Gene losses, parallel evolution and heightened expression confer adaptations to dedicated cleaning behaviour
BMC Biology volume 21, Article number: 180 (2023)
Cleaning symbioses are captivating interspecific interactions in which a cleaner fish removes ectoparasites from its client, contributing to the health and diversity of natural fish communities and aquaculture systems. However, the genetic signatures underlying this specialized behaviour remain poorly explored. To shed light on this, we generated a high-quality chromosome-scale genome of the bluestreak cleaner wrasse Labroides dimidiatus, a dedicated cleaner with cleaning as primary feeding mechanism throughout its life.
Compared with facultative and non-cleaner wrasses, L. dimidiatus was found with notable contractions in olfactory receptors implying their limited importance in dedicated cleaning. Instead, given its distinct tactile pre-conflict strategies, L. dimidiatus may rely more heavily on touch sensory perception, with heightened gene expression in the brain in anticipation of cleaning. Additionally, a reduction in NLR family CARD domain-containing protein 3 might enhance innate immunity of L. dimidiatus, probably assisting to reduce the impacts from parasite infections. In addition, convergent substitutions for a taste receptor and bone development genes across cleaners (L. dimidiatus and facultative cleaners) may provide them with evolved food discrimination abilities and jaw morphology that differentiate them from non-cleaners. Moreover, L. dimidiatus may exhibit specialized neural signal transductions for cleaning, as evidenced by positive selection in genes related to the glutamatergic synapse pathway. Interestingly, numerous glutamate receptors also demonstrated significantly higher expression in L. dimidiatus not engaged in cleaning, as compared to those involved in cleaning. Besides, apparent contractions in L. dimidiatus for protocadherins, which are responsible for neuronal development, may further promote specialized neural signal transductions in this species.
This study reveals that L. dimidiatus harbours substantial losses in specific gene families, convergent evolutions across cleaners and a large-scale high gene expression in preparation for cleaning, allowing for adaptation to the dedicated cleaning behaviour.
Cleaning symbioses are cooperative interspecific interactions between a cleaner and a usually larger client, where a cleaner removes and consumes materials that can negatively impact a client. These cleaning interactions are widespread among marine animals, and a total of 208 fish and 51 shrimp species have so far been described as cleaners . A cleaner fish can ingest a variety of food items including ectoparasites, mucus, injured tissue, or other particles from many client fish species [2, 3]. These cleaning interactions provide benefits for both partners: the cleaner fish receives food, while the client experiences a reduction in ectoparasite load which may otherwise lead to disease, growth reduction and diminished reproductive success . Cleaner fish have been experimentally proven to directly influence fish abundance and species richness in coral reef fish communities [5, 6]. Moreover, they have been shown to enhance aquaculture productions by serving as a biological control against ectoparasites [7, 8].
According to the persistence of cleaning behaviour throughout their life stages, cleaner fishes can be categorized as either dedicated or facultative cleaners. Dedicated cleaners are specialized in feeding almost exclusively by cleaning throughout their entire lives. In contrast, facultative cleaners rely only in small part on cleaning as a food source and/or perform cleaning only during juvenile period [1, 9]. Among 208 reported cleaner fish species, the majority (93%) are facultative cleaners, while only 16 species are dedicated cleaners that depend on cleaning for nearly all of their food [1, 4]. Cleaning behaviours in fish are restricted to the two fish families, the Gobiidae (gobies: 9 dedicated and 5 facultative cleaners) and Labridae (wrasses: 6 dedicated and 62 facultative cleaners) .
Several distinct behavioural and morphological characteristics have been proposed as contributions for adaptations to cleaning, such as client attraction methods, colouration and mouth morphology. To initiate cleaning interactions, some cleaners, such as the bluestreak cleaner wrasse Labroides dimidiatus, normally stay in their territories called cleaning stations and perform a “dance”, a vertical zig-zag swimming pattern, to attract clients [10, 11]. Furthermore, the striking colouration of cleaner fish, characterized by vivid lateral stripes (typically blue or yellow), enhances their visibility and effortlessly attracts clients for cleaning [4, 12]. In addition, a subterminal mouth and small body size of cleaners may enable them to effectively remove ectoparasites from larger clients [13, 14]. Limited molecular studies on cleaner fish indicate that immediate early genes, glutamate receptors and neurohormones exhibit significant expression changes in brain regions of L. dimidiatus during cleaning interactions, and hence play critical roles in cleaning behaviours . Despite parasites appearing as tiny dark dot on clients’ surface and potentially transfer to cleaners causing infections during close-contact cleaning [16, 17], it remains unknow how cleaners can detect parasites efficiently and minimize the infection risk.
The bluestreak cleaner wrasse L. dimidiatus, intensively studied for its cleaning behaviour, can interact with over 100 reef fish species  and have over 2000 interactions with client fish individuals each day , including large predators [12, 19]. However, a major conflict is present since L. dimidiatus prefer to eat the protective mucus from their clients, which constitutes cheating [20, 21]. This is the likely cause for the evolution of highly sophisticated cognition and decision rules used by cleaners during interactions. Thus, L. dimidiatus are used as a prime example of sophisticated fish cognition, due to their capacity to display self-recognition [22, 23], remember their previous interactions with clients , social tool use , social learning  and reputation management . To investigate the genetic basis underlying these behavioural adaptations needed to perform cleaning, we assembled a high-quality chromosome-level genome of L. dimidiatus. By comparing L. dimidiatus genome to genomes of seven other Labridae species including five facultative cleaners (Thalassoma bifasciatum, Symphodus melops, Tautogolabrus adspersus, Semicossyphus pulcher, Labrus bergylta)  and two non-cleaners (Cheilinus undulatus, Notolabrus celidotus), we examine gene family size, positive selections and convergences, as well as transcriptional regulation for dedicated cleaner L. dimidiatus, to pinpoint the molecular adaptations and convergent genomic features underlying the essential cleaning behaviour.
Genome sequencing assembly and annotation
For the de novo genome assembly of L. dimidiatus, 22.6 gigabase (Gb) PacBio CCS (HiFi) reads (approximately 23-fold coverage) were generated. After removing haplotigs and contig overlaps, the assembly length is 726.36 megabases (Mb) with 256 scaffolds and N50 length of 10.36 Mb. 42.2 Gb of Omni-C (Dovetail) data corresponding to approximately 42-fold coverage were integrated for a final 726.38 Mb chromosome-scale assembly with 57 scaffolds and an N50 of 33.59 Mb (Additional file 1: Table S1), and the 24 largest scaffolds containing 92% of protein-coding genes were deemed as chromosomes (Additional file 2: Fig. S1; Additional file 1: Table S2). The L. dimidiatus genome was annotated based on OrthoDB proteins and RNA-seq data, which predicted 37,023 and 61,565 genes respectively. By combining the two predictions, the final annotations indicated 28,138 genes, of which 23,551 (83.7%) genes showed homology with the proteins in Swiss-Prot or vertebrate_other (2021–09-12) in RefSeq. BUSCO analysis  on 28,138 predicted genes in L. dimidiatus and its genome assembly, revealed that 94.9% (4,349) and 96.7% (4,430) conserved and complete genes were detected using 4584 Actinopterygii genes as reference (Additional file 1: Table S3). Along the L. dimidiatus genome, around 35.53% (258.08 Mb) is repeat content, and 27.35% (198.67 Mb) is composed of transposable elements (Additional file 1: Table S4).
Dynamics of gene family size and evolution of Labridae fish species
Of 22,528 orthogroups among fourteen Labridae fish species with available reference genomes (Fig. 1), 13,826 conserved gene families were retained for investigating the dynamics of gene family size. The eight Labridae fish species diverged ~ 75.29 Million years ago (Mya) and shared a most recent common ancestor with stickleback around 92.17 Mya (Additional file 2: Fig. S2) based on the phylogenetic tree constructed with 2915 single-copy genes (Fig. 1A). Among eight Labridae fish species, L. dimidiatus as the only dedicated cleaner fish exhibited 96 contracted gene families (Additional file 1: Table S5) and 32 expanded gene families. The contractions in L. dimidiatus were more notable when compared to its two closely related fishes, a facultative cleaner T. bifasciatum and a non-cleaner fish N. celidotus, such as olfactory receptors (ORs), NLR family CARD domain-containing protein 3 and protocadherins (Additional file 1: Table S5).
Among eight Labridae fishes, the dedicated cleaner L. dimidiatus might experience a stronger purifying selection with a relatively lower dN/dS (nonsynonymous-synonymous substitution ratio, mean dN/dS = 0.1106, Fig. 1B], which was only higher than one non-cleaner species C. undulatus (mean dN/dS = 0.0986, Wilcoxon rank sum test, p = 0.1407) and one facultative cleaner T. bifasciatum (mean dN/dS = 0.0944, p < 2.2e − 16).
Sensory receptor genes
As a main food of cleaner fish, gnathiid isopods are ectoparasites that appear as small dark dots on clients . Olfaction and vision might be critical for cleaner fish to locate the parasites efficiently. Hence, we test for changes in copies of olfactory receptors (ORs) and vision opsin genes. The dedicated cleaner L. dimidiatus has 46 ORs, a minimum number among eight Labridae fish species (Additional file 1: Table S6). Compared with two closely related fish species, the facultative cleaner T. bifasciatum and the non-cleaner: N. celidotus, with 121 and 101 ORs, respectively. The number of subfamilies δ and ζ are contracted in L. dimidiatus, with 34 δ and 3 ζ ORs (Fig. 2, Additional file 2: Fig. S3). The facultative cleaner T. bifasciatum has 94 δ and 10 ζ OR s, while the non-cleaner N. celidotus has 69 δ and 17 ζ ORs. In addition, we examined the visual opsin genes across dedicated, facultative cleaners and non-cleaners for a potential divergence in visual senses. Surprisingly, a non-cleaner C. undulatus possessed 16 opsins (Additional file 2: Fig. S4), the largest number among all Labridae species, which was also reported in a previous study . Asides from C. undulatus, the other Labridae species displayed a similar number of opsin genes. The only differences between L. dimidiatus and its two closely related species (T. bifasciatum, N. celidotus) lie in the number of rhodopsin 2 (RH2). L. dimidiatus exhibited three RH2, whereas T. bifasciatum and N. celidotus possess four RH2. In addition to olfactory receptors and opsin genes, we also examined the gene number of glutamate receptors (GluRs), comprising ionotropic (iGluRs) and metabotropic (mGluRs) types, which were found to be similar between cleaner and non-cleaner species (Additional file 1: Table S7). Regarding iGluRs, the non-cleaner N. celidotus has the fewest number (19 genes), while others have 23–26. As for mGluRs, the eight Labridae species exhibited 9–19 genes, with L. dimidiatus (14 genes) similar to facultative cleaners (S. pulcher: 13, T. adspersus: 14) and non-cleaners (C. undulatus: 15, N. celidotus: 14).
Innate immune system
Pathogen recognition receptors (PRRs) are critical for the detection of pathogens to initiate innate immune defense . Here we investigated the repertoires of three major PRR families: RIG-like receptors (RLRs), NOD-like receptors (NLRs) and Toll-like receptors (TLRs). L. dimidiatus has 31 PRRs (2 RLRs, 16 NLRs, 13 TLRs; Additional file 2: Fig. S5), while T. bifasciatum and N. celidotus have 52 (3 RLRs, 33 NLRs, 16 TLRs) and 42 PRRs (2 RLRs, 29 NLRs, 11 TLRs), respectively. In particular, L. dimidiatus displayed contractions in a NLR gene (NLRC3: NLR family CARD domain-containing protein 3) with a minimum number (five copies of NLRC3) among eight Labridae fish species (Fig. 3A), and T. bifasciatum and N. celidotus have 19 and 18 respectively.
Protocadherins are homophilic cell adhesion molecules required for neuronal development and synaptic specificity . However, the dedicated cleaner L. dimidiatus may have experienced contractions in protocadherins, notable for the alpha and gamma subunits. The dedicated cleaner L. dimidiatus genome encodes 22 protocadherin α (Additional file 2: Fig. S6) and 32 protocadherin γ genes (Additional file 2: Fig. S7), while facultative cleaner T. bifasciatum and non-cleaner N. celidotus have 34 α and 35 γ genes, and 25 α and 41 γ genes respectively. The contractions in L. dimidiatus were most notable for an alpha (protocadherin alpha-2: PCDA2, Fig. 3B) and a gamma subunit (protocadherin gamma-A11: PCDGB, Fig. 3C). Among eight Labridae fish species, L. dimidiatus exhibited a minimum number of PCDA2 (one PCDA2) and the second minimum number of PCDGB (ten PCDGB), while T. bifasciatum has eight PCDA2 and 17 PCDGB, and N. celidotus has two PCDA2 and 19 PCDGB.
Estimation of positive selection and convergence
The dedicated cleaner L. dimidiatus exhibits 162 positively selected genes (PSGs) identified by PAML . Accounting for the impacts of multi-nucleotide substitutions on natural selection detection, BUSTED-MH in HYPHY  was also applied to detect the PSGs, which indicated 45 genes were under positive selection. The 41 genes detected as PSGs by both PAML and HYPHY were considered as the final PSGs (Additional file 1: Table S8). Among these PSGs, it is noteworthy that glutamate receptor 3 (GRIA3) and adenylate cyclase type 1 (ADCY1) are linked to glutamatergic synapse pathway (Fig. 4A), growth/differentiation factor 2 (GDF2) also known as bone morphogenetic protein 9 (BMP9), plays a role in regulating cartilage and bone development.
We detected convergence at conservative sites (CCS)  to estimate genomic convergences among cleaners. The simulation of amino acid sequences of fish species in the phylogeny of this study (Fig. 4A) indicated that the CCS method reduced random convergences and false convergences by 93.3% (from 10,277 to 687) and 100% (from 1040 to 0, Additional file 1: Table S9). To further remove random convergences, only the genes with CCS in all non-cleaners but not detected in any cleaners were treated as convergent evolving genes (CEGs) in cleaners, which revealed 38 parallel amino acid residue substitutions in 38 genes (Additional file 1: Table S10) across all cleaners (including dedicated and facultative cleaners). Among 38 CEGs, taste receptor type 1 member 3 (TAS1R3) is associated with taste sensory perception, and four genes (bone morphogenetic protein: BMP10; leucine-rich repeat-containing protein 17: LRRC17; chondroadherin: CHAD; thrombospondin-4-B: THBS4B) are involved in bone morphogenesis (Fig. 4B).
Cleaning behaviour gene expression in brain regions
To evaluate molecular signals related to cleaning interactions in the genome of dedicated cleaner L. dimidiatus, transcriptomic data were compared between interacting and non-interacting L. dimidiatus individuals, which revealed 2735, 1582 and 512 differentially expressed genes (DEGs, Additional file 1: Table S11) in the forebrain (FB), hindbrain (HB) and midbrain (MB), respectively. The majority of genes were reduced in expression in interacting L. dimidiatus (Fig. 5A), with 65, 70 and 61% downregulated DEGs in FB, HB and MB when compared to non-interacting L. dimidiatus individuals. A total of 4004 DEGs were detected in at least one of three brain regions, of which 56.9% (2,278 DEGs) exhibited a reduced expression across all regions of L. dimidiatus during cleaning interactions.
The number of DEGs with reduced expression was even more remarkable in signal transductions especially in the sensory perception of touch and glutamate transmission (Fig. 5B, Additional file 1: Table S12), of which 100% (23 of 23 genes) and 93.5% (43 of 46 genes) have a decreased expression during cleaning. In addition, the majority of DEGs related to behaviours, such as locomotory exploration behaviour (93.3%, 28 of 30 genes) and social behaviour (84.8%, 56 of 66 genes), were also downregulated in interacting L. dimidiatus individuals. Among 66 DEGs underlying social behaviour (Additional file 1: Table S13), two glutamate receptors (GRIN1, GRID1) and three glutamate decarboxylases (2 GAD1, GAD2) were downregulated in L. dimidiatus individuals during cleaning (Fig. 5C). Moreover, twenty-seven glutamate receptors (GluRs) and two glutamate receptor-interacting proteins (GRIPs) were significantly expressed between interacting and non-interacting L. dimidiatus individuals, and all these GluRs and GRIPs were lower expressed in interacting L. dimidiatus individuals (Fig. 5D, Additional file 1: Table S14). In addition, we also examined expression levels in L. dimidiatus for olfactory receptors (Additional file 2: Fig. S8) and vision opsin genes (Additional file 2: Fig. S9), which displayed a low expression with no significant differences between interacting and non-interacting L. dimidiatus individuals. Similar results were also found in protocadherins (Additional file 2: Fig. S10) and pathogen recognition receptors (PRRs, Additional file 2: Fig. S11) apart from a protocadherin (protocadherin alpha-C2: PCDC2) and a PRR (Toll-like receptor 7: TLR7) with a significantly different expression when interacting.
Cleaning behaviour is a critical foraging strategy of cleaner fish with tremendous importance in maintaining a healthy reef ecosystem and aquaculture systems . A chromosome-scale reference genome of L. dimidiatus, a prime candidate with dedicated cleaning behaviour, allowed for novel insights into the molecular adaptations of this important and fascinating behaviour. A variety of substantial gene family contractions were detected in L. dimidiatus, with positively selected genes and transcriptional changes in key functions. We further detected convergent evolution in key functions between dedicated and facultative cleaners allowing to pinpoint the genetic and functional basis of this complex cleaning behaviour.
Sensory perception of vision and olfaction
Animals commonly use olfaction and vision for foraging and to avoid predators [36, 37]. As the cleaner fish mainly eats parasites which appear as small dark dots on the surface of clients , it can be assumed that cleaner fish would have good olfaction or vision. Surprisingly, the number of vision opsin genes is similar across dedicated, facultative and non-cleaners showing no evidence of vision-related changes involved in cleaning behaviour. However, the olfactory receptors (ORs) exhibited distinctive massive contractions in the dedicated cleaner L. dimidiatus. These ORs contractions may reveal a decreased importance of olfaction in dedicated cleaners compared to other facultative and non-cleaner fish species. Fish species which feed on a variety of food sources may rely heavily on olfaction to detect different foods [38, 39]. However, the dedicated cleaner fish mainly depends on advertising their services through dancing and conspicuous colourful stripes to gain access to client fish [1, 10,11,12, 40]. Since cleaners inhabit areas with generally good visibility and client fishes directly approach cleaners asking for service, the requirement for olfaction might be reduced in cleaning interactions. In fact, the contraction of ORs can be linked to dietary transitions , and the observed difference between facultative and dedicated cleaners in the number of ORs may be a result of specialization in cleaning as facultative species only get half of their food from cleaning interactions . In addition, olfaction is essential for fish species in predator recognition as well as alarm cues from conspecifics warning of predation danger [42, 43]. However, predation on cleaners especially the dedicated cleaners have been rarely observed during cleaning even when they serve large predators [4, 44, 45]. Due to high specialization and dependency on cleaning, predator avoidances are not as key to survival as for the vast majority of other coral reef fishes. As such, olfactory receptors may have lost importance leading to the observed contractions in ORs.
Sensory perception of touch and taste
In addition to vision and olfaction, fish can also depend on tactile or taste sensory information to find and ingest food [46, 47]. Tactile stimuli also play an important role in the exploratory and social behaviour of fish through their tactile organs such as fins, barbels and dermal teeth [47,48,49]. Cleaner wrasses L. dimidiatus developed the capacity to use tactile stimulation (“massage”-like behaviour, where cleaners rub the body surface of client with their pectoral fins) as a pre-conflict managing strategy after dishonesty and as a client stress managing strategy [50, 51]. This pre-conflict strategy only evolved in cleaners from the clade labrichthyines (only L. dimidiatus in our studies species). These cleaners prefer to eat mucus instead of parasites (i.e., being dishonest), enhancing the need to employ pre-conflict strategies such as tactile stimulation to maintain interactions . Notably, the dedicated cleaner L. dimidiatus displayed twenty-three genes involved in sensory perception of touch with significantly increased expression when not interacting and decreased expression in interacting L. dimidiatus, indicating a possible preparedness for interaction. Of these twenty-three touch sensory perception genes, a glutamate receptor (GRIN1) and seven a-type potassium channel genes are known for extracting information from sensory inputs  and pain-sensing  respectively. Besides, feeding behaviour is also affected by their taste-discrimination capacity . The dedicated and facultative cleaners were detected with convergent evolving signals in a taste receptor TAS1R3, which is responsible for the sensory perception of sweet taste  and may thus contribute to differences in the food they ingest and even cause dietary transitions between cleaners and non-cleaners. Hence, dedicated cleaner fish may pre-heighten the touch sensory perception for cleaning interactions and food ingestions by their evolved taste sensory system.
Low-displacement and fast jaw movements enable cleaner fish to rapidly and dexterously touch clients using their subterminal mouths for the removal ectoparasites on clients [1, 14]. Bone morphogenetic proteins (BMPs) are pivotal morphogenetic signals related to the formation of bone and cartilage in fish [56, 57]. Here we found that the dedicated and facultative cleaners have a convergent evolution in BMP10, which was also seen to induce jaw deformity in golden pompano larvae when this gene was lowly expressed . Moreover, three other genes (LRRC17, CHAD, THBS4B) involved in bone development also exhibited convergences among all cleaners in our study. For instance, LRRC17  and CHAD  may contribute to effective amelioration of bone loss. Highly expressed THBS4B in articular chondrocytes is essential for maintaining articular cartilage integrity . Furthermore, the dedicated cleaner displayed positive selection in GDF2 (also known as BMP9), which can modulate dentinogenesis and tooth development [62, 63]. Therefore, the amino acid convergences observed in four genes across cleaners may play a role in the differences in jaw structures between cleaners (dedicated and facultative cleaners) and non-cleaners . Additionally, GDF2 could potentially contribute to further jaw divergence specifically in dedicated cleaners.
Enhanced immune defense
Parasites have been widely documented with negative impacts on the growth, survival and reproductive success of their fish hosts [65, 66]. Therefore, the cleaning benefits appear to be obvious; clients benefit from the removal of parasites and cleaners benefit with a source of food. However, the benefit probably comes with a cost for cleaners due to the transmission of the ectoparasites onto themselves from the clients during close-contact cleaning interactions [16, 17]. In addition, parasite infection pressure may be even higher around cleaning stations due to the occurrence of clients with a higher risk of parasitic infection . Hence, detecting the pathogens and initiating an innate immune defense is critical for cleaners to refrain from the parasitism risk. We found that NLRC3 (NLR family CARD domain-containing protein 3), a pathogen recognition receptor, exhibited substantial contractions in the dedicated cleaner L. dimidiatus when compared to its closely relative facultative and non-cleaner. NLRC3 can attenuate toll-like receptor signalling  and negatively regulate innate immune signalling induced by interferon genes , which could be helpful for L. dimidiatus in adjusting to elevated parasite loads. Since dedicated cleaners engage in more cleaning interactions than facultative and non-cleaners, dedicated cleaners are exposed to a higher parasitism risk; thus, a strong selective pressure to evolve a stronger innate immunity to avoid the negative impacts from parasites might be present.
Specializations in neural signalling transduction
As a dedicated cleaner fish that prefers to eat mucus (cheat), L. dimidiatus evolved a set of highly sophisticated cognition and decision-rules techniques to manipulate their clients. This highly sophisticated cognition may require a complex nervous system. However, the dedicated cleaner L. dimidiatus exhibited contractions in protocadherins, compared to facultative cleaner and non-cleaner. Protocadherins are homophilic cell adhesion molecules required for neuronal development as well as synaptic specificity . Substantial expansions of protocadherins have been reported to play a critical role in the formation of large and complex nervous systems of octopus . Hence, a reduced number of protocadherins may lead to a simplified or more specialized nervous system in dedicated cleaner L. dimidiatus. We further observed a divergence of glutamatergic transmission genes between cleaners (dedicated and facultative cleaners) and non-cleaners that might contribute to the specializations in neural signalling transduction of cleaners. As the most abundant excitatory neurotransmitter in the brain, glutamate plays a major role in learning and memory [71, 72], which can consolidate taste-recognition memory [73, 74]. In particular, dedicated cleaner L. dimidiatus showed positive selections in a glutamate receptor (GRIA3) and ADCY1, both of which are crucial for the release of glutamate . Furthermore, we found that twenty-nine glutamate receptors with significant expression differences between interacting and non-interacting L. dimidiatus, and twenty-seven glutamate receptors have significantly increased expression when not interacting as possible preparedness for interaction. Thus, protocadherin reductions and evolved glutamatergic transmission genes may allow for the dedicated cleaner L. dimidiatus to have evolved specializations in neural signalling transduction for social learning and cognition.
Cleaner wrasses specialized their behaviour and morphology to obtain food from social interactions. Here we produced a high-quality chromosome-scale genome of L. dimidiatus to investigate molecular adaptations underlying the dedicated cleaning behaviour by the comparison with facultative and non-cleaner fishes. L. dimidiatus experienced substantial contractions in olfactory receptors (ORs), innate immune receptors and protocadherins. Meanwhile, cleaners (dedicated and facultative) displayed divergences in genes associated with taste-discriminate, glutamatergic transmission and bone formation. In addition, the dedicated cleaner L. dimidiatus is more likely to exhibit elevated gene expression in brain regions when not interacting as a possible preparedness for cleaning, especially notable for genes related to touch sensory perception and glutamatergic transmission. Therefore, we conclude that, compared with facultative and non-cleaners, dedicated cleaning behaviour have a higher dependency on touch and taste than olfaction and vision for sensory perception, a distinct jaw for food ingestion, an enhanced immune response to lessen the impact of parasitism, and a specialized nervous system for neural signal transduction. Our results provide novel and important insights into molecular adaptations underlying dedicated cleaning behaviour.
High molecular weight DNA extraction and PacBio sequencing
A single blue-streak cleaner wrasse fish (6.8 cm), obtained from a local store, was used to obtain sample tissues. To obtain sufficient high-quality genomic DNA for the whole genome sequencing, brain (23 mg), gills (30.2 mg) and muscle (367.1 mg) were aseptically dissected out, snap-frozen in liquid nitrogen for at least 1 h and then stored at − 80 °C. DNA were extracted from muscle tissues according to the Qiagen Genomic DNA Handbook (Qiagen) and Genomic-Tip 500/G (Qiagen) procedure. The quality of the DNA was checked by agarose gel electrophoresis, and excellent integrity DNA molecules were observed.
HiFi PacBio library preparation and sequencing
DNA purity was assessed on a NanoDrop NP-1000 spectrophotometer (NanoDrop Technologies), DNA concentration was measured with a Qubit dsDNA high-sensitivity assay and DNA size was validated by pulsed-field gel electrophoresis (PFGE). Ten micrograms of DNA was sheared to the appropriate size range (10–30 kb) using a Covaris g-TUBE for the construction of PacBio HiFi sequencing libraries, followed by bead purification with PB Beads (PacBio). Sequencing libraries were constructed following the manufacturer’s protocol using a SMRTbell Express Template Prep Kit 2.0. Libraries were quantified using the Qubit dsDNA high-sensitivity assay, and size was checked on a Femto Pulse System (Agilent). Sequencing was performed on PacBio Sequel II systems in circular consensus sequencing (CCS) mode for 30 h.
Omni-C library preparation and sequencing
For each Omni-C library, chromatin was fixed in place with formaldehyde in the nucleus and then extracted. Fixed chromatin was digested with DNase I, chromatin ends were repaired and ligated to a biotinylated bridge adapter followed by proximity ligation of adapter containing ends. After proximity ligation, crosslinks were reversed and the DNA purified. Purified DNA was treated to remove biotin that was not internal to ligated fragments. Sequencing libraries were generated using NEBNext Ultra enzymes and Illumina-compatible adapters. Biotin-containing fragments were isolated using streptavidin beads before PCR enrichment of each library. The library was sequenced on an Illumina HiSeqX platform to produce ~ 30 × sequence coverage. Then HiRise used (See read-pair above) MQ > 50 reads for scaffolding.
De novogenome assembly and scaffolding
22.6 gigabase-pairs of PacBio CCS reads were used as an input to Hifiasm v0.15.4-r347  with default parameters. The initial assembly was mapped against the nucleotide sequence database by Blastn, and the results were used as input for blobtools v1.1.1  to remove the scaffolds that were identified as possible contamination in the assembly. Finally, purge_dups v1.2.5  was used to identify and remove both haplotigs and heterozygous overlaps.
The de novo assembly and Omni-C library reads were used as input for HiRise, a software pipeline designed specifically for using proximity ligation data to scaffold genome assemblies . Omni-C sequences were aligned to the draft input assembly using bwa (https://github.com/lh3/bwa). The read pairs mapped within draft scaffolds were applied to produce a likelihood model for genomic distance between scaffolds for joining. Based on Actinopterygii odb9 database comprising 4584 single-copy orthologs, the completeness of genome assembly was assessed by BUSCO v3.1.0  using the predicted genes and whole genome assembly respectively.
Mitochondrial genome assembly
Mitochondrial DNA is a useful and particularly popular marker for molecular ecology, population genetics and phylogenetic studies , while traditional genome assembly software can hardly assemble complete mitogenomes . With Notolabrus celidotus mitochondrial genome as the reference and a readpool by sampling 10% Ominc reads, we assembled the mitochondrial genome of L. dimidiatus using MITObim v1.9.1  by 30 iterations “-start 1 -end 30”, and MITObim reached a stationary read number after 30 iterations. The resulting assembly was annotated for genes using MitoAnnotator . To confirm the species of the sampled L. dimidiatus individual, we constructed a phylogeny based on Cytochrome c oxidase subunit I gene (COI), 12S and 16S of 10 Labridae species (L. dimidiatus, the individual in our study and an individual from NCBI; L. phthirophagus; L. bicolor; L. pectoralis; L. rubrolabiatus; Symphodus melops; Labrus bergylta; Thalassoma bifasciatum; Cheilinus undulatus; Notolabrus celidotus) and three other species (Medaka, Oryzias latipes; Fugu, Takifugu rubripes; Stickleback, Gasterosteus aculeatus; Zebrafish, Danio rerio; Platyfish, Xiphophorus maculatus; Spotted gar, Lepisosteus oculatus) (Additional file 1: Table S15).
RNA sequencing and gene expression analyses
For genome annotation and expression analyses, fore-, mid- and hindbrain tissue were obtained from 30 adult individuals of L. dimidiatus for RNA sequencing. The specimens were collected in the Maldives Islands and transported to the aquatic facilities of Laboratorio Maritimo da Guia in Cascais, Portugal by TMC-Iberia, further details can be found in  and Ramirez-Calero et al. (2023, unpublished). For the experiment evaluating transcriptional changes for cleaning behaviour, L. dimidiatus (N = 6) or clients (N = 6) were kept alone in the observation tank (control) as no-interaction treatment, while one cleaner (N = 6) and one client (N = 6) were kept together in the observation tank as the interaction treatment, allowing them to have close contact. Their interactions were filmed for 40 min since it is documented that this is the time frame in which neurohormones, and peptides are activated during cleaner interactions . At the end of the observation period, three separated regions (forebrain, midbrain, hindbrain) of the brain were immediately dissected out for each L. dimidiatus individuals for RNA sequencing. The details of sequencing and reads quality control can be found in .
To identify gene expression differences of three brain tissues between non-interacting and interacting L. dimidiatus individuals, high-quality reads were mapped against the L. dimidiatus genome assembled in this study with HISAT2 v2.1.0 . Read number matrices of all genes were generated using FeatureCounts v2.0.0  and then were used as input for DEseq2  to estimate the differential expressed genes (DEGs) between non-interacting and interacting L. dimidiatus individuals. DEGs should display with an FDR adjusted P value ≤ 0.05 and the average of the normalized count values (basemean) ≥ 10 as well as the effect size (Log2FoldChange) ≥ 0.3. Functional enrichment analyses were performed for all gene sets of interest by comparison with the whole gene data set with Fisher’s exact test in OmicsBox v2.0.29. Functions were accepted as significantly enriched with a false discovery rate (FDR) Padj < 0.05 and reduced to most specific terms.
Repeat elements and gene annotation
To compare the genomes of fish with cleaning behaviours, we obtained the sequences of seven other Labridae fish species, including five facultative cleaner (Thalassoma bifasciatum, GenBank: RPOG00000000.1; Symphodus melops, GenBank: MWVA00000000.1; Tautogolabrus adspersus, GCA_020745685.1; Semicossyphus pulcher, GCA_022749685.1; Labrus bergylta, GenBank: FKLU00000000.1) and two non-cleaner (Cheilinus undulatus, GenBank: GCA_018320785.1; Notolabrus celidotus, GenBank: GCA_009762535.1). Their assemblies and L. dimidiatus assembly were annotated by searching repeated elements using RepeatModeler v2.0.2 . The assembled genome sequences were used as input to generate a de novo library for RepeatModeler. Then, RepeatModeler was run (-LTRStruct) by combining the results of LTR structural discovery pipeline (LTR_Harvest and LTR_retreiver) and RepeatScout/RECON pipeline. Consensus sequences of the repeated elements were subsequently used to mask repeats in the assembly using RepeatMasker v4.1.2-p1 .
The soft-masked genome of L. dimidiatus was then predicted by BRAKER v2.1.6. [89,90,91,92,93,94,95,96,97], BRAKER ran twice respectively using OrthoDB proteins (–softmasking; –AUGUSTUS_ab_initio; –gff3; –prot_seq) and RNA-seq data on the fore-, mid- and hindbrain of 30 adult L. dimidiatus individuals (–softmasking; –AUGUSTUS_ab_initio; –gff3; –UTR = on; –bam). The two results obtained were combined using TSEBRA v1.0.3  with default parameters. For the five Labridae species, we also re-predicted their protein-coding genes using OrthoDB proteins. The longest transcripts of all predicted coding genes were selected for homology annotation with the proteins from Swiss-Prot and vertebrate_other by diamond v0.9.24.125 .
Estimation of gene family expansion and contraction
Gene family expansion or contraction is thought to be an important driving force to evolutionary novelties, such as cleaning behaviour. To examine and compare the gene family dynamics, in addition to the eight Labridae fish species, the protein sequences of another six reference fish species with high-quality genome were downloaded by BioMart from Ensembl (Japanese Medaka, Oryzias latipes ASM223467v1; Fugu, Takifugu rubripes fTakRub1.2; Stickleback, Gasterosteus aculeatus BROAD S1; Zebrafish, Danio rerio GRCz11; Platyfish, Xiphophorus maculatus X_maculatus-5.0-male; Spotted gar, Lepisosteus oculatus LepOcu1). For all of 14 fish species, the longest protein per gene was selected, the genes with protein sequences shorted than 30 amino acids or have early stop codons in the coding regions were removed. The protein sequences of remaining genes were applied to detect orthologous genes using the default parameters in OrthoFinder v2.3.3. The 13,861 of 22,528 orthogroups including at least one gene from zebrafish and genes among at least four reference fish species were deemed as conserved gene families. And the matrix including gene numbers of these 13,861 conserved gene families were parsed to identify gene family dynamics by CAFE v4.2.1 .
Phylogenetic analysis and divergence time estimation
The 2915 one-to-one orthologous genes of all 14 fish species were selected to construct the phylogenetic tree. MUSCLE v3.8.31  was applied to align the protein sequences; regions with poor quality were trimmed using trimAl v1.4.rev22  by the following parameters: -gt 0.8 -st 0.001 -cons 60. Then the sequences of proteins were concatenated for phylogeny detection by RaxML v8.2.11 . The parameter PROTGAMMAAUTO was used to select the optimal amino acid substitution model. Then MCMCtree  was used to investigate the divergence time under a relaxed-clock model (correlated molecular clock) with approximate likelihood calculation. First, the coding nucleotide sequences were aligned by MUSCLE v3.8.31  and trimmed by trimAl v1.4.rev22  with “-gt 0.9 -st 0.001 -cons 60”. Based on the above resulting phylogenetic tree and nucleotide sequences, substitution rate was roughly estimated by baseml under the general time reversible (GTR) model suggested by jModelTest . Then MCMCtree was run for the first time to estimate the gradient and Hessian, the output (out.BV) was used for the second running of MCMCTree to perform approximate likelihood calculations. The final Markov chain Monte Carlo process was run by “burnin = 50000; sampfreq = 100; nsample = 2000000”. We set two fossil calibrations: O. latipes–T. nigroviridis (~ 96.9–150.9 Ma), D. rerio–G. aculeatus (~ 149.85–165.2 Ma) and a time for the root (< 700 Ma).
Due to the potential involvement of olfaction and vision in the interaction behaviours displayed by L. dimidiatus, we compared the olfactory and vision genes among species. The olfactory receptor (OR) genes  were downloaded as query sequences and mapped to genome of fourteen fish species by tblastn with “1e − 5”. Solar (in-house software, version 0.9.6) was used to join the high-scoring segment pairs (HSPs) between each pair of protein mapping results if overlaps were detected in the mapping regions of two hits. The best alignments to the same mapping regions with lengths longer than 200 in the genome were kept. Subsequently, GeneWise v2-4–1  was applied to map the protein sequences to genome regions which extended 280 bp upstream and downstream of their mapping genome region. If a genomic nucleotide region were mapped by multiple query proteins, the one with highest GeneWise score were kept and the predicted protein sequence was annotated by Swiss-Prot. The genes with description including keywords “olfactory” or “odorant” were retained as putative OR genes. The hmmscan was applied to search against Pfam database to identify the domain with the highest score for each putative OR gene. The putative OR genes with a coding sequence from start codon to a stop codon were considered as intact OR genes, which were BLASTP to OR query sequences for OR subfamilies classification. These intact OR genes were aligned using MUSCLE, and the alignments were used to construct a phylogenetic tree by FastTree2 for the verification and correction of these putative OR genes. For the final maximum likelihood phylogenetic tree, the following eight non-OR genes  were used as the outgroup sequences: alpha-1A adrenergic receptor isoform 1 (NP_000671.2), beta-1 adrenergic receptor (NP_000675.1), adenosine receptor A2b (NP_000667.1), histamine H2 receptor isoform 2 (NP_071640.1), 5-hydroxytryptamine receptor 1B (NP_000854.1), 5-hydroxytryptamine receptor 1F (NP_000857.1), 5-hydroxytryptamine receptor 6 (NP_000862.1), galanin receptor type 1 (NP_001471.2) and somatostatin receptor type 4 (NP_001043.2). The putative OR gene sequences of eight Labridae fish species and eight non-OR genes were aligned by MUSCLE and trimmed by trimAl v1.4.rev22  by “ -gt 0.8 -st 0.001 -cons 60,” the resulting alignments were used to construct the maximum likelihood phylogenetic tree by RaxML v8.2.11  with PROTGAMMAAUTO to select the optimal amino acid substitution model.
The opsin genes (RH1: rhodopsin; RH2: middle-wavelength-sensitive opsin rhodopsin 2; LWS: long-wavelength-sensitive opsin; SWS1: short-wavelength-sensitive opsin 1; SWS2: short-wavelength-sensitive opsin 2) of spotted gar, zebrafish, medaka, platyfish, fugu and stickleback were used as query protein sequences. And these query protein sequences were mapped to genome of fourteen fish species by tblastn with “1e − 5”. Solar (in-house software, version 0.9.6) was used to join the high-scoring segment pairs (HSPs) between each pair of protein mapping results if overlaps were detected in the mapping regions of two hits. The best alignments to the same mapping regions with lengths longer than 50 in the genome were kept. Subsequently, Genewise was applied to map the protein sequences to genome regions which extended double of the query proteins length along upstream and downstream of their mapping genome region. If a genomic nucleotide region were mapped by multiple query proteins, the one with highest Genewise score were kept and its predicted protein sequences was annotated by Swiss-Prot; the genes were annotated as opsin genes which covered at least 70% and with e-value < 1e − 20 of the corresponding proteins in Swiss-Prot were retained as putative opsin genes, which were aligned by MUSCLE, and trimmed by trimAl v1.4.rev22  by “ -gt 0.8 -st 0.001 -cons 60” to construct the maximum likelihood phylogenetic tree by RaxML v8.2.11  with PROTGAMMAAUTO to select the optimal amino acid substitution model.
Protocadherin, innate immunity family and glutamate receptors
Using the same method as opsin gene identification, gene family members for protocadherin, innate immunity families and glutamate receptors were identified in fourteen fish species from the phylogenetic tree (Fig. 1A). Protocadherin alpha and gamma genes from Swiss-Prot were used as query proteins. Based on this, a domain search against the Pfam database was conducted using hmmscan, and genes with over six extracellular cadherin domains were considered putative protocadherin alpha and gamma genes . These genes were used to build a phylogenetic tree following the opsin gene identification method.
For innate immunity families, protein sequences of toll-like receptors (TLRs), RIG-like receptors (RLRs) and NOD-like receptors (NLRs) from six reference fish species (spotted gar, zebrafish, medaka, platyfish, fugu, stickleback) were used as queries. Predicted genes covering at least 70% and having an e-value < 1e − 20 were considered innate immunity family members and used for phylogenetic tree construction.
For glutamate receptors, protein sequences from six reference fish species (spotted gar, zebrafish, medaka, platyfish, fugu, stickleback) were used. The identified ionotropic glutamate receptors (iGluRs) with the SYTANLAAF motif  and metabotropic glutamate receptors (mGluRs) with seven transmembrane regions were retained .
The orthogroups generated by OrthoFinder v2.3.3  were divided into sub-orthogroups by Possvm v1.1 . The genes within sub-orthogroups were classified into different groups according to their gene names, and only the groups containing all of the 14 species were kept. For species with more than two genes in the subgroup, the one with highest blast score was kept. The genes in the 6929 subgroups were considered as orthologous genes for the following evolutionary analysis. The protein sequences of orthologous genes in the subgroup were aligned by Clustal Omega-1.2.4 (-t Protein; –outfmt = fa) . The protein alignments and corresponding coding nucleotide sequences were used as input for pal2nal v.14  to construct the protein–cDNA sequence pairs, and poorly aligned positions and divergent regions of cDNA were removed by Gblocks v.0.91b (options: -b4, 10; -b5, n; –b3, 5; –t = c) .
To explore the possibility of cleaners evolving more rapidly compared to non-cleaners, we used the free-ratio model of codeml in PAML v.4.9  to estimate the dN/dS (nonsynonymous-synonymous substitution ratio) along each lineage based on all 6929 orthologous genes. The results of each lineages under study were curated to reduce errors  by removing genes with any one of the following values: dS > 1, N > sequence length, N + S > sequence length by 50 or more bp and N*dN or S*dS < 1. The final gene number to estimate evolutionary rate were 5215 (L. dimidiatus), 5342 (T. bifasciatum), 5563 (S. melops), 4655 (T. adspersus), 5789 (S. pulcher), 5654 (L. bergylta), 5961 (C. undulatus), 5972 (N. celidotus), 6256 (Japanese Medaka: O. latipes), 6264 (Fugu: T. rubripes), 6259 (Stickleback: G. aculeatus), 4514 (Zebrafish: D. rerio), 6227 (Platyfish: X. maculatus) and 2219 (Spotted gar: L. oculatus). The mean dN/dS values of the qualified orthologous genes in the eight Labridae fish species were employed to compare their respective evolutionary rates.
To identify genes evolving under positive selection, the branch-site model of codeml in PAML v.4.9  was applied to investigate the positively selected genes (PSGs) for L. dimidiatus based on the species tree. The terminal branch of L. dimidiatus was set as the foreground branch, a likelihood ratio test (LRT) was used to estimate whether the branch-site model containing positively selected codons (omega > 1) fits better than the null model including neutral selection or negative selection (omega ≤ 1). Chi-square statistics wrapped in PAML were performed to estimate the P values for model comparison, and the P values were corrected by the false discovery rate (FDR) using R version 3.6.3. Only the genes with an LRT FDR < 0.05 and containing codon sites with a posterior probability of positive selection over 0.95 by the Bayes empirical Bayes (BEB) method were treated as PSGs. To account for the impact of multi-nucleotide substitutions on natural selection detection , we also used BUSTED-MH method implemented in HYPHY v2.5.51  to detect positively selected genes in L. dimidiatus with parameters “hyphy busted –alignment –tree –multiple-hits Double + Triple –starting-points 5 –branches”. With L. dimidiatus as the foreground branch, the LRT P value for episodic diversifying positive selection was corrected for FDR using R version 3.6.3. Genes with an LRT FDR < 0.05 were considered positively selected. Only genes detected by both PAML and HYPHY were deemed as the final positively selected genes.
To unveil convergent molecular evolution at the amino acid variation level underlying the cleaning behaviour in fish, we applied a method in Xu et al.  for detecting convergence at conservative sites (CCS) where all non-cleaner species (two non-cleaner Labridae fish and six reference species) shared the same amino acid. The noise estimation for the CCS method involved the following steps: (1) Sequence simulation: the amino acid sequences of 6,353 orthologous genes (7,720,731 amino acids) were concatenated to estimate branch lengths, amino acid frequencies and the best shape parameter for variable rates among sites (alpha) using codeml in PAML  with the JTT + gamma model. Using these parameters, we simulated amino acid sequences of the same length (7,720,731 amino acids) as the real data set with the evolver tool from PAML; (2) Ancestral state inference: the ancestral amino acid sequences were inferred using the empirical Bayesian ancestral reconstruction in codeml with the same parameters as the sequence simulation; (3) Accuracy estimation: sites were considered as convergences across cleaner fish if they were shared by at least three cleaner species and differed from any non-cleaner species. Among these convergences, sites were classified as random convergences if more than two cleaners showed the same amino acid as non-cleaners, and as false convergences if the ancestral amino acids of all non-cleaners (excluding the spotted gar Lepisosteus oculatus) were different with the spotted gar. The ancestral accuracy was estimated by comparing the amino acids of the spotted gar, and the ancestor of all non-cleaners (excluding spotted gar). The inferences of amino acids in ancestor were correct if they were identical with spotted gar; if not, the inferences were incorrect. To remove the random convergent sites, only sites in six cleaners that are not consistent with the CCS in all non-cleaners were considered as potential convergent evolution across six cleaners.
Availability of data and materials
Genomic sequences (HiFi PacBio and Omni-C) and the de novo genome assembly have been deposited in NCBI under BioProject PRJNA937036 (https://www.ncbi.nlm.nih.gov/bioproject/PRJNA937036) , the transcriptomic sequences can be retrieved under BioProject PRJNA726349 (https://www.ncbi.nlm.nih.gov/bioproject/PRJNA726349) .
Million years ago
Nonsynonymous-synonymous substitution ratio
Pathogen recognition receptors
Toll-like receptor 7
NLR family CARD domain-containing protein 3
Positively selected genes
Glutamate receptor 3
Adenylate cyclase type 1
Growth/differentiation factor 2
Bone morphogenetic protein 9
Convergence at conservative sites
Convergent evolving genes
Taste receptor type 1 member 3
Bone morphogenetic protein
Leucine-rich repeat-containing protein 17
Differentially expressed genes
Glutamate receptor-interacting proteins
Ionotropic glutamate receptors
Metabotropic glutamate receptors
Bone morphogenetic proteins
High-scoring segment pairs
Middle-wavelength-sensitive opsin rhodopsin 2
Short-wavelength-sensitive opsin 1
Short-wavelength-sensitive opsin 2
Likelihood ratio test
False discovery rate
Bayes empirical Bayes
Vaughan DB, Grutter AS, Costello MJ, Hutson KS. Cleaner fishes and shrimp diversity and a re-evaluation of cleaning symbioses. Fish Fish. 2017;18(4):698–716.
Grutter AS, Murphy JM, Choat JH. Cleaner fish drives local fish diversity on coral reefs. Curr Biol. 2003;13(1):64–7.
Bansemer C, Grutter AS, Poulin R. Geographic variation in the behaviour of the cleaner fish Labroides dimidiatus (Labridae). Ethology. 2002;108(4):353–66.
Cote IM. Evolution and ecology of cleaning symbioses in the sea. Oceanogr Mar Biol. 2000;38:311–55.
Waldie PA, Blomberg SP, Cheney KL, Goldizen AW, Grutter AS. Long-term effects of the cleaner fish Labroides dimidiatus on coral reef fish communities. PLoS ONE. 2011;6(6):e21201.
Demairé C, Triki Z, Binning SA, Glauser G, Roche DG, Bshary R. Reduced access to cleaner fish negatively impacts the physiological state of two resident reef fishes. Mar Biol. 2020;167:1–10.
Costello MJ. Ecology of sea lice parasitic on farmed and wild fish. Trends Parasitol. 2006;22(10):475–83.
Costello M. The global economic cost of sea lice to the salmonid farming industry. J Fish Dis. 2009;32(1):115.
Limbaugh C. Cleaning symbiosis. Sci Am. 1961;205(2):42–9.
Potts GW. The ethology of Labroides dimidiatus (cuv. & val.)(Labridae, Pisces) on Aldabra. Anim Behav. 1973;21(2):250–91.
Gorlick DL, Atkins PD, Losey GS Jr. Cleaning stations as water holes, garbage dumps, and sites for the evolution of reciprocal altriusm? Am Nat. 1978;112(984):341–53.
Cheney KL, Grutter AS, Blomberg SP, Marshall NJ. Blue and yellow signal cleaning behavior in coral reef fishes. Curr Biol. 2009;19(15):1283–7.
Huie JM, Thacker CE, Tornabene L. Co-evolution of cleaning and feeding morphology in western Atlantic and eastern Pacific gobies. Evolution. 2020;74(2):419–33.
Baliga VB, Mehta RS. Linking cranial morphology to prey capture kinematics in three cleaner wrasses: Labroides dimidiatus, Larabicus quadrilineatus, and Thalassoma lutescens. J Morphol. 2015;276(11):1377–91.
Ramírez-Calero S, Paula J, Otjacques E, Rosa R, Ravasi T, Schunter C. Neuro-molecular characterization of fish cleaning interactions. Sci Rep. 2022;12(1):1–12.
Kabata Z. Copepoda (Crustacea) parasitic on fishes: problems and perspectives. Adv Parasitol. 1982;19:1–71.
Narvaez P, Yong RQ-Y, Grutter AS, Hutson KS. Are cleaner fish clean? Mar Biol. 2021;168(5):59.
Grutter A. Parasite removal rates by the cleaner wrasse Labroides dimidiatus. Mar Ecol Prog Ser. 1996;130:61–70.
Grutter AS, Poulin R. Cleaning of coral reef fishes by the wrasse Labroides dimidiatus: influence of client body size and phylogeny. Copeia. 1998;1:120–7.
Grutter AS. Spatiotemporal variation and feeding selectivity in the diet of the cleaner fish Labroides dimidiatus. Copeia. 1997;2:346–55.
Grutter AS, Bshary R. Cleaner wrasse prefer client mucus: support for partner control mechanisms in cleaning interactions. Proc R Soc Lond, Ser B: Biol Sci. 2003;270(suppl_2):S242–4.
Kohda M, Bshary R, Kubo N, Awata S, Sowersby W, Kawasaka K, Kobayashi T, Sogawa S. Cleaner fish recognize self in a mirror via self-face recognition like humans. Proc Natl Acad Sci U S A. 2023;120(7):e2208420120.
Kohda M, Sogawa S, Jordan AL, Kubo N, Awata S, Satoh S, Kobayashi T, Fujita A, Bshary R. Further evidence for the capacity of mirror self-recognition in cleaner fish and the significance of ecologically relevant marks. PLoS Biol. 2022;20(2):e3001529.
Tebbich S, Bshary R, Grutter A. Cleaner fish Labroides dimidiatus recognise familiar clients. Anim Cogn. 2002;5:139–45.
Wismer S, Grutter A, Bshary R. Generalized rule application in bluestreak cleaner wrasse (Labroides dimidiatus): using predator species as social tools to reduce punishment. Anim Cogn. 2016;19:769–78.
Truskanov N, Emery Y, Bshary R. Juvenile cleaner fish can socially learn the consequences of cheating. Nat Commun. 2020;11(1):1159.
Binning SA, Rey O, Wismer S, Triki Z, Glauser G, Soares MC, Bshary R. Reputation management promotes strategic adjustment of service quality in cleaner wrasse. Sci Rep. 2017;7(1):1–9.
Simao FA, Waterhouse RM, Ioannidis P, Kriventseva EV, Zdobnov EM. BUSCO: assessing genome assembly and annotation completeness with single-copy orthologs. Bioinformatics. 2015;31(19):3210–2.
Grutter AS. Spatial and temporal variations of the ectoparasites of seven reef fish species from Lizard Island and Heron Island. Australia Mar Ecol Prog Ser. 1994;115:21–30.
Liu D, Wang X, Guo H, Zhang X, Zhang M, Tang W. Chromosome-level genome assembly of the endangered humphead wrasse Cheilinus undulatus: Insight into the expansion of opsin genes in fishes. Mol Ecol Resour. 2021;21(7):2388–406.
Tan M, Redmond AK, Dooley H, Nozu R, Sato K, Kuraku S, Koren S, Phillippy AM, Dove AD, Read T. The whale shark genome reveals patterns of vertebrate gene family evolution. Elife. 2021;10:e65394.
Zipursky SL, Sanes JR. Chemoaffinity revisited: dscams, protocadherins, and neural circuit assembly. Cell. 2010;143(3):343–53.
Yang Z. PAML: a program package for phylogenetic analysis by maximum likelihood. Comput Appl Biosci. 1997;13(5):555–6.
Murrell B, Weaver S, Smith MD, Wertheim JO, Murrell S, Aylward A, Eren K, Pollner T, Martin DP, Smith DM. Gene-wide identification of episodic selection. Mol Biol Evol. 2015;32(5):1365–71.
Xu S, He Z, Guo Z, Zhang Z, Wyckoff GJ, Greenberg A, Wu C-I, Shi S. Genome-wide convergence during evolution of mangroves from woody plants. Mol Biol Evol. 2017;34(4):1008–15.
Lin Q, Fan S, Zhang Y, Xu M, Zhang H, Yang Y, Lee AP, Woltering JM, Ravi V, Gunter HM. The seahorse genome and the evolution of its specialized morphology. Nature. 2016;540(7633):395–9.
Musilova Z, Cortesi F, Matschiner M, Davies WI, Patel JS, Stieb SM, de Busserolles F, Malmstrøm M, Tørresen OK, Brown CJ. Vision using multiple distinct rod opsins in deep-sea fishes. Science. 2019;364(6440):588–92.
Velez Z, Hubbard PC, Welham K, Hardege JD, Barata EN, Canário AV. Identification, release and olfactory detection of bile salts in the intestinal fluid of the Senegalese sole (Solea senegalensis). J Comp Physiol, A. 2009;195:691–8.
Yacoob S, Browman H. Olfactory and gustatory sensitivity to some feed-related chemicals in the Atlantic halibut (Hippoglossus hippoglossus). Aquaculture. 2007;263(1–4):303–9.
Randall JE. A review of the labrid fish genus Labroides, with descriptions of two new species and notes on ecology. Pac Sci. 1958;XII:327–47.
Niimura Y, Matsui A, Touhara K. Acceleration of olfactory receptor gene loss in primate evolution: possible link to anatomical change in sensory systems and dietary transition. Mol Biol Evol. 2018;35(6):1437–50.
McCormick M, Manassa R. Predation risk assessment by olfactory and visual cues in a coral reef fish. Coral Reefs. 2008;27:105–13.
Kelley JL, Magurran AE. Learned predator recognition and antipredator responses in fishes. Fish Fish. 2003;4(3):216–26.
Feder HM. Cleaning symbiosis in the marine environment. Symbiosis. 1966;1(S M):327–80.
Darcy GH, Maisel E, Ogden JC. Cleaning preferences of the gobies Gobiosoma evelynae and G. prochilos and the juvenile wrasse Thalassoma bifasciatum. Copeia. 1974;2:375–9.
Kasumyan AO. Tactile reception and behavior of fish. J Ichthyol. 2011;51:1035–103.
Kasumyan AO. The taste system in fishes and the effects of environmental variables. J Fish Biol. 2019;95(1):155–78.
Soares MC, Oliveira RF, Ros AF, Grutter AS, Bshary R. Tactile stimulation lowers stress in fish. Nat Commun. 2011;2(1):534.
Jamieson AJ, Bailey DM, Wagner H-J, Bagley P, Priede I. Behavioural responses to structures on the seafloor by the deep-sea fish Coryphaenoides armatus: implications for the use of baited landers. Deep Sea Res Part I Oceanogr. 2006;53(7):1157–66.
Bshary R, Würth M. Cleaner fish Labroides dimidiatus manipulate client reef fish by providing tactile stimulation. Proc R Soc Lond, Ser B: Biol Sci. 2001;268(1475):1495–501.
Grutter AS. Cleaner fish use tactile dancing behavior as a preconflict management strategy. Curr Biol. 2004;14(12):1080–3.
Stevens CF, Zador AM. Input synchrony and the irregular firing of cortical neurons. Nat Neurosci. 1998;1(3):210–7.
Chien L-Y, Cheng J-K, Chu D, Cheng C-F, Tsaur M-L. Reduced expression of A-type potassium channels in primary sensory neurons induces mechanical hypersensitivity. J Neurosci. 2007;27(37):9855–65.
Morais S. The physiology of taste in fish: potential implications for feeding stimulation and gut chemical sensing. Rev Fish Sci. 2017;25(2):133–49.
Lemon CH, Margolskee RF. Contribution of the T1r3 taste receptor to the response properties of central gustatory neurons. J Neurophysiol. 2009;101(5):2459–71.
Hogan B. Bone morphogenetic proteins: multifunctional regulators of vertebrate development. Genes Dev. 1996;10(13):1580–94.
Rajaram S, Murawala H, Buch P, Patel S, Balakrishnan S. Inhibition of BMP signaling reduces MMP-2 and MMP-9 expression and obstructs wound healing in regenerating fin of teleost fish Poecilia latipinna. Fish Physiol Biochem. 2016;42:787–94.
Ma Z, Hu J, Yu G, Qin JG. Gene expression of bone morphogenetic proteins and jaw malformation in golden pompano Trachinotus ovatus larvae in different feeding regimes. J Appl Anim Res. 2018;46(1):164–77.
Liu F, Yuan Y, Bai L, Yuan L, Li L, Liu J, Chen Y, Lu Y, Cheng J, Zhang J. LRRc17 controls BMSC senescence via mitophagy and inhibits the therapeutic effect of BMSCs on ovariectomy-induced bone loss. Redox Biol. 2021;43:101963.
Capulli M, Olstad OK, Önnerfjord P, Tillgren V, Muraca M, Gautvik KM, Heinegård D, Rucci N, Teti A. The C-terminal domain of chondroadherin: a new regulator of osteoclast motility counteracting bone loss. J Bone Miner Res. 2014;29(8):1833–46.
Jeschke A, Bonitz M, Simon M, Peters S, Baum W, Schett G, Ruether W, Niemeier A, Schinke T, Amling M. Deficiency of thrombospondin-4 in mice does not affect skeletal growth or bone mass acquisition, but causes a transient reduction of articular cartilage thickness. PLoS ONE. 2015;10(12):e0144272.
Huang X, Wang F, Zhao C, Yang S, Cheng Q, Tang Y, Zhang F, Zhang Y, Luo W, Wang C. Dentinogenesis and tooth-alveolar bone complex defects in BMP9/GDF2 knockout mice. Stem Cells Dev. 2019;28(10):683–94.
Lowery JW, Intini G, Gamer L, Lotinun S, Salazar VS, Ote S, Cox K, Baron R, Rosen V. Loss of BMPR2 leads to high bone mass due to increased osteoblast activity. J Cell Sci. 2015;128(7):1308–15.
Baliga VB, Mehta RS. Linking cranial morphology to prey capture kinematics in three cleaner wrasses: L abroides dimidiatus, L arabicus quadrilineatus, and T halassoma lutescens. J Morphol. 2015;276(11):1377–91.
Kotob MH, Menanteau-Ledouble S, Kumar G, Abdelzaher M, El-Matbouli M. The impact of co-infections on fish: a review. Vet Res. 2017;47(1):1–12.
Iwanowicz DD. Overview on the effects of parasites on fish health. In: Proceedings of the Third Bilateral Conference between Russia and the United States Bridging America and Russia with Shared Perspectives on Aquatic Animal Health: 12–20 July, 2009; held in Shepherdstown, West Virginia. 2011. pp. 176–84.
Narvaez P, Vaughan DB, Grutter AS, Hutson KS. New perspectives on the role of cleaning symbiosis in the possible transmission of fish diseases. Rev Fish Biol Fish. 2021;31(2):233–51.
Schneider M, Zimmermann AG, Roberts RA, Zhang L, Swanson KV, Wen H, Davis BK, Allen IC, Holl EK, Ye Z. The innate immune sensor NLRC3 attenuates Toll-like receptor signaling via modification of the signaling adaptor TRAF6 and transcription factor NF-κB. Nat Immunol. 2012;13(9):823–31.
Zhang L, Mo J, Swanson KV, Wen H, Petrucelli A, Gregory SM, Zhang Z, Schneider M, Jiang Y, Fitzgerald KA. NLRC3, a member of the NLR family of proteins, is a negative regulator of innate immune signaling induced by the DNA sensor STING. Immunity. 2014;40(3):329–41.
Albertin CB, Simakov O, Mitros T, Wang ZY, Pungor JR, Edsinger-Gonzales E, Brenner S, Ragsdale CW, Rokhsar DS. The octopus genome and the evolution of cephalopod neural and morphological novelties. Nature. 2015;524(7564):220–4.
Riedel G, Platt B, Micheau J. Glutamate receptor function in learning and memory. Behav Brain Res. 2003;140(1-2):1–47.
Storm-Mathisen J, Danbolt N, Ottersen O, Localization of glutamate and its membrane transport proteins. In: CNS neurotransmitters and neuromodulators: glutamate. Boca Raton, FL: CRC Press; 1995. pp. 1–18.
Ferreira G, Gutierrez R, De La Cruz V, Bermúdez-Rattoni F. Differential involvement of cortical muscarinic and NMDA receptors in short-and long-term taste aversion memory. Eur J Neurosci. 2002;16(6):1139–45.
Miranda MI, Ferreira G, Ramírez-Lugo L, Bermúdez-Rattoni F. Glutamatergic activity in the amygdala signals visceral input during taste memory formation. Proc Natl Acad Sci U S A. 2002;99(17):11417–22.
Kind PC, Neumann PE. Plasticity: downstream of glutamate. Trends Neurosci. 2001;24(10):553–5.
Cheng H, Concepcion GT, Feng X, Zhang H, Li H. Haplotype-resolved de novo assembly using phased assembly graphs with hifiasm. Nat Methods. 2021;18(2):170–5.
Laetsch DR, Blaxter ML. BlobTools: Interrogation of genome assemblies. F1000Research. 2017;6(1287):1287.
Guan D, McCarthy SA, Wood J, Howe K, Wang Y, Durbin R. Identifying and removing haplotypic duplication in primary genome assemblies. Bioinformatics. 2020;36(9):2896–8.
Putnam NH, O’Connell BL, Stites JC, Rice BJ, Blanchette M, Calef R, Troll CJ, Fields A, Hartley PD, Sugnet CW. Chromosome-scale shotgun assembly using an in vitro method for long-range linkage. Genome Res. 2016;26(3):342–50.
Hahn C, Bachmann L, Chevreux B. Reconstructing mitochondrial genomes directly from genomic next-generation sequencing reads—a baiting and iterative mapping approach. Nucleic Acids Res. 2013;41(13):e129–e129.
Meng G, Li Y, Yang C, Liu S. MitoZ: a toolkit for animal mitochondrial genome assembly, annotation and visualization. Nucleic Acids Res. 2019;47(11):e63–e63.
Iwasaki W, Fukunaga T, Isagozawa R, Yamada K, Maeda Y, Satoh TP, Sado T, Mabuchi K, Takeshima H, Miya M, et al. MitoFish and MitoAnnotator: a mitochondrial genome database of fish with an accurate and automatic annotation pipeline. Mol Biol Evol. 2013;30(11):2531–40.
Soares MC. The neurobiology of mutualistic behavior: the cleanerfish swims into the spotlight. Front Behav Neurosci. 2017;11:191.
Kim D, Langmead B, Salzberg SL. HISAT: a fast spliced aligner with low memory requirements. Nat Methods. 2015;12(4):357–60.
Liao Y, Smyth GK, Shi W. featureCounts: an efficient general purpose program for assigning sequence reads to genomic features. Bioinformatics. 2014;30(7):923–30.
Love MI, Huber W, Anders S. Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2. Genome Biol. 2014;15(12):1–21.
Flynn JM, Hubley R, Goubert C, Rosen J, Clark AG, Feschotte C, Smit AF. RepeatModeler2 for automated genomic discovery of transposable element families. Proc Natl Acad Sci U S A. 2020;117(17):9451–7.
RepeatMasker (ISB, 2019) http://www.repeatmasker.org/webrepeatmaskerhelp.html.
Hoff KJ, Lange S, Lomsadze A, Borodovsky M, Stanke M. BRAKER1: unsupervised RNA-Seq-based genome annotation with GeneMark-ET and AUGUSTUS. Bioinformatics. 2016;32(5):767–9.
Brůna T, Hoff KJ, Lomsadze A, Stanke M, Borodovsky M. BRAKER2: automatic eukaryotic genome annotation with GeneMark-EP+ and AUGUSTUS supported by a protein database. NAR Genom Bioinform. 2021;3(1):lqaa108.
Brůna T, Lomsadze A, Borodovsky M. GeneMark-EP+: eukaryotic gene prediction with self-training in the space of genes and proteins. NAR Genom Bioinform. 2020;2(2):lqaa026.
Lomsadze A, Ter-Hovhannisyan V, Chernoff YO, Borodovsky M. Gene identification in novel eukaryotic genomes by self-training algorithm. Nucleic Acids Res. 2005;33(20):6494–506.
Buchfink B, Xie C, Huson DH. Fast and sensitive protein alignment using DIAMOND. Nat Methods. 2015;12(1):59–60.
Gotoh O. A space-efficient and accurate method for mapping and aligning cDNA sequences onto genomic sequence. Nucleic Acids Res. 2008;36(8):2630–8.
Iwata H, Gotoh O. Benchmarking spliced alignment programs including Spaln2, an extended version of Spaln that incorporates additional species-specific features. Nucleic Acids Res. 2012;40(20):e161–e161.
Stanke M, Diekhans M, Baertsch R, Haussler D. Using native and syntenically mapped cDNA alignments to improve de novo gene finding. Bioinformatics. 2008;24(5):637–44.
Stanke M, Schöffmann O, Morgenstern B, Waack S. Gene prediction in eukaryotes with a generalized hidden Markov model that uses hints from external sources. BMC Bioinformatics. 2006;7(1):1–11.
Gabriel L, Hoff KJ, Brůna T, Borodovsky M, Stanke M. TSEBRA: transcript selector for BRAKER. BMC Bioinformatics. 2021;22(1):1–12.
Han MV, Thomas GW, Lugo-Martinez J, Hahn MW. Estimating gene gain and loss rates in the presence of error in genome assembly and annotation using CAFE 3. Mol Biol Evol. 2013;30(8):1987–97.
Edgar RC. MUSCLE: multiple sequence alignment with high accuracy and high throughput. Nucleic Acids Res. 2004;32(5):1792–7.
Capella-Gutiérrez S, Silla-Martínez JM, Gabaldón T. trimAl: a tool for automated alignment trimming in large-scale phylogenetic analyses. Bioinformatics. 2009;25(15):1972–3.
Darriba D, Taboada GL, Doallo R, Posada D. jModelTest 2: more models, new heuristics and parallel computing. Nat Methods. 2012;9(8):772–772.
Liu H, Chen C, Lv M, Liu N, Hu Y, Zhang H, Enbody ED, Gao Z, Andersson L, Wang W. A chromosome-level assembly of blunt snout bream (Megalobrama amblycephala) genome reveals an expansion of olfactory receptor genes in freshwater fish. Mol Biol Evol. 2021;38(10):4238–51.
Birney E, Durbin R. Using GeneWise in the Drosophila annotation experiment. Genome Res. 2000;10(4):547–8.
Niimura Y. Identification of olfactory receptor genes from mammalian genome sequences. In: Crasto CJ, editor. Olfactory Receptors: Methods and Protocols. Totowa, NJ: Humana Press; 2013. p. 39–49.
Traynelis SF, Wollmuth LP, McBain CJ, Menniti FS, Vance KM, Ogden KK, Hansen KB, Yuan H, Myers SJ, Dingledine R. Glutamate receptor ion channels: structure, regulation, and function. Pharmacol Rev. 2010;62(3):405–96.
Ramos-Vicente D, Ji J, Gratacos-Batlle E, Gou G, Reig-Viader R, Luis J, Burguera D, Navas-Perez E, Garcia-Fernandez J, Fuentes-Prior P. Metazoan evolution of glutamate receptors reveals unreported phylogenetic groups and divergent lineage-specific events. Elife. 2018;7:e35774.
Emms DM, Kelly S. OrthoFinder: solving fundamental biases in whole genome comparisons dramatically improves orthogroup inference accuracy. Genome Biol. 2015;16(1):157.
Grau-Bové X, Sebé-Pedrós A. Orthology Clusters from Gene Trees with Possvm. Mol Biol Evol. 2021;38(11):5204–8.
Sievers F, Wilm A, Dineen D, Gibson TJ, Karplus K, Li W, Lopez R, McWilliam H, Remmert M, Söding J. Fast, scalable generation of high-quality protein multiple sequence alignments using Clustal Omega. Mol Syst Biol. 2011;7(1):539.
Suyama M, Torrents D, Bork P. PAL2NAL: robust conversion of protein sequence alignments into the corresponding codon alignments. Nucleic Acids Res. 2006;34(suppl_2):W609–12.
Talavera G, Castresana J. Improvement of phylogenies after removing divergent and ambiguously aligned blocks from protein sequence alignments. Syst Biol. 2007;56(4):564–77.
Goodman M, Sterner KN, Islam M, Uddin M, Sherwood CC, Hof PR, Hou Z-C, Lipovich L, Jia H, Grossman LI. Phylogenomic analyses reveal convergent patterns of adaptive evolution in elephant and human ancestries. Proc Natl Acad Sci U S A. 2009;106(49):20824–9.
Lucaci AG, Zehr JD, Enard D, Thornton JW, Kosakovsky Pond SL. Evolutionary Shortcuts via Multinucleotide Substitutions and Their Impact on Natural Selection Analyses. Mol Biol Evol. 2023;40(7):1–20.
Labroides dimidiatus isolate:JK-2023 Genome sequencing and assembly. NCBI BioProject PRJNA937036. https://www.ncbi.nlm.nih.gov/bioproject/PRJNA937036. 2023.
High-throughput molecular characterization of two species' social interaction under normal and climate change conditions. NCBI BioProject PRJNA726349. https://www.ncbi.nlm.nih.gov/bioproject/PRJNA726349. 2021.
We thank the team in the lab of C.S. at HKU for all their support and feedback, and also thank Dr. Kang Du (Texas State University, USA) and Dr. Haifeng Jiang (Northwest A&F University, China) for their help in the gene family detection. The computational analyses were performed using research computing facilities offered by Information Technology Services, the University of Hong Kong
This study was supported by the University of Hong Kong start-up grant to C.S. (C.S., J.L.K.) and by FCT – Fundação para a Ciência e Tecnologia to JRP (2021.01030.CEECIND; PTDC/BIA-BMA/0080/2021 – ChangingMoods; UIDB/04292/2020 & LA/P/0069/2020).
Ethics approval and consent to participate
This project was completed under approval of the Committee on the Use of Live Animals in Teaching and Research (5581–20) of the University of Hong Kong, Faculdade de Ciências da Universidade de Lisboa animal welfare body (ORBEA—Statement 01/2017), and Direção-Geral de Alimentação e Veterinária (DGAV—Permit 2018–05-23–010275) following the requirements imposed by the Directive 2010/63/EU of the European Parliament and of the Council of 22 September 2010 on the protection of animals used for scientific purposes.
Consent for publication
The authors declare that they have no competing interests.
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Metrics for Labroides dimidiatus assembly. Table S2. The length and number of coding genes in Labroides dimidiatus genome. Table S3. BUSCO results (The lineage dataset is: actinopterygii_odb9) using predicted genes in genome of each species as input. Table S4. Percentage of different category of repeat elements in Labroides dimidiatus genome. Table S5. Dynamics of Gene family size of Labroides dimidiatus. Table S6. The number of predicted intact olfactory receptor (OR) in all of 14 species in our study. Table S7. The number of glutamate receptors identified among eight Labridae fishes. Table S8. Positively selected genes (41 genes) of Labroides dimidiatus detected by both PAML and HYPHY. Table S9. Noise Estimated by Simulations—Number of sites of random/false convergence. Table S10. Genes with convergent substitutions in dedicated cleaner (Labroides dimidiatus) and facultative cleaners (Symphodus melops, Semicossyphus pulcher, Tautogolabrus adspersus, Thalassoma bifasciatum, Labrus bergylta). Table S11. Differentially expressed genes (DEGs) between non-interaction (solo) and interaction (inte) groups of Labroides dimidiatus. Table S12. GO enrichment of all 4,004 differentially expressed genes (DEGs) between non-interaction (solo) and interaction (inte) groups of Labroides dimidiatus. Table S13. Genes related to social behavior of differentially expressed genes (DEGs) between non-interaction (solo) and interaction (inte) groups of Labroides dimidiatus. Table S14. Differentially expressed glutamate receptors and glutamate receptor-interacting proteins (from top to bottom in heatmap) between non-interaction (solo) and interaction (inte) groups of Labroides dimidiatus. Table S15. The sequences of 12S, 16S and COI used in the construction of phylogenetic tree to confirm the species of the sampled Labroides dimidiatus individual.
Gene number along the whole genome. Figure S2. The divergence time of the Labridae lineage based on 2,915 single-copy gene families that only have one gene copy for all species. Figure S3. Phylogenetic tree of olfactory receptors (ORs) subfamily ζ gene sequences with 100 bootstraps and the non-ORs as the outgroup. Figure S4. Visual opsin genes among the dedicated cleaner L. dimidiatus, five facultative and two non-cleaners in the fish family Labridae. Figure S5. Phylogenetic tree of NOD-like receptors (NLRs) with 100 bootstraps and rooting at the midpoint. Figure S6. Phylogenetic tree of protocadherin alpha gene sequences with 100 bootstraps and rooting at the midpoint. Figure S7. Phylogenetic tree of protocadherin gamma gene sequences with 100 bootstraps and rooting at the midpoint. Figure S8. Expression of olfactory receptors (ORs) in the three brain regions (forebrain: FB, hindbrain: HB, midbrain: MB) of interacting and non-interacting L. dimidiatus individuals. Figure S9. Expression of opsin genes in the three brain regions (forebrain: FB, hindbrain: HB, midbrain: MB) of interacting and non-interacting L. dimidiatus individuals. Figure S10. Expression of protocadherins α and γ genes in the three brain regions (forebrain: FB, hindbrain: HB, midbrain: MB) of interacting and non-interacting L. dimidiatus individuals. Figure S11. Expression of ten pathogen recognition receptors in the thsree brain regions (forebrain: FB, hindbrain: HB, midbrain: MB) of interacting and non-interacting L. dimidiatus individuals.
About this article
Cite this article
Kang, J., Ramirez-Calero, S., Paula, J.R. et al. Gene losses, parallel evolution and heightened expression confer adaptations to dedicated cleaning behaviour. BMC Biol 21, 180 (2023). https://doi.org/10.1186/s12915-023-01682-3
- Gene family contraction
- Sensory receptors
- Immune system
- Neural signal transduction
- Labroides dimidiatus