Molecular Drivers of Mutualistic Association Between Anemone and Anemonefish

Abstract

The anemone–anemonefish mutualism is one of the most iconic in the marine environment. While the evolution of this mutualistic relationship has contributed to the ecological success of both partners, the underlying molecular processes that establish and maintain it remain poorly understood, particularly how anemonefish tolerate anemone venom. Here, we characterize the transcriptional dynamics in both the anemonefish Amphiprion clarkii and its host anemone Entacmaea quadricolor 48 h after association, providing a rare insight into the coordinated molecular processes in both partners that underlie symbiosis establishment. Upon acclimation with an anemone, anemonefish showed differential regulation of sensory perception and memory genes in key brain regions, indicating activation of neural pathways that may facilitate host recognition and mutualism establishment. In the fish's skin, altered expression of genes involved in neurotransmitter release, cytoskeleton organization, and venom receptor proteins points to mechanisms of resistance to anemone venom. This resistance is particularly remarkable since anemone hosting fish exhibited increased expression of genes encoding mechanoreceptors, putative venom-associated proteins, and ion channels involved in nematocyst discharge, indicating the anemone does indeed mount an active response to their mutualistic partner. By simultaneously capturing the molecular responses of both symbiotic partners, our results reveal the complex, coordinated interplay of molecular events in both species that play a pivotal role in establishing this mutualistic relationship.

Significance.

Mutualistic partnerships between species are essential for ecosystem functioning, yet we know little about the molecular mechanisms enabling animals to recognize partners and establish these relationships. The anemonefish–anemone symbiosis represents one of nature's most remarkable mutualisms: anemonefish live unharmed among the venomous tentacles of sea anemones, but the molecular processes underlying fish venom resistance and mutualism establishment remain poorly understood. By examining gene expression changes in both partners during their initial association, we discovered that anemonefish activate specific memory and sensory pathways in their brain and deploy multiple complementary defense mechanisms against anemone venom, facilitating their mutualistic relationship, as anemones continue producing venom despite hosting their partner. Remarkably, while anemones benefit from the partnership, they simultaneously experience physiological stress, revealing a complex cost-benefit dynamic. These findings demonstrate that successful mutualisms require coordinated molecular adjustments in both partners, fundamentally advancing our understanding of how animals establish and maintain such relationships.

Introduction

Mutualistic interactions are major drivers of biodiversity and ecosystem function, influencing evolutionary trajectories across diverse taxa (Hale and Valdovinos 2021). These complex associations evolve through various processes (Weiblen and Treiber 2015; Chomicki et al. 2020) and, once established, can drive evolutionary transitions and species diversification (West et al. 2015; Chomicki et al. 2020). For instance, mutualisms between plants and mycorrhizal fungi, plants and pollinators, corals and dinoflagellates, sea anemones and fish, and primates and plants have all contributed to evolutionary diversifications and increased species richness (Gómez and Verdú 2012; Weiblen and Treiber 2015; Campoy et al. 2020).

Among marine mutualisms, the association between anemonefishes (Pomacentridae) and giant sea anemones is especially intriguing, with all anemonefish species associating with only ten host anemone species spanning three families (Actiniidae, Stichodactylidae, and Thalassianthidae; Kashimoto et al. 2024). Sea anemones produce venom both in their mucous secretions and through nematocysts discharged upon mechanical or chemical stimulation as a defense mechanism (Anderson and Bouchard 2009). While the venom produced by sea anemones is lethal to most marine organisms, anemonefishes exhibit a unique resistance enabling them to form mutualistic relationships with one or more of ten sea anemone species (Fautin and Allen 1993; Litsios et al. 2012). This mutualistic association has enhanced the ecological success of both partners (Holbrook and Schmitt 2005). Anemonefish gain protection from predation and a secure breeding site (Holbrook and Schmitt 2005), while the host anemones benefit from defense against predators, nutrient input, and increased oxygenation that promote growth and reproduction (Szczebak et al. 2013; Hoepner et al. 2022). Notably, while anemonefishes are obligate symbionts and never found solitary in nature, anemones can survive independently (Fautin and Allen 1993). This obligate association was contingent on anemonefishes developing distinct molecular and physiological adaptations to live unharmed among the anemone's stinging tentacles (Marcionetti et al. 2019). Several hypotheses have been proposed to explain this resistance, with the anemonefish's mucous coating widely considered to play a key role in protecting them from anemone toxins (Lubbock 1980; Balamurugan et al. 2014; Abdullah and Saad 2015). The mucus layer of anemonefishes is proposed to be thicker, chemically distinct from other coral reef fishes, and lacking compounds that trigger nematocyst discharge (Lubbock 1980; Hoepner et al. 2022). In particular, anemonefish mucus contains lower levels of sialic acids, a compound that triggers nematocyst firing (Abdullah and Saad 2015; Heim et al. 2025; Roux et al. 2025), and may also be molecularly similar to anemone mucus, effectively acting as a “chemical camouflage” to inhibit nematocyst discharge (Schlichter 1976; Elliott et al. 1994).

A key aspect of this resistance also involves a behavioral process called “acclimation” during initial contact with a host anemone. During acclimation, fish perform stereotyped behaviors in which they repeatedly touch the anemone tentacles first with only their fins and then gradually increase body contact with the tentacles until they are no longer stung (Elliott et al. 1994). During this acclimation process, molecular changes likely occur within the fish that confer venom resistance (Hoepner et al. 2022). Recent genomics studies identified genes that underwent positive selection during anemonefish diversification, such as Versican core protein and Protein O-GlcNAse, which alters the structure of N-acetylated sugars in fish mucus glycoproteins thereby preventing nematocyst discharge and contributing to anemonefish venom resistance (Marcionetti et al. 2019). However, the dynamic molecular changes occurring in the anemonefish during the critical acclimation period remain uncharacterized, limiting our understanding of the mechanisms of venom resistance.

While most research has focused on the fish partner, the host anemone may also undergo physiological and molecular changes during acclimation. Recent studies have shown that the microbiomes of both the host anemone and anemonefish are altered after association (Pratte et al. 2018; Roux et al. 2019; Titus et al. 2020; Émie et al. 2021). Host venom production may similarly change upon mutualism establishment, although the direction of this change is unclear. On one hand, if acclimated anemonefish no longer triggers nematocyst discharge, this could be reflected in the host anemone through reduced venom production upon mutualistic association. On the other hand, because anemone venom serves as a general defense against diverse predators, venom production could instead increase upon symbiotic establishment to better protect both the host and its partner. Therefore, investigating how venom production changes upon mutualism establishment can provide crucial insights into the molecular mechanisms underlying the formation and maintenance of this complex mutualistic relationship.

The present study investigates the molecular underpinnings of mutualism between the sea anemone Entacmaea quadricolor and the anemonefish Amphiprion clarkii, both of which are generalist species (Ollerton et al. 2007; Litsios et al. 2012). E. quadricolor is a natural host of A. clarkii (Elliott and Mariscal 1997; Hoepner et al. 2022), and A. clarkii requires an acclimation period in order to fully associate with E. quadricolor (Elliott and Mariscal 1997; Mebs 2009; Roux et al. 2019). Additionally, when separated from its host anemone for prolonged durations, A. clarkii, similar to other anemonefish, needs to reacclimate (Mariscal 1970; Schlichter 1976). This acclimation period plays a crucial role in establishing the symbiotic relationship, and we hypothesize that during acclimation gene expression patterns are altered in both partners. By investigating the transcriptional landscape of both species upon acclimation, we provide a holistic view of the bidirectional molecular mechanisms underlying this symbiosis. Specifically, we aim to identify the molecular changes that facilitate mutualistic association in both partners and elucidate the molecular processes conferring anemonefish resistance to anemone venom.

Results

To investigate the molecular changes underlying mutualistic association between A. clarkii and E. quadricolor, we employed a comparative experimental design where fish and anemones were assigned to one of two treatments: a control treatment, in which fish (n = 4) were maintained with artificial anemones, and live anemones (n = 5) were kept in isolation, and an interaction treatment, in which fish (n = 5) and live anemones (n = 5) were housed in the same tank allowing them to interact. After 48 h, a duration sufficient for full acclimation (Mariscal 1970; Hoepner et al. 2022), the transcriptomic profiles from the skin and five brain regions of A. clarkii and the tentacles of E. quadricolor were examined. Of the nine total fish sampled, four were females, and five were males (Table S1). An overview of the experimental design and sampled tissues is provided in Fig. 1.

Fig. 1.

Schematic representation of the experimental design showing from left to right: control fish (with fake anemone), control anemone and physical interaction between the fish and anemone, and the sampled tissues for RNA sequencing. BS, brain stem; CB, cerebellum; DI, diencephalon; OT, optic tectum; TE, telencephalon.

A. clarkii Acclimation Behavior

Visual examination of recorded videos of fish housed with live sea anemones (n = 5) was used to confirm that full acclimation had occurred prior to tissue sampling. The fish were observed to show characteristic acclimation behaviors upon introduction of the anemone into tanks, with the fish initially maintaining distance before gradually approaching the anemone, sometimes hovering near the anemone for several minutes before making first contact. Initial contact involved fish touching anemone tentacles with their mouth or fins, often swimming away after being stung (identified by a visible jolting motion away from the anemone) but then returning to continue acclimation. Fish engaged in more frequent and closer contact with the anemone over time, transitioning from tentatively touching the anemone with just their fins and mouth to rubbing their entire body against the tentacles. Additionally, fish were observed nibbling the tentacles and fanning the anemone. Within 5 h of introducing the anemone into the tanks, four out of five fish were nearly fully acclimated; however, one individual remained unacclimated, only touching the anemone with its caudal fin and showing avoidance. Within 24 h of introducing the anemone to the tanks, all fish achieved full acclimation, swimming freely among the tentacles without showing any visible signs of avoidance or distress upon contact.

Transcriptomic Responses in A. clarkii Skin Upon Acclimation

Differential expression analysis identified significant changes in A. clarkii skin gene expression upon acclimation to anemones. Samples clustered primarily by condition rather than sex (Fig. S1(a)), and a likelihood ratio test (LRT) identified a significant effect of sex on only two genes (0.01% of total expressed genes in skin; adjusted P-value [Padj] < 0.05; Table S2). Comparing control fish to those acclimated with anemones using the Wald test in DESeq2 identified a total of 38 significantly differentially expressed (DE) genes (false discovery rate [FDR] adjusted P-value < 0.05, absolute log2 fold change >0.3, and baseMean >10), with the majority (36 genes) being downregulated in fish acclimated with anemones (Table S3). Two of the 38 DE genes were also significant in the LRT for sex; however, because sex was included as a covariate in the model, the estimated condition effects were independent of sex. The 38 DE genes were associated with diverse functional categories such as cytoskeleton, vesicle transport, metabolism, and cellular stress (Fig. 2). Notably, a mannose receptor (MRC1) that binds C-type lectins and RalBP1-interacting protein 2 (REPS2), which is involved in receptor endocytosis, were both downregulated in fish acclimated with anemones (Fig. 2). Functional enrichment analysis identified two significantly enriched gene ontology (GO) terms (sarcomere and myofibril assembly) among the 38 DE genes (Table S4).

Fig. 2.

Functional categories associated with the DE genes in A. clarkii skin. Each point represents one gene, and the color represents the FDR corrected P-value (Padj) reported by DESeq2.

Neurotranscriptomic Responses in A. clarkii Brain Regions Upon Acclimation

The five brain regions (diencephalon, telencephalon, optic tectum, cerebellum, and brain stem) displayed distinct gene expression patterns upon acclimation and mutualism establishment (Fig. 3(a)). The samples clustered primarily by brain regions rather than by sex (Fig. S1(b)), with an LRT in DESeq2 identifying a significant effect of sex on 365 genes (1.5% of total expressed genes in brain; Padj < 0.05; Table S2). Given the strong effect of brain region on gene expression, differential expression analysis between control and interaction treatments was carried out separately for each of the five brain regions using the Wald test in DESeq2. The magnitude of transcriptional response varied considerably among regions, with the brain stem showing the largest transcriptional response (21 DE genes), followed by telencephalon (14 DE genes), cerebellum (9 DE genes), diencephalon (3 DE genes), and optic tectum (2 DE genes; Table S3). Notably, there was minimal overlap in DE genes across the five brain regions (Fig. 3(b)), with only one gene (PRUNE2) commonly DE among brain stem, telencephalon, and optic tectum, and one other DE gene (unannotated) shared between brain stem, cerebellum, and telencephalon. Of the 45 unique DE genes identified across the five regions, only six were also affected by sex (LRT, Padj < 0.05); but since sex was included as a covariate in the model, the condition effects for these genes were estimated independent of sex. No GO terms were significantly enriched among DE genes in each individual brain region.

Fig. 3.

a) Overall pattern of gene expression across the five brain regions of A. clarkii in control conditions and associated with anemones. The principal component analysis with 95% confidence ellipses was done using the regularized log-transformed (rlog) counts of all expressed genes. b) Number of DE genes between A. clarkii in control and associated with anemones for each of the five brain regions.

The forebrain regions (telencephalon and diencephalon) showed evidence of reduced cellular stress, indicated by downregulation of CYP1A, CRHBP, and NRDE2 in the telencephalon, and ERCC4 in the diencephalon of anemone-acclimated fish compared to controls. Additionally, PMEL, which is involved in melanin biosynthesis and pigmentation, was upregulated in the diencephalon, and CFAP74, which is involved in olfactory signaling, was upregulated in the telencephalon of fish upon acclimation. In contrast, the hindbrain subregions (cerebellum and brain stem) exhibited differential expression of genes primarily associated with somatosensory functions, sensory memory, immune response, and neurogenesis (Fig. 4; Table S3). This regional specificity suggests that distinct neural functions are activated in each of the brain subregions in fish upon acclimation.

Fig. 4.

Expression changes (log2 fold change) of DE genes between A. clarkii in control and acclimated with anemones across the five brain regions. BS, brain stem; CB, cerebellum; DI, diencephalon; OT, optic tectum; TE, telencephalon.

E. quadricolor Transcriptome Profile Upon Mutualism Establishment

A de novo transcriptome assembly was constructed for E. quadricolor because no reference genome is currently available for this species, with assembled contigs clustered into genes to enable gene-level expression analysis (see Materials and Methods). E. quadricolor displayed a substantially more extensive transcriptional response than A. clarkii upon mutualism establishment, with 965 DE genes identified (578 upregulated, 387 downregulated; Table S5). However, functional annotation was available for only 321 upregulated and 209 downregulated DE genes based on the UniRef90 database (https://www.uniprot.org/uniref?query=%28identity%3A0.9%29), limiting functional enrichment analysis to these annotated genes.

There were no significantly enriched GO terms among the downregulated DE genes while a total of 145 GO terms were significantly enriched among the upregulated DE genes (Table S6). These could be grouped into 11 broader functional categories which included signaling, G-protein coupled receptor (GPCR), putative venom-associated proteins, immune response, TGF-beta family, stress response, gene expression regulation, ion channels, molecular trafficking, cell cycle, and structural genes (Fig. 5; Table S7).

Fig. 5.

Functional categories associated with the DE genes in E. quadricolor tentacles. Each point represents one gene, and the color represents the FDR corrected P-value (Padj) reported by DESeq2.

Discussion

The initial acclimation period is pivotal for establishing the mutualistic relationship between anemones and anemonefishes, during which the fish acquires protection from the host's stinging tentacles through coordinated physiological and molecular changes in both partners (Elliott and Mariscal 1997; Balamurugan et al. 2014). Here, we investigated the transcriptional responses of both A. clarkii and E. quadricolor during mutualism establishment to identify the molecular processes underlying this symbiosis. In our study, all fish displayed typical acclimation behavior upon introduction to an anemone and were fully acclimated within 24 h. However, the timescales at which behavioral and molecular changes occur can differ, with environmentally induced shifts in gene expression often taking hours to days to establish following a stimulus (Maruska and Fernald 2010; Rittschof and Hughes 2018). We therefore sampled tissues from both species at 48 h following introduction, to capture a consolidated rather than a transient transcriptional state. The resulting changes in the transcriptional landscape observed in both partners highlight key molecular processes underlying mutualism establishment.

In anemonefish, acclimation was accompanied by modulation of sensory pathways in the brain. Differential expression of genes involved in sensory perception, signaling, and sensory memory across the telencephalon, cerebellum, and brain stem suggests that modulation of sensory processes plays a central role during acclimation. This molecular signature aligns with the known reliance of anemonefishes on olfactory cues to recognize and locate their host anemones (Elliott et al. 1995; Arvedlund and Nielsen 1996; Arvedlund et al. 1999). Supporting this, we observed upregulation of CFAP74 in the telencephalon, a cilia- and flagella-associated protein that supports olfactory function (Jenkins et al. 2009; Dong et al. 2017). Furthermore, the downregulation of XIRP1, an inhibitor of actin depolymerization, in the cerebellum may enhance associative olfactory memory in anemonefish via actin remodeling, a mechanism also observed in honeybees (Ganeshina et al. 2012; Reiprich et al. 2021). Together, these patterns suggest that olfactory learning and memory, perhaps similar to processes occurring during embryonic imprinting (Arvedlund et al. 1999), may be reactivated in adults during reacclimation to an anemone. The detection of these olfactory cues could possibly contribute to the behavioral responses necessary for acclimation in the fish. In fact, aromatase (CYP19A1A), which has been implicated in sensory processing and sensory memory within the social behavior network, was upregulated in the brain stem, where it potentially facilitates translation of sensory cues into behavioral responses (Garcia-Segura 2008; Diotel et al. 2010; Spool et al. 2022). Overall, these coordinated changes in sensory genes and signaling pathways are consistent with a role for olfactory processing in guiding acclimation behavior; however, as this study relies solely on transcriptomic data, additional work incorporating behavioral and physiological data is needed to confirm that fish are actively integrating olfactory cues to facilitate mutualism establishment. Furthermore, as the fish used in our study were previously associated with anemones, prior to a seven-month separation, we propose that the molecular signatures associated with sensory pathways may partly reflect reactivation of an olfactory memory established during their first encounter with an anemone, consistent with behavioral evidence that prior experience with a host anemone facilitates faster reacclimation (Balamurugan et al. 2014). Modulation of sensory genes may therefore differ between naive fish encountering a host anemone for the first time and reacclimating fish such as those used here, and further studies are needed to test this. This sensory-driven partner identification hypothesis parallels findings in other mutualisms, such as visual recognition of cleaner fish by clients through distinctive color patterns (Caves 2021) and olfactory recognition of mutualistic partners in cardinalfish-bacteria (Gould et al. 2015) and bee-plant (Leonhardt et al. 2014) associations, underscoring that sensory communication represents a fundamental mechanism for establishing mutualistic partnerships across diverse taxa.

The most intriguing aspect of the anemone–anemonefish mutualism is the resistance of the anemonefish to anemone venom. Here, we found differential expression of various cytoskeletal genes in the anemonefish skin upon acclimation, which could potentially contribute to venom resistance by inducing structural modifications in the epidermal plasma membrane receptors potentially targeted by venom proteins, thereby possibly reducing their binding affinity (Hoepner et al. 2022). This mechanism of target alteration providing venom tolerance is also documented in other species (Holding et al. 2016; Hoepner et al. 2022). Cytoskeleton remodeling has also been implicated in the defense response of plants against pathogens (Wang et al. 2022). However, since this study focuses on transcriptomic data, further studies are needed to confirm that these expression changes translate to functional changes in receptor structure which provides venom resistance. Interestingly though, cytoskeleton genes were identified to be under positive selection at the origin of anemonefish radiation, potentially contributing to the evolution of this iconic mutualistic relationship and the subsequent adaptive radiation of anemonefishes (Marcionetti et al. 2019), and hence they could potentially be important in venom tolerance of anemonefishes. Furthermore, other genes with annotated functions potentially related to venom resistance were DE in the skin of acclimated anemonefish. These include downregulation of a mannose receptor (MRC1), which binds C-type lectins, a common component of anemone venom (Geijtenbeek and Gringhuis 2009; Delgado et al. 2022), and an endocytosis-related gene (REPS2), whose expression is correlated with receptor internalization (Oosterhoff et al. 2005). A similar mechanism has been observed in opossums, where alterations in C-type lectin protein binding sites potentially provide protection from venomous snake predators (McCabe and Mackessy 2017), though direct functional evidence in anemonefish is lacking. Additionally, downregulation of synaptotagmin and taxilin, both involved in neurotransmitter release (Petrenko et al. 1991; Shoji-Kasai et al. 1994; Nogami et al. 2003), may provide tolerance to neurotoxins and other pore forming toxins found in anemone venom, possibly by preventing uncontrolled neurotransmitter release (Rivera-de-Torre et al. 2019; Delgado et al. 2022). Together with previous findings that anemonefish actively reduces sialic acid levels in their mucus as a protective mechanism against nematocyst discharge (Roux et al. 2025), our results highlight several candidate genes and pathways that may contribute to venom tolerance and, by extension, mutualism establishment in anemonefish, although functional validation is needed to confirm this.

Venom resistance of anemonefish is particularly crucial, as changes in the transcriptional landscape of host anemones upon acclimation suggest that they possibly mount an active response to the presence of the fish. The upregulation of genes associated with mechanosensory functions and structural components of mechanoreceptors in anemones hosting A. clarkii suggests possible modulation of mechanosensory pathways in response to fish movements. Anemones are known to rely on both chemical and mechanical sensory systems to detect movements around them and discharge nematocysts in response (Watson and Hessinger 1989), and the length and density of hair bundle mechanoreceptors in anemone tentacles increase in response to physical stimulation (Campbell et al. 2018). Furthermore, upregulation of genes associated with GPCR and hedgehog signaling, two signaling pathways known to be associated with sensory cilia in mechanoreceptors (Sigg et al. 2017), further supports the activation of mechanosensory systems; however, direct measurement of mechanoreceptor activity in the anemones is needed to confirm this. This mechanosensory activation is potentially followed by nematocyst discharge as we observed increased expression of genes encoding calcium and potassium channels, ion channels involved in nematocyst exocytosis (Lubbock et al. 1981; Gitter et al. 1994; Sutton et al. 2004; Babonis et al. 2023), and putative venom-associated proteins in anemones hosting fish. Additional physiological or histological validation is however needed to confirm that the anemones are indeed actively discharging nematocysts. While previous studies have proposed that anemonefish mucus molecularly mimics that of the anemone (Schlichter 1976; Elliott et al. 1994; Hoepner et al. 2022) and might lack certain components that triggers nematocyst firing (Abdullah and Saad 2015; Hoepner et al. 2022), our findings suggest the alternative possibility that fish movements could continue to stimulate cnidocyte mechanoreceptors in anemones even after acclimation, to which the fish responds through its multifaceted venom resistance mechanisms.

Mutualistic association induced comprehensive physiological adjustments in both species. In anemonefish, acclimation may trigger pigmentation changes, evidenced by the upregulation of PMEL in the diencephalon, a melanocyte protein involved in melanosome morphogenesis and pigmentation. While we did not directly measure phenotypic color changes, changes in melanism are known to occur in some anemonefishes upon association with certain anemone species (Fautin and Allen 1993), and while their adaptive value remains unclear, it may be linked to host toxicity tolerance (Merilaita and Kelley 2018). Alternatively, given that melanomacrophages play a role in fish immunity (Steinel and Bolnick 2017; Bjørgen and Koppang 2024), upregulation of PMEL may reflect an immune response to anemone venom exposure rather than, or in addition to, pigmentation changes. Supporting this, complement component 7B (C7B), also involved in immune response (Shen et al. 2012), was upregulated in the brain stem. However, broader immune gene upregulation was otherwise absent in A. clarkii at 48 h, suggesting that either the immune response was limited or had not yet fully mobilized at this sampling point. We also detected molecular signatures of stress alleviation in the fish following successful acclimation indicated by the upregulation of CRHBP, an inhibitor of the stress hormone corticotropin-releasing hormone, and downregulation of various other cellular stress response genes in the telencephalon, diencephalon, and skin. This pattern likely reflects a shift from a state of increased predation vulnerability when fish were separated from their host anemones, similar to conditions experienced when host anemones undergo bleaching (Beldade et al. 2017), to a more secure state once the mutualism was reestablished. Furthermore, shifts in metabolic gene expression in the skin of acclimated fish suggest possible metabolic changes resulting from nutrient exchange between partners (Cleveland et al. 2011; Alan Verde et al. 2015). In anemones, association with fish was accompanied by upregulation of immune response and stress response genes, revealing potential physiological costs for the anemone. The increased expression of immune genes may aid tolerance to microbiome shifts that occur upon mutualism establishment (Émie et al. 2021), while elevated expression of stress genes, including antioxidant genes such as thioredoxin reductase, glutathione S-transferase, paraoxonase, and selenoprotein P, suggests that the anemone experiences oxidative stress upon mutualism establishment. We note that this interpretation is based on gene expression alone, and direct biochemical measurement of reactive oxygen species or antioxidant enzyme activity would be needed to confirm oxidative stress. Like corals, anemones harbor symbiotic dinoflagellates whose photosynthetic activity imposes a baseline level of oxidative stress on the host despite providing energetic benefits (Wooldridge 2010). The additional physical disturbance imposed by the fish acclimation behaviors, including nibbling of the tentacles, likely elevates this oxidative stress burden further, although this remains to be tested directly. Therefore, while anemones receive clear benefits from the fish, our results suggest that they may simultaneously experience physiological costs. Anemones thus may navigate a complex balance between physiological costs and mutualistic benefits. Intriguingly, the shift in anemone microbiome composition upon mutualistic establishment could serve as a compensatory mechanism, mitigating oxidative stress through microbial processing of free radical metabolites (Titus et al. 2020). Collectively, these findings highlight that the establishment and maintenance of this mutualistic relationship likely involve coordinated physiological adjustments in both partners, underscoring the complexity of this mutualistic partnership.

Conclusion

Our findings reveal that the establishment of the iconic mutualism between sea anemones and anemonefish involves coordinated molecular changes in both partners during the critical acclimation period. In A. clarkii, acclimation induced transcriptional shifts in genes involved in sensory perception and memory across brain regions, suggesting that host recognition and acclimation may be facilitated by olfactory-mediated processes, although complementary physiological evidence is needed to confirm this. While it has been hypothesized that anemonefish do not trigger nematocyst discharge (Hoepner et al. 2022), our findings raise the alternative possibility that anemones hosting fish may continue to discharge nematocysts even after the fish has fully acclimated. Anemones hosting fish upregulated genes encoding putative venom-associated proteins and ion channels associated with nematocyst discharge, possibly due to the activation of mechanoreceptors, though further research is needed to confirm the functional significance of these transcriptional changes. The fish in turn modulated the expression of cytoskeleton genes and putative venom-associated protein receptors in the skin which may confer tolerance to the anemone venom, although this remains to be functionally validated. Both partners also exhibited molecular signatures of broader physiological changes which could be key to driving the mutualistic association. The potential costs to anemones, suggested by stress and immune gene upregulation, alongside the clear benefits anemonefish provide, suggest a dynamic cost-benefit balance that has likely shaped the coevolutionary dynamics of this partnership. Overall, our results show that mutualism establishment is underpinned by dynamic transcriptomic changes in both the anemonefish and its host anemone, underscoring the complex interplay of molecular events that underlie their mutualistic association.

Materials and Methods

Sample Collection and Experimental Design

Amphiprion clarkii and Entacmaea quadricolor were chosen as model species for our study as both are widely distributed, and E. quadricolor is the most common host anemone for A. clarkii in Hong Kong (See 2017). Adult A. clarkii individuals were collected by SCUBA diving from a depth of 5 to 7 m using barrier nets and hand nets in September 2021 from two locations (22°15.525′N 114°20.981′E and 22°15.545′N 114°20.930′E) in South Ninepin Island, Hong Kong. Fish were transported to the aquarium facility at The University of Hong Kong and maintained in pairs for a period of 7 months (until April 2022) in complete isolation from anemones, which is sufficient time for anemonefish to lose their protection from anemone venom (Mariscal 1970; Hoepner et al. 2022). One week before the start of the experiment, fish were separated and housed in individual tanks.

Ten individuals of E. quadricolor, originally collected in Hong Kong waters, were purchased from a local aquarium store and maintained in isolation from anemonefish in separate tanks for a period of 5 days prior to the start of the experiment. A shorter isolation time was chosen for anemones as it has been previously established that reacclimation of fish with anemones is not influenced by the length of time anemones have been in isolation (Mariscal 1970). Both fish and anemones were maintained in 30 L tanks with a photoperiod of 12/12 h (light/dark cycle). Fish and anemones were fed ad libitum with commercial feed once daily, with fish diet supplemented with brine shrimp once per week. Temperature, pH, dissolved oxygen, and salinity were kept constant throughout the experiment while mimicking the natural parameters of the collection site (Table S8).

At the start of the experiment, five anemones were transferred to five different tanks, each containing one fish. For the control treatment, four artificial anemones (made from rubber bands) were transferred to four other tanks, each containing one fish. Five control anemones (kept in isolation throughout the experiment) were also transferred between tanks to account for movement effects. The interaction between fish and anemones was recorded for 5 h per day for two consecutive days using SQ23 Wireless HD 1080P mini cameras to observe the acclimation behavior. Video recordings were used to qualitatively confirm completion of acclimation prior to tissue sampling. Acclimation was considered complete when fish could swim freely among anemone tentacles without any visible distress or avoidance. Quantitative assessment of the behavioral dataset was not done as the behavioral acclimation process in anemonefish is already well documented in the literature (Mariscal 1970; Balamurugan et al. 2014). No water changes were conducted during the experiment, and the water flow was turned off to avoid altering olfactory cues that anemonefish rely on to recognize anemones. Fish and anemones were allowed to interact for a period of 48 h, a duration shown to be sufficient for full acclimation (Mariscal 1970; Hoepner et al. 2022), which was also confirmed by observations in our study. A similar time frame of interaction was used in a previous study investigating whether resistance of fish to anemones is facilitated by transfer of mucous antigens from anemones to fish (Elliott et al. 1994). After 48 h, all fish were euthanized by severing the spine, standard length was measured, and skin and five brain regions (diencephalon, telencephalon, optic tectum, cerebellum, and brain stem) were dissected and snap frozen in liquid nitrogen. All fish were sampled on the same day between 11 AM and 2:10 PM, alternating between control and interaction individuals, to minimize circadian effects on gene expression (Table S1). Skin tissue was sampled from one lateral side of the body, spanning approximately from the anterior end of the dorsal fin to the anterior end of the caudal peduncle, capturing all the distinctive band colors. All fish were sexed by visual examination of their gonads. A total of ten tentacles were sampled from each individual anemone at the end of the experiment and snap frozen in liquid nitrogen. The pedal disc diameter of all anemones was measured prior to dissection. All tissue samples were stored at −80 °C until further processing.

RNA Extraction, Sequencing, and Sequence Processing

Frozen skin tissue samples from A. clarkii were homogenized in buffer RLN (Qiagen), which was precooled to 4 °C, for 5 min at 30 cycles/s using Qiagen TissueLyzer II with 10 to 15 single-use silicon beads per sample. Frozen brain regions were homogenized in buffer RLN (Qiagen), which was precooled to 4 °C, using sterile micropestles. Homogenates were then centrifuged, and the supernatant was mixed with buffer RTL. The Qiagen RNeasy Mini Kit was subsequently used to extract total RNA. On-column DNase I treatment (Qiagen) was incorporated during the extraction process, according to the manufacturer's instructions, to remove potential DNA contamination.

Total RNA was extracted from frozen anemone tentacles following the protocol from Warner and Röttinger (2021), with slight modifications using Phasemaker tubes (Invitrogen). Frozen tissue samples were directly homogenized in TRIzol (Life Technologies) using sterile micropestles, and a second extraction step was performed by adding 250 µL of TRIzol and 250 µL of chloroform:isoamyl alcohol (24:1; Sigma-Aldrich). To remove any DNA contamination, in-solution DNase digestion was performed using DNAse 1 (Qiagen) followed by RNA cleanup (Qiagen) according to the manufacturer's instructions.

The concentration of extracted RNA from all the samples was measured using Qubit, and RNA quality was determined using Agilent 2100 Bioanalyzer. Samples with an RNA integrity number (RIN) > 6.5 were used for cDNA library preparation using KAPA mRNA HyperPrep Kit with 100 ng of total RNA as starting material. The mRNA samples were subsequently sequenced on Illumina NovaSeq 6000 to generate 151 bp paired-end reads at the Centre for PanorOmic Sciences, The University of Hong Kong.

Quality of all raw reads was determined using FastQC (Andrews, Simon 2010) v0.11.8. Adapters and low-quality reads (Phred-scaled quality score <20 for fish skin and brain and <25 for anemone) were removed using Trimmomatic (Bolger et al. 2014) v0.39 with a 4-base sliding window approach using the parameters ILLUMINACLIP:adapters.fa:2:30:15:8:true MINLEN:40. Different quality thresholds were chosen for anemone and fish samples due to differences in average quality of raw reads. Additionally, for anemone samples, overrepresented sequences reported by FastQC were also removed. The trimmed sequences from both species were further filtered to remove potential contaminant sequences from bacteria, fungus, and virus using Kraken (Wood and Salzberg 2014) v2.0.8-beta with a confidence score of 0.7 and 0.3 for A. clarkii and E. quadricolor, respectively.

An average of 24 ± 4.1 and 31 ± 2.2 million high-quality read pairs from A. clarkii skin and brain tissues (Table S1) were then mapped to the reference genome (Moore et al. 2023) using HISAT2 (Kim et al. 2019) v2.2.1 (parameters --no-mixed --no-discordant). On average, 70.68 ± 12.61% and 83.68 ± 4.23% of the sequenced reads from skin and brain samples, respectively, mapped to the reference genome (Table S1). FeatureCounts (Liao et al. 2014) v2.0.6 was subsequently used to obtain raw read counts per gene with multimapped reads being assigned fractional counts (parameters: -M --fraction -B -p --countReadPairs).

De Novo Transcriptome Assembly and Annotation for E. quadricolor

Currently, no reference genome exists for E. quadricolor, so a de novo transcriptome assembly was constructed. Adapter-free, quality trimmed, and decontaminated sequences of E. quadricolor were further filtered using the “filter_illumina” script from DRAP (Cabau et al. 2017) v1.92 to identify and remove reads containing ambiguous bases (Ns). Specifically, for all reads containing ambiguous bases, the longest subsequence without “N” was extracted and retained if it constituted at least half of the original sequence length and was 40 bp or longer. An average of 23 ± 1.5 million read pairs from eight individuals (which retained the most reads after all filtering steps; Table S1) were merged and assembled de novo using DRAP with Trinity (Grabherr et al. 2011) as the assembler using the pooled assembly strategy (parameters: -c “-x” --no-trim --no-norm -m bwa --no-rate). The initial transcriptome assembly consisted of 463,812 contigs.

Since sea anemones contain endosymbiotic dinoflagellates (Symbiodiniaceae), the initial transcriptome assembly contained a mixture of anemone and dinoflagellate transcripts, as evidenced by two peaks in the GC content plot (Fig. S2(a)). Anthozoans and dinoflagellates have drastically different GC contents, with Anthozoa having ∼40% and Symbiodiniaceae having ∼55% (Kashimoto et al. 2022). This difference in GC content can be used to separate transcripts belonging to dinoflagellates from those of the anemone as done previously (Kashimoto et al. 2022). To disentangle the meta-transcriptome, the initial transcriptome assembly was blasted (using BLASTN; parameters: -evalue 1e-20 -max_hsps 1) against a collection of Symbiodiniaceae genomes obtained from the NCBI (Sayers et al. 2022) and SAGER (Yu et al. 2020) databases (Table S9). All transcripts with BLAST hits to Symbiodiniaceae and GC content above 40% were removed from the transcriptome assembly. The resulting “cleaned” transcriptome consisted of 397,169 contigs and only a single peak in the GC content plot (Fig. S2(b)). Of the 66,643 transcripts identified as dinoflagellate sequences and removed from the initial de novo assembly, the majority were from the genus Cladocopium (Fig. S2(c)). This is similar to the findings of Kashimoto et al. (2022), who found that sea anemones from the fringing reefs of Okinawa form symbiotic relationships predominately with Cladocopium.

To reduce redundancy in the transcriptome assembly, TransDecoder (Haas et al. 2013) v5.5.0 was used on the “cleaned” assembly to identify transcripts containing coding regions (open reading frames [ORFs]), and only those containing ORFs over 100 amino acids were retained. This threshold corresponds to the TransDecoder default and was chosen to minimize false positive ORF predictions (Haas et al. 2013), but we note that this threshold will have excluded any short cnidarian peptide toxins (e.g. Kunitz peptides and ShK toxins) whose ORFs fall below this cutoff. These were then blasted against the UniProtKB/Swiss-Prot Release 2023_01 database (-evalue 1e-5 -max_target_seqs 1), and only the best ORF from each contig was retained, prioritized first based on blast homology and then based on ORF length. A final clustering step was performed using Corset (Davidson and Oshlack 2014) v1.09, which hierarchically clusters contigs to genes, and the longest transcript per gene cluster was extracted using the fetchClusterSeqs.py script from Corset-tools (https://github.com/Adamtaranto/Corset-tools). The final transcriptome assembly consisted of 38,887 high-quality contigs, each corresponding to a single gene. Completeness of the final transcriptome assembly was determined using BUSCO v4.1.3 (Manni et al. 2021), using the Metazoa_odb10 set of Benchmarking Universal Single-Copy Orthologs. The percentage of complete BUSCOs was above 90% with only a small fraction (6.4%) being duplicated (C: 91.5% [S: 85.1%, D: 6.4%]; Fig. S3), indicating that the assembly is highly complete. Basic statistics of the transcriptome assembly was determined using “TrinityStats.pl” from Trinity accessory scripts (https://github.com/trinityrnaseq/trinityrnaseq/tree/master/util), rnaQUAST (Bushmanova et al. 2016) v2.1.0, and “stats.sh” script from BBMap (Bushnell 2014) v38.87. Assembly contiguity was also sufficiently good with an N50 of 2,802 (see Table S10 for additional statistics). Filtered RNA-Seq reads from all ten anemone samples were mapped to the final de novo assembly using Bowtie2 (Langmead and Salzberg 2012) v2.5.1, and the transcript expression levels were quantified using RSEM (Li and Dewey 2011) v1.3.1.

The final de novo transcriptome was annotated by blasting against the UniRef90 database (downloaded on 2023 July 19; Suzek et al. 2007) using DIAMOND (Buchfink et al. 2021) v2.1.8 (blastx --max-target-seqs 1 -e 1e-5). Functional annotation of transcripts with GO terms was carried out in OmicsBox (Gotz et al. 2008) v1.4.11. Additionally, eggNOG-Mapper (Huerta-Cepas et al. 2017) v2.1.0 with EggNOG (Huerta-Cepas et al. 2019) v5.0.2 was used for annotation based on orthology predictions, and the results were merged with the GO annotations, resulting in a total of 21,216 (54.56%) annotated transcripts.

Differential Expression Analyses

Principal component analysis using regularized log-transformed (rlog) counts was performed in R v4.2.1 (R Core Team 2022) to explore gene expression patterns of A. clarkii skin and brain samples and E. quadricolor tentacles and detect and remove any outlier samples. To assess the influence of sex on the overall gene expression pattern in A. clarkii brain and skin samples, Pearson correlation values between samples were computed from the rlog transformed counts and visualized as hierarchically clustered heatmaps annotated by sex and condition (and additionally by brain region for the brain samples) using the pheatmap function in R. A LRT was then performed in DESeq2 (Love et al. 2014) v.1.32.0 in R v4.2.1, comparing a full model against a reduced model without the sex factor. For brain, the full model (∼Sex+Condition+Brain_region) was compared against a reduced model (∼Condition+Brain_region); and for skin, the full model (∼Sex+Condition) was compared against the reduced model (∼Condition). A significant effect of sex on gene expression (Padj < 0.05) was detected for two genes in the skin and 365 genes in the brain. To identify genes responding to the interaction treatment, pairwise comparisons were carried out between the control and interaction treatments separately for skin, brain regions, and tentacle samples using the Wald test in DESeq2. For A. clarkii, sex was included as a fixed-effect covariate in the design (∼Sex+Condition) to account for its effect on gene expression. For all pairwise comparisons, genes with FDR adjusted P-value less than 0.05, absolute log2 fold change >0.3, and baseMean >10 were considered to be significantly DE. To determine the independence of the interaction treatment effect from sex-biased expression in A. clarkii, the genes significantly affected by sex in the LRT were compared with the significant Wald test results separately for brain (across all five regions combined) and skin. Functional enrichment analyses were performed for all significantly DE gene sets using Fisher's exact test in OmicsBox v1.4.11 with FDR corrected P-value cutoff 0.05 and the option to “reduce to most specific” GO terms. For A. clarkii, significantly DE genes in the skin and brain regions were annotated and categorized into broader functional groups, where possible, based on their NCBI gene database descriptions (https://www.ncbi.nlm.nih.gov/gene/). For E. quadricolor, due to the substantially larger number of DE genes, only the genes associated with each enriched GO term were further categorized into broader functional groups when possible based on their description from the NCBI gene database (https://www.ncbi.nlm.nih.gov/gene/). In both species, these functional groupings should be interpreted as broad descriptive groupings rather than confirmed protein identities.

Supplementary Material

Supplementary material is available at Genome Biology and Evolution online.

Author Contributions

The experiment was designed by S.S. and C.S. and run by S.S. with input from C.S. and help from D.R. Anemone and fish dissections were carried out by all authors. Molecular lab work was performed by S.S. S.S. carried out the RNA-Seq data analysis with input from C.S. S.S. lead the writing of the manuscript with input from C.S., and all authors read, edited, and approved the final manuscript.

Funding

This study was supported by a seed fund to C.S. from the University of Hong Kong as well as the General Research Fund (17104424) by the Research Grants Council Hong Kong. D.R. was funded by the Hong Kong PhD fellowship.

Data Availability

The RNA-Seq raw sequences are deposited in NCBI under BioProject ID PRJNA1056457. Code used for analyses is available at https://github.com/sneha100895/Anemone-anemonefish-mutualism-RNA-Seq

Ethics

Sample collection was carried out following all institutional and national law guidelines. The experiment was completed under the approval of the Committee on the Use of Live Animals in Teaching and Research (CULATR), University of Hong Kong (# 5830-21), and the Animals (Control of Experiments) Ordinance (cap. 340), Director of Health, Government of the Hong Kong Special Administrative Region (# 21-1041/1042/1043/1044).

Declaration of AI Use

The authors used OpenAI's GPT-5 (ChatGPT) and Claude Sonnet 4 for minor wording improvements. No unpublished or sensitive data were shared.