bechsgaard et al. 2016). Some IAPs, such as DIAP2 in Drosophila and cIAP2 in humans, have been shown to regulate immunity and NF-κB activation (Huh et al. 2007; Sharma et al. 2017).

In our dataset, 89% of protostomes have at least one DIAP1-like IAP, with two BIR domains from the E and C subgroups combined with one RING domain (Fig. 8 and Data S12). DIAP1 is often referred to as the major apoptosis inhibitor in D. melanogaster, since disruption of its activity or expression is sufficient to induce apoptosis in virtually all cells (Orme and Meier 2009). The two BIR domains of DIAP1 can interact with several caspases (ie Dronc, DrICE, and Dcp-1), which can then be degraded following RING-mediated ubiquitination (Tenev et al. 2005; Orme and Meier 2009). The presence of a DIAP1 homolog in most protostomes suggests that this function is maintained and underscores its biological importance. Importantly, two types of DIAP1-like proteins are found in protostomes. One is similar to the one of chordates and Drosophila, found in most species including mollusks, and a second one, limited to Mollusca and Brachiopoda, has an N-terminal domain with characteristics of subgroup E2, ie more similar to the N-terminal BIR domain of XIAP/DIAP-2-like IAP and featuring amino acids that have been shown to interfere with IBM binding in the IBM binding groove (Fig. 9). These results suggest that at least one additional and independent duplications of a XIAP/DIAP-2-like IAP happened after the divergence of platyhelminthes from other Lophotrochozoa, with loss of the domain in second position. Supporting this, we found evidence, in C. gigas and L. anatina for instance, for duplication of a XIAP/DIAP-2-like IAP followed by the loss of a ∼1,200 bp region containing the BIR domain in second position, probably through unequal crossing-over (Fig. S18). The Lophotrochozoa-specific DIAP1-like IAPs should have interaction potentials distinct from those of the canonical DIAP1 homologs and may have evolved new functions. Additionally, the N-terminal domains of DIAP1-like proteins in hymenopterans form a separate cluster in the tree (Fig. 9), which also includes N-terminal BIRs from Formicidae IAPs that contain only two BIRs (without RING domain), likely arising from a duplication of the hymenopteran DIAP1-like IAP. Supporting this, in Formicidae ants, DIAP1-like proteins and their two-BIR counterparts are systematically found in tandem, indicating that the latter emerged from a tandem duplication of the former. Moreover, there is no evidence for the subsequent loss of the RING-encoding exon, indicating that the absence of the RING domain is more likely due to an incomplete duplication rather than a secondary loss.

The ancestral bilaterian XIAP/DIAP2-like IAP (with three BIR domains belonging to subgroups E2, E1, and C associated with one RING domain) was conserved in all deuterostomes. By contrast, it has been lost in 37% of protostomes, including all nematodes and platyhelminthes, Tephritidae flies, Halictidae bees, Acrididae grasshoppers, aphids, water fleas, and most Anopheles mosquitoes (Note S3). In D. melanogaster, DIAP2 functions primarily as a regulator of the NF-kB-related innate immune response and plays only a minor role in apoptosis, when flies are submitted to mild stress or following inhibition of DIAP1 (Leulier et al. 2006; Ribeiro et al. 2007). The lack of a XIAP/DIAP2-like IAP in Tephritidae flies, Acrididae grasshoppers and Anopheles mosquitoes can be explained by the loss of one domain: they have two DIAP1-like IAPs, one shared across their order and one that emerged after the secondary loss of the N-terminal BIR domain of a DIAP2-like IAP (Data S4 and S13). The anti-apoptotic function of DIAP2 relies on the specific binding of its BIR3 domain to DrICE (Ribeiro et al. 2007), and loss of the N-terminal domain should therefore only have a limited impact on its ability to inhibit apoptosis. Aphids also present several interesting features: (i) only three out of the nine species included in our dataset have an IAP with three BIRs (Fig. 8 and Data S4), (ii) contrary to what is observed in other insects, the N-terminal and middle domains of these IAPs are part of the same cluster within the E subgroup, suggesting that their three-BIR architecture arose through intra-IAP BIR duplication rather than through duplication of a DIAP2/XIAP-like IAP, (Fig. 9 and Data S12), and (iii) with only one exception (a sequence from A. gossypii), these IAPs, and their homologs in other aphid species, have two RING domains in tandem, a feature unique to aphids and likely resulting from an intra-IAP domain duplication in the common ancestor. Intra-IAP domain duplication has further driven the diversification in this IAP group: the aforementioned E subgroup cluster includes adjacent BIR pairs from M. persicae and A. gossypii sequences with more than three BIR domains, which are more closely related to each other than to the corresponding domains of their homologs. Members of this IAP subfamily, designated as Ap-IAP-A in a previous study (Ribeiro Lopes et al. 2020), have been shown to inhibit apoptosis more potently than homologs with only one RING. The four pea aphid representatives of this IAP subfamily are differentially regulated depending on the aphid morphs and specifically upregulated in aging bacteriocytes (see section Expanded IAP repertoire support lineage-specific functional diversification) supporting a functional specialization of these proteins.

The expansion of the protostome IAP repertoire is largely due to the duplication of XIAP/DIAP2-like and DIAP1-like IAPs, as seen in L. anatina with 14 XIAP/DIAP2-like proteins or M. yessoensis and Bombus vancouverensis nearcticus with 18 and 7 DIAP1-like proteins, respectively (Data S4 and S13). Furthermore, these gene duplications are often accompanied by the loss of BIR and/or RING domains, contributing to architectural diversification of the IAP repertoire. Notably, while loss of BIR domains from the E subgroup is frequent, domains from the C subgroup are typically retained and drive further duplications. Gastropods, bivalves and aphids, for instance, possess multiple IAPs with a unique C subgroup BIR domain, with or without RING domain (Data S13 and Fig. S19). In the black soldier fly H. illucens, tandem duplication of a C subgroup BIR led to architectures with up to five BIR domains (Fig. 9 and Data S12). In both D. melanogaster and mammals, members of the C subgroup (C-terminal BIR domains of DIAP1, DIAP2, XIAP, and cIAPs) have been reported to interact with caspases and IAP antagonists (Ribeiro et al. 2007; Eckelman et al. 2008; Berthelet and Dubrez 2013) and their amplification might underlie an elevated need for this specific function in some species.

Conclusion

Our study reveals that IAPs constitute a highly dynamic family of proteins, whose evolutionary history has been shaped by complex patterns of gene gains, losses, and structural variations, even among closely related species. By investigating the mechanisms and relative timing of IAP expansion across a broad taxonomic range, we show that duplication dynamics vary markedly among lineages: while tandem duplication is widespread, its contribution differs substantially across species, and other mechanisms can represent major, sometimes dominant, drivers of expansion in specific taxa. Indeed, transposition represent the major driver of IAP expansion in several hemipterans or decapods, while segmental duplication is dominant in X. laevis (chordata) or A. gifuensis (arthropoda). This shows that IAP expansion is not driven by a single process but instead reflects diverse evolutionary trajectories across metazoans. These trajectories range from ancient, conserved orthogroups with variable copy numbers to recent, taxon- or species-specific duplications, suggesting that distinct selective regimes govern IAP diversification in different lineages. Accordingly, our study reveals distinct functions for expanded IAPs across lineages. Gastropods and bivalves both show differential regulation of IAPs from multiple, evolutionarily distinct subfamilies following exposure to biotic and abiotic stressors. These results underscore the evolutionary diversification and stress-specific deployment of IAPs as a key molecular strategy enabling mollusks to cope with heterogeneous environments. By contrast, arthropods exposed to similar stressors present no, or a more limited, transcriptional response of IAPs, suggesting that, in these species, expanded IAP repertoires may be maintained for functions other than acute stress responsiveness. In line with this, we show that, in aphids, IAP expansion may support more specialized physiological innovations unrelated to immediate stress responses such as polymorphism and symbiosis.

To better understand how this diversity arose over evolutionary time, we used phylogenetic analyses to reconstruct the evolutionary history of IAPs, with a focus on the key steps underlying their diversification in early metazoans, deuterostomes, and protostomes. We showed that Survivin/Deterin-like, Bruce-like IAPs, and RING-containing IAPs have been conserved since before the divergence of bilaterian species, underscoring their essential role in metazoans (Fig. 10). However, their evolutionary trajectories differ. Survivin/Deterin-like and Bruce-like IAPs have undergone few duplications and remain architecturally stable, whereas RING-containing IAPs frequently expand and undergo architectural remodeling. This remodeling includes the acquisition of new domains (eg CARD, NACHT, FIIND) likely linked to immune and inflammatory regulation, especially in gnathostomes, as well as intra-IAP domain loss or duplication, observed in numerous protostomes (eg bivalves, gastropods, and various insect groups in particular aphids, ants, and Anopheles mosquitoes). These changes have repeatedly given rise to IAPs with similar architectures but independent origins, potentially resulting in distinct functional roles (Fig. 10). We also identify lineage-specific IAP losses, such as the near-complete absence of RING-containing IAPs in Placozoa, Nematoda and Platyhelminthes and the degradation of Bruce-like IAPs in hemipteran insects, nematodes, and chelicerates (Fig. 10). As homologs of these IAPs have been shown to play key roles in the control of apoptosis in model organisms, these losses suggest significant differences in apoptosis regulation that warrant further study.

Figure 10.

Proposed scenario for the evolution of metazoan IAPs. Cross, inclined bars, vertical arrows, and curved arrows represent predicted gene loss, domain loss, domain insertion, and domain duplication, respectively. When the architecture of the ancestral IAP at the origin of one the aforementioned events was identified, the symbol was colored accordingly (survivin/Deterin-like, red; Bruce-like, blue; XIAP/DIAP2-like, green; DIAP1-like, orange) or was left black if not. Similar architecture that arose multiple independent times across metazoan evolution are numbered. Specific examples of events that have marked the evolution of arthropods are shown in a box. Deutero., Deuterostomia; Proto., Protostomia; Lophotro., Lophotrochozoa; Ecdyso, Ecdysozoa.

Together, these findings provide an evolutionary framework for understanding how gene duplication and diversification contribute to the functional versatility of IAPs and highlight the need for future work linking lineage-specific expansions to their physiological and ecological roles.

Materials and methods

Sequence data collection

The analysis of IAP evolution was conducted at two taxonomic levels, by generating two distinct datasets: “META” and “PROTO”. The “META” dataset comprises sequences from 41 representative species, spanning the entire metazoan phylogenetic tree. It was generated to identify the ancestral repertoire of metazoan IAPs and the early stages of their diversification. The “PROTO” dataset is more comprehensive and was assembled to gain a better understanding of the major evolutionary steps that drove IAP diversification in the protostome group.

The two datasets were generated as follows. First, the proteomes of the 795 metazoans for which annotated genomes were available in the RefSeq database (the NCBI Reference Sequence database, https://www.ncbi.nlm.nih.gov/) at the time of analysis were downloaded (accession numbers and details on genomes are available in Data S11). The proteomes of two additional metazoans, from the Ctenophora phylum, for which there was no representative in the RefSeq database, were downloaded from dedicated databases (the Mnemiopsis Genome Project Portal, https://research.nhgri.nih.gov/mnemiopsis/ and the Beroe ovata genome release, https://github.com/josephryan/Beroe_genome/releases) (Data S11). In total, these 797 species represent 14 phyla and 41 classes (Data S1). To assess the quality of the corresponding annotated genomes, we analyzed the gene content of the 797 metazoan species with BUSCO (Benchmarking Universal Single-Copy Orthologs V 5.4.6) (Manni et al. 2021) using as a query the “metazoa_odb10” lineage-specific dataset, ie the set of 954 marker genes that are present in at least 90% of metazoan species, and present in a single copy in 90% of these species (Data S11).

For the “META” dataset, we selected one representative species for each of the 41 animal classes. D. melanogaster, C. elegans, and humans were chosen as representatives of insects, chromadorean nematodes and mammals, due to the extensive research carried out on the regulation of apoptosis in these species. For other animal classes, the species with the highest number of research articles on PUBMED was chosen, as this suggested a more comprehensive understanding of its ecological and physiological characteristics (Data S11). The genome of the selected species had BUSCO completeness scores above 85%, except those of C. elegans (78.5%), Mnemiopsis leidy (73.9%), B. ovata (73.1%), S. mansoni (70.9%), E. granulosus (69.2%), and T. spiralis (62.3%). These species were nevertheless retained in the analysis, because they were the only representatives of their respective class or belonged to a class in which other species had lower scores.

For the “PROTO” dataset, the 289 protostomes among the 797 aforementioned metazoan species were considered. Of these species, only 13 (the arachnid Stegodyphus dumicola, the hemipteran D. citri, all nematodes, and platyhelminthes) had BUSCO completeness scores below 85%. This dataset includes 18 species that were also present in the “META” dataset.

Overall, when considering the two datasets (“META” and “PROTO”), IAPs were analyzed in 312 species, which are listed in Data S1.

Identification of IAP proteins

For each of the 312 species, putative IAPs were identified using a combination of sequence similarity and HMM profile-based methods using, respectively, the BLASTP tool (V 2.8.1) (Camacho et al. 2009) and HMMsearch of the HMMer package (V 3.1) (Eddy 2011) (see the complete pipeline in Fig. S20). For the BLASTP search, we used as query the sequences of IAPs identified in previous studies in A. pisum, Aedes aegypti, Bombyx mori, C. elegans, Danio rerio, D. melanogaster, Gallus gallus, H. sapiens, Spodoptera frugiperda, and Trichoplusia ni (Data S14). For the HMM search, the curated Pfam BIR domain model (PF00653) was used to query the selected proteomes (Johnson et al. 2010). In all cases, a relaxed E-value of 0.1 was used as a threshold to ensure that no IAP was missed due to the distance between species, which could result in high divergence between IAP sequences.

Protein sequences identified by BLASTP or HMM were retrieved and further analyzed using InterProScan (V 5.59-97.0) (Paysan-Lafosse et al. 2023). Proteins were considered putative IAPs if they contained at least one BIR domain, ie if they provided a positive match with one of the following signatures: PS50143 (“BIR repeat profile”), PS01282 (“BIR repeat”), SM00238 (“bir_2”), or IPR001370 (“BIR repeat”). If other domains were predicted at the same position, only sequences with at least one clearly identified BIR domain (E-value of the BIR domain inferior to the E-value of other domains at the same position) were retained. In this way, 509 putative IAPs were identified in the “META” dataset and 2,660 in the “PROTO” dataset. These represented a total of 2,843 different IAPs, accounting for IAPs from species present in the two datasets (Data S1).

To identify potential gene prediction omissions that might cause IAPs to be overlooked in protein set searches, we performed additional tblastn searches using IAP query sequences from closely related species. These searches targeted the raw genomic scaffolds of species that were missing at least one protein from the core set of metazoan IAPs, ie a survivin/Deterin-like, a Bruce-like, or a RING-containing IAP. Significant hits from these tblastn searches were further examined by (i) verifying whether the alignment corresponded to the expected BIR domain and, when relevant, to additional conserved domains, ie RING (for RING-containing IAPs) or UBC (for Bruce-like IAPs); (ii) comparing the genomic context of the hits with that of their putative homologs in closely related species, specifically checking for conserved synteny (ie whether they are flanked by the same neighboring genes); and (iii) assessing their expression, when transcriptomic data were available, to determine whether these loci were transcriptionally active. To account for potential genome assembly errors or incompleteness, we also performed tblastn searches against the whole-genome shotgun contig database in the aforementioned species.

Characterization of duplication modes

To investigate the mechanisms underlying the expansion of IAPs, we selected 20 species possessing at least 10 IAPs and chromosome-level genome assemblies. Duplicated gene pairs in each selected species and their putative duplication modes were inferred with the R/Bioconductor package doubletrouble v1.10.0 (Almeida-Silva and Van de Peer 2025). Doubletrouble was run using the longest peptide isoform for each gene and default arguments (DIAMOND (Buchfink et al. 2021) similarity searches, top hit = 5 and E-value = 1e-10). Duplicates were categorized into four types: segmental (paralogs in syntenic blocks), tandem (adjacent repeats), proximal (duplicates located on the same chromosome within 10 genes of each other but not immediately adjacent), and dispersed (all other duplicates). Dispersed duplicates were further designated as putative retrotransposition events if one copy lacked introns.

To estimate the evolutionary divergence of duplicated genes, we constructed distance matrices for all pairwise comparisons of IAPs within each of the 20 species identified above. Pairwise sequence identities were computed using Biopython PairwiseAligner (Cock et al. 2009) in local alignment mode, using the longest transcript sequence for each gene and using an identity-based scoring scheme (match = +1, mismatch = 0; Durbin et al. 1998) with affine gap penalties (gap opening = −1, gap extension = −0.5; Gotoh 1982). For each transcript pair, the best local alignment was retained and all non-contiguous aligned blocks were included for identity calculations. Identity scores were summed across all aligned residues and normalized by the maximum possible score for the aligned length. With this strategy, relying solely on local alignment identity would overestimate the relatedness of sequences that share only short conserved regions but differ significantly in their overall sequence (Rost 1999). While standard comparative genomics and sequence clustering methodologies typically employ strict coverage thresholds to filter out partial alignments (Konstantinidis and Tiedje 2005; Li and Godzik 2006; Miele et al. 2011; Steinegger and Söding 2017), establishing a single arbitrary threshold was not suitable to capture the continuous spectrum of size and structural divergence in our dataset. Therefore, to penalize alignments spanning only small fractions of the sequences and mitigate the associated bias, we weighted sequence identity by coverage by defining a “dissimilarity score” calculated as follows: 1—(Identity × Coverage). This approach allows ranking transcript pairs while penalizing partial alignments. Because recently duplicated genes have had less time to accumulate mutations, they should on average exhibit lower dissimilarity scores within this framework. For each species, dissimilarity score distributions were visualized using the R/tidyr package (Wickham 2014). In the case of H. illucens, the full distance matrix was displayed as an annotated clustermap in Seaborn (Waskom 2021). The heatmap shows pairwise distances, with both rows and columns grouped by Ward's hierarchical clustering (Ward 1963) and is displayed with a dendrogram. Additional genomic annotations, such as scaffold location, presence or absence of introns, and the number of BIR domains encoded, were incorporated as categorical tag columns.

To situate duplication events within a phylogenetic framework, we used Orthofinder version 3.1.0 (Emms and Kelly 2019) to identify putative orthologs among the above duplicates. When orthologous genes were detected in more than one species, we inferred that the duplication event that produced them occurred prior to the divergence of those species. For each species, the longest peptide isoform of each gene in the complete proteome was provided as input, and OrthoFinder was run with default parameters. OrthoFinder assigned 351,075 genes (91.5% of all input genes) to 26,354 orthogroups. Among these, 3,210 orthogroups contained representatives from every species, and 310 of these were composed exclusively of single-copy genes. OrthoFinder results were then exploited in two complementary ways: (i) to identify IAPs from different species that cluster within the same orthogroups, thereby inferring shared evolutionary origin; and (ii) to evaluate the genomic context of each IAP by assessing whether their flanking genes are assigned to the same orthogroups across species, which provides additional support for conserved synteny and shared ancestry.

IAP domain architecture prediction

Unraveling the evolutionary history of metazoan IAPs implies looking at the diversity of domain architectures that characterize these proteins, in order to determine how and when they arose. Therefore, for each predicted IAP, the presence of putative functional domains was predicted by InterProScan (Paysan-Lafosse et al. 2023), as was the presence of signatures suggesting membership of known protein families or superfamilies (Data S15). As InterProScan uses predictive models that may be biased toward certain taxa, particularly mammals, the presence of the most commonly found domains (ie BIR, RING, CARD, UBA, UBC, and NACHT) was verified by HMM search with in-house HMM profiles. These profiles were generated using the HMMbuild tool of the HMMer package on the domain sequences found in the 2,843 IAPs (Fig. S20), then used to perform new searches against those IAPs, and were updated iteratively using the newly identified domains until no new domain was identified (n = 5 iterations). A P-value threshold of 1 × 10⁻⁵ was chosen to define the presence or absence of a domain. The results obtained for the BIR, CARD, UBC, and NACHT domains were identical to those obtained with InterProScan. For the RING and UBA domains, this new analysis enabled us to identify new domains, which were not identified by InterProScan, and eliminate others, which were identified with InterProScan but not with the HMM search. The new in-house HMM profile for the BIR domain was also used to research metazoan proteomes, confirming that no IAPs had been missed in the first analysis.

A comprehensive description of all putative IAPs identified in this work (ie accession numbers, protein length, protein domain composition, and exon numbers) is provided in Data S3 and S4, for the “META” and “PROTO” datasets, respectively.

BIR domain phylogeny

The high modularity of IAPs, in terms of domain composition, precludes the possibility of inferring the phylogeny of these proteins from their complete sequence. This is because events such as partial gene duplications and internal duplications or losses of individual domains may have resulted in homologous domains appearing at different positions within full-length IAPs. Consequently, aligning entire protein sequences would inevitably force non-homologous regions into alignment, compromising phylogenetic accuracy. However, this difficulty can be overcome by analyzing the BIR domains, which are present in all IAP sequences (see Introduction section). We found 880 and 4,464 BIR domains in the 509 and 2,660 putative IAPs identified in the “META” and “PROTO” datasets, respectively. The boundaries of each individual domain were determined using the signature SM00238 (bir_2) given by InterProScan. When multiple BIR domains were present within an IAP, they were numbered relative to their respective positions in the protein sequence (eg BIR_Y/X, with Y the relative position of the domain from the N-terminus to the C-terminus and X the total number of BIR domain). Each individual BIR sequence therefore had a unique identifier, composed of the protein ID (eg XP_033739516.1) followed by an indication of the domain position within the IAP (eg BIR_1/2 and BIR_2/2 when two BIR domains are present in the protein). BIR sequences were then extracted and, from this point onward, alignments and phylogenetic inferences were conducted on the individual domains.

To limit sequence redundancy, BIR domains with very similar sequences in each dataset were clustered using the mmseq2 software (V 13.45111) (Steinegger and Soding 2018), with the following parameters: 80% identity and 70% coverage, coverage mode 1 (alignment covers at least 70% of target, recommended when clustering protein fragments) and cluster mode 2 (takes the longest sequence as representative of the cluster, recommended to suppress partial sequences from becoming representative sequences). When a cluster contained sequences from species of several orders, one representative sequence per order was selected. This allowed us to retain 532 (out of 880) and 1,801 (out of 4,464) BIR sequences in the “META” and “PROTO” datasets, respectively (Data S10 and S16).

To further ensure the quality of the alignments, outlier sequences were omitted from the phylogenetic analysis when (i) their size was significantly different from the average size of BIR domains (<60 or >80 amino acids), (ii) they presented substitutions at amino acids that are key to BIR domain function (ie cysteines and histidine involved in the formation of the BIR fold), or (iii) they were identified by only one of the several BIR domain signatures used by InterProScan (suggesting a strong sequence divergence from the domains used to generate the predictive models). When one of the outlier sequences had been selected as representative of a cluster by the mmseq2 software, a new representative sequence was chosen wherever possible. In the end, 479 and 1,642 BIR sequences were retained in the “META” and “PROTO” datasets, respectively (Data S10 and S16).

These BIR sequences were aligned using the MAFFT multiple alignment program V 7.505 (Katoh and Standley 2013) with the accurate L-INSI option. Alignments were visually inspected using the alignment editor Seaview V 5.0.5 (Gouy et al. 2010), and graphical representations of the alignments were made with ESPript V 3.0 (Robert and Gouet 2014). Phylogenetic inferences were then conducted under maximum likelihood optimization using (i) FastTree V 2.1.11 (Price et al. 2010) with the LG + gamma 20 substitution model and (ii) IQ-TREE 2V 2.3.4 (Minh et al. 2020) with the LG model as selected by ModelFinder (Kalyaanamoorthy et al. 2017) (as implemented in IQ-TREE 2 V2.3.4). Node robustness was measured with SH-like local support for trees inferred with FastTree and aLRT and 1,000 UltraFast Bootstrap for trees inferred with IQ-TREE. Phylogenetic trees were visualized and annotated using Itol V 6 (https://itol.embl.de/itol.cgi) (Letunic and Bork 2019).

Species tree of metazoans and protostomes

Understanding the evolutionary history of a gene family means reconciling the information obtained at the gene level with the species tree in order to formulate hypotheses about gene duplications and losses, and about the shuffling of functional domains at the origin of the huge diversity of present-day IAPs. We therefore set out to reconstruct species phylogenies for both the “META” (Fig. 5) and “PROTO” (Fig. 8) datasets. This was necessary because, although phylogenies exist for metazoans and protostomes, no tree including our specific selection of species was available in the literature. We used a dedicated Python pipeline (McGowan et al. 2023) to assemble sets of near-universal single-copy protein families corresponding to the BUSCO “metazoa_odb10” lineage-specific dataset, since all species from our two datasets are metazoans. The protein families that are single copy in at least 70% of the input samples were retained and aligned with MUSCLE V 5.1 (Edgar 2004), trimmed with trimAI V 1.4 (Capella-Gutierrez et al. 2009), and used to build a supermatrix. By applying this pipeline, two supermatrices were assembled: SMmetazoa (341 protein families, 103,197 amino acid positions) and SMprotostomia (680 protein families, 215,568 amino acid positions). Maximum likelihood phylogenies were inferred using IQ-TREE V 2.2.2.5 (Minh et al. 2020), with 1,000 bootstrap replicates and the best-fit model as determined by the ModelFinder feature (Kalyaanamoorthy et al. 2017). The resulting trees were overall consistent with known phylogenies (Misof et al. 2014; Telford et al. 2015), with species clustering together according to their taxonomic affiliation.

Processing of publicly available expression data

Raw sequencing reads from 16 distinct BioProjects (Data S6) comprising 208 RNA-seq libraries from ten organisms (P. vulgata, B. glabrata, P. clarkii, P. chinensis, T. palmi, A. pisum, A. gossypii, M. persicae, C. felis, and I. scapularis) were retrieved and processed using TrimGalore v0.6.10 (https://github.com/FelixKrueger/TrimGalore) with default settings to remove low-quality bases and adapters. The filtered read pairs were then aligned to the corresponding species’ genomes using HISAT2 v2.2.1 (Kim et al. 2019), and gene-level counts were generated with HTSeq-count v1.99.2 in union mode (Anders et al. 2015). Differential expression analysis was performed using DESeq2 v1.42.0 (Love et al. 2014). Genes were considered differentially expressed if they had a false discovery rate adjusted P-value (P-adj) below 0.05. For further analysis, only genes with a log2 fold change (Log2FC) greater than 1 (upregulated) or less than −1 (downregulated) were retained. For visual representation of IAP expression in different conditions, we applied a Z-score normalization to each gene expression vector. This method is a well-recognized normalization technique that allows a robust and direct comparison of the expression of genes across different experiments (Cheadle et al. 2003). This normalization aims at rescaling the expression levels to reduce the gene-specific baseline expression scale differences and rather focus on condition-dependent variations. In practice, Z-score normalizes each gene vector by subtracting its mean expression and by dividing each element from the vector by its standard deviation. Hence Z-score expresses gene expression levels in terms of standard deviation units.

Statistical analysis

The R software V 4.0.5 (R Core Team 2022) was used to assess the correlation between the number of IAPs and the number of protein-coding genes in metazoans. As the count distributions are not normal, a logarithmic transformation was applied before calculating the non-parametric Spearman correlation coefficient. P-values < 0.05 were considered significant.

Author contributions

M.R.L and F.C. secured funding. M.R.L., N.P., C.B.-A., and F.C. designed this study. M.R.L., N.P., S.P., F.R., P.B.P., E.J., H.C., P.C., C.B-A., and F.C. performed data analyses and interpreted the results. M.R.L., N.P., S.P., and C.B-A. prepared the figures and tables. M.R.L., C.B.-A., and F.C. wrote the manuscript with input from all authors. All authors contributed to the critical reading of the manuscript and accepted its final version.

Supplementary material

Supplementary material is available at Molecular Biology and Evolution online.

Funding

This work was supported by INRAE (Institut National de Recherche pour l’Agriculture, l’Alimentation et l’Environnement), INSA Lyon (Institut National des Sciences Appliquées de Lyon) and the INSA BQR grant CarBaDyn.

Data availability

All data generated or analyzed during this study are included in this published article (and its supplementary information files) or available from the corresponding author on reasonable request.