Tropical rock lobster (Panulirus ornatus) uses chemoreception via the antennular lateral flagellum to identify conspecific ecdysis

The tropical rock lobster, Panulirus ornatus, is a commercially important aquaculture species exhibiting complex social interactions in laboratory culture, including cannibalism of moulting conspecifics. Cannibalism of soft-shelled post-moult stage individuals is a major limitation during the juvenile stage of culture. Not limited to P. ornatus, cannibalism is widespread across farmed decapods, limiting stocking densities in crab, freshwater crayfish, and prawn species. To understand the mechanisms driving this behaviour and reduce its prevalence, we have investigated the role of chemoreception via the aesthetasc-bearing region of the lateral antennular flagellum, in the recognition of conspecific moulting cues. Differential expression analysis of several tissues in P. ornatus shows an upregulation of 70 ionotropic receptor isoforms, including co-receptors (IR25a and IR93a) and divergent receptors (IR4, IR7, and IR21a) in the aesthetasc-bearing region of the antennules. Deafferentation of the aesthetascs via deionised water exposure prevents juveniles from responding to conspecific moulting cues in a two-current choice flume, suggesting chemoreception, possibly olfaction, plays a role in identifying moulting juveniles. This is the first step in understanding the mechanisms via which cannibalism is triggered in juvenile P. ornatus culture. Further work in this area will help discover means to limit cannibalism in laboratory and commercial culture.


Materials and methods
Transcriptomic analysis. Sample preparation and sequencing. Multiple tissues were previously sampled and used for a transcriptomic assembly, published in Ventura, 2020 32 (available in the NCBI BioProject: PRJNA903480). Raw reads from this project were included in the current study, from brain (n = 2), eye stalk (n = 2), antennal gland (n = 2), epithelium from beneath the dorsal carapace (n = 2) and hepatopancreas (n = 6) tissues. Adult P. ornatus were purchased from Torres Strait wild-caught stocks and reared at the University of the Sunshine Coast. Tissue from the aesthetasc-bearing region of the lateral flagellum of the antennules was sampled from inter-moult adult lobsters (male n = 1, female n = 3), snap-frozen in liquid nitrogen and stored at − 80 °C until further analysis. Frozen tissue was ground with mortar and pestle in liquid nitrogen and total RNA extracted with RNAzol (MRC) and β-Mercaptoethanol. Extracted RNA were assessed using NanoDrop 2000 (ThermoFisher) for yield and purity. For each sample a minimum 10 µl of clean RNA were combined with equal parts RNAstable LD (Sigma-Aldrich), then dried in Concentrator plus (Eppendorf) for 3 h at 60 °C. Desiccated RNA were sent to Novogene for quality control, library preparation, and RNA Sequencing (HiSeq2500, paired end 150 bp). RNA sequence data were submitted to the NCBI Short Read Archive under the BioProject: PRJNA903480 (BioSamples: SAMN32802504, SAMN32802505, SAMN32802506 and SAMN32802507).
Transcriptome assembly and alignment. The newly sequenced antennules, as well as the previously sequenced brain, eye stalk, antennal gland, epithelial tissue and hepatopancreas were used for transcriptomic analysis in this study. Raw paired-end reads for the antennule samples were quality checked using FASTQC 33 and low quality bases (quality score < 20) were removed with Trimmomatic 34 . The trimmed reads were quality checked, concatenated, and de novo assembled alongside the previously sampled tissues with Trinity v.2.9.1 35 as the draft reference genome and transcriptomes currently available for P. ornatus are preliminary and do not include the antennules. A total of 577,782,989 trimmed paired-end reads from all tissues were assembled, with a minimum contig length of 200 bp and a minimum K-mer count of 2. Transcript abundance of the 18 samples was then quantified using the trimmed paired end reads against the multiple tissue reference assembly with RSEM and Bowtie2 36,37 . BUSCO v3 was run on the de novo transcriptome assembly to assess its completeness using the Bioinformatics analyses. The edgeR package in OmicsBox was used to conduct pairwise differential expression analysis comparing antennule tissue samples (n = 4) to non-antennule tissue samples (n = 14) 40,41 . Low counts were filtered to CPM1, normalised with TMM and the significance threshold set to 0.05. Transcripts from this analysis with a logFC > 1 were considered up-regulated and logFC < − 1 considered down-regulated. A Fisher's exact test was conducted within the OmicsBox software to establish statistical significance between a test-set of differentially expressed genes and an annotated reference set. Transcripts classified as differentially expressed were then annotated within OmicsBox software using NCBI-BLAST + 2.10.0 nucleotide database with an Arthropoda taxonomy filter (e-values < 0.01) 42 . The resulting transcripts were further annotated using Gene Ontology (GO) mapping and a Gene Set Enrichment Analysis (GSEA) was performed on TMM normalised gene counts to identify enriched GO terms within the differentially expressed transcripts alongside all annotated features in the antennule transcriptomic assembly in OmicsBox 43 . Results were filtered to meet FDR < 0.05 significance threshold and a heatmap was generated using the OmicsBox software to visualise expression patterns of transcripts of interest which were differentially expressed in antennule tissue compared to non-antennule tissues (brain, eyestalk, hepatopancreas, antennal gland and epithelium tissue). From this list of differentially expressed transcripts, 70 putative IRs were identified and open reading frames were predicted using OrfPredcitor v2.3 44 . A multiple sequence alignment was then performed with ClustalW 45 including previously characterised IR25a sequences from P. argus and Procambarus clarkii, and IR93a and IR21a sequences from H. americanus sourced from NCBI. A maximum likelihood phylogenetic tree was produced with MEGA11 46 with 1000 bootstrap replications using the JTT model. Alignments are provided in Supplementary File S1. The maximum likelihood tree was visualised using iTOL 47 . Olfactory response bioassay. Experimental animals. Panulirus ornatus individuals were reared from eggs at Ornatas commercial aquaculture facility following established procedures [48][49][50] . A total of 18 P. ornatus juveniles (instar J3, 1.63 g ± 0.1, carapace length 11.6 mm ± 0.2) were stocked in three groups of six in 18 L opaque plastic tanks containing PVC tube hides lined with fly mesh for shelter. Water quality was monitored daily and maintained at an average temperature of 27.1 °C ± 0.1, dissolved oxygen 100.9% ± 0.3, salinity 33.9 ppt ± 0.07, and pH 8.2 ± 0.01. Lobsters were fed IMAS commercial-in-confidence feed, rationed at 5% of total tank biomass per 24 h, over six feeds.
Tanks were syphoned twice daily to remove uneaten feed and waste. Housing tanks were kept under a 12:12 L:D photoperiod, with purple (70% red, 30% blue) LED light (1950 lm) positioned 2.5 m above tanks to provide day phase. Lobsters were assigned an ID number and tagged with waterproof paper, glued to the dorsal carapace using gel super glue. At stocking the sex and wet weight (g) of juveniles was recorded, and wet weight was rerecorded the day following ecdysis throughout the experimental period. Pre-moult lobsters were identified up to 18 h prior to ecdysis by the darkening of their ecdysial suture line, on the dorsal and anterior edges of the gill cover ( Supplementary Fig. S1).
Experimental protocol. Design of the two-current choice flume was based on recommendations by Jutfelt, 2017 31 and detailed in Kelly 10 (Fig. 1a,b). Water was supplied to two header tanks (2 L min −1 each) upstream of the flume chamber. Three layers of honeycomb collimators created two laminar currents in the choice arena. Laminar flow was confirmed by adding dye to each header tank and observing no mixing of currents within the choice arena (Fig. 1c). All tank elements were thoroughly cleaned between experimental runs. Removable pieces, such as honeycomb collimators and plastic tubing, were submerged in a chlorine bath (sodium hypochlorite 50 ppm) for 4 h, then rinsed under a flow of saltwater. The flume and header tanks were drained, rinsed with diluted chlorine, and flushed with saltwater. Water quality in the flume tank was monitored before and after all experiments and maintained at an average temperature of 27.2 °C ± 0.1, dissolved oxygen 101.8% ± 0.2 salinity 33.7 ppt ± 0.09, and pH 8.2 ± 0.02.
The experiment followed the procedures of Kelly 10 in which a pre-moult juvenile is placed in a randomly assigned flume header tank and an inter-moult juvenile who is socially naïve to the pre-moult lobster is placed in the choice arena. Time-lapse footage (1 fps) of both the header tank and choice arena is captured overnight to identify the exact time lobsters moult in the header tank. To examine the impact of olfactory deafferentation on P. ornatus behavioural response to conspecific moulting cues, temporary aesthetasc ablation of inter-moult lobsters was achieved by submerging the distal lateral antennule flagellum, featuring the aesthetasc, in distilled water for 30 s. Choice arena lobsters in non-ablation treatments and control replicates had their aesthetascs submerged in saltwater for 30 s. This process was carried out 3 h prior to lobster placement in choice arena. To limit handling and air exposure stress, lobsters were transferred to the header tank and choice arena in jugs of water (2 L) 1 h before recording began to allow for acclimatisation. Three experimental groups were used: ablation treatments (n = 6) and non-ablation treatments (n = 5) featuring an inter-moult lobster in the choice arena and a pre-moult lobster in one header tank, and experimental controls (n = 3) featuring an inter-moult lobster in the choice arena with no lobster in either header tank (Table 1).
Tracking. Time-lapse images were written to AVI format on MATLAB R2020b, with a frame rate of 6 fps. The ImageJ plugin, AnimalTracker, was used to track the activity of lobsters within the two laminar currents, which were designated as regions of interest using AnimalTracker Zone Designer 51,52 . Each frame was filtered using www.nature.com/scientificreports/ www.nature.com/scientificreports/ Animal Tracker's inbuilt background subtraction and thresholding, then postprocessed to remove excess noise. A single reference point, the centroid of the detected lobster, was followed by AnimalTracker without concern for body orientation. Tracks were manually edited if tracker lost the reference point on the body for several frames. A baseline observation hour was analysed from 1 h after lobsters were placed in the flume. Movement of choice arena lobsters in treatment replicates was tracked for 1 h before and 1 h after the lobster in the header tank moulted. Lobsters in the choice arena during control replicates were tracked for 1 h before and 1 h after the mean moult time observed in treatment replicates, approximately 3.5 h after commencement of dark phase.
Statistical analyses. Data were assessed using non-parametric tests in R 53 due to restricted sample size. Water current preference of lobsters in the choice arena was assessed with a Mann-Whitney U test comparing mean time spent in the conspecific moulting cue during the 10 min after an upstream moult for ablation and non-ablation treatments. Time spent in a single current was also assessed for the baseline hour and 1 h before upstream moult using a Kruskal-Wallis test for ablation and non-ablation treatments and control replicates.

Results
Transcriptomic analysis. De  Differential expression analysis. From 120,812 filtered features, 3205 were significantly differentially expressed (FDR < 0.05). This included, 2481 upregulated transcripts (logFC > 1), and 724 downregulated transcripts (logFC < − 1) in the aesthetasc-bearing antennule tissue compared to other tissues. NCBI nucleotide BLAST found successful hits for 1450 sequences (Supplementary File S4) and 70 ionotropic receptor transcripts were identified with upregulated expression in antennule tissue (Fig. 2). Annotation via the NCBI-BLAST + 2.10.0 nucleotide database with an Arthropoda taxonomy filter found 14 transcripts shared high identity with putative genes, while the remaining 56 share high identify with predicted homologues of ionotropic glutamate receptors. We have identified homologues for kainate-like receptors, delta-like receptors, two co-IRs (IR25a and IR93a), several divergent IRs (IR4, IR7 and IR21a) and one NMDA receptor (Fig. 3). IR25a was consistently upregulated in the aesthetasc-bearing region of the antennule and downregulated in all other tissues (Fig. 2). IR25a has been well described as a co-receptor expressed in P. argus olfactory receptor neurons (ORNs), and the sequence determined here in P. ornatus shares high identity with P. argus (92.49% ident., e-value 0.0, Supplementary File S4). Single cell transcriptome research has previously identified co-IRs, IR25a and IR93a in P. argus ORNs 54 . Our analysis similarly revealed an upregulation of these two co-IRs in the P. ornatus antennule tissue but failed to observe increased expression of the non-olfactory antennular co-IRs, IR8a and IR76b. Forty-three Biological Process GO IDs were observed, with five over-represented (electron transport chain, cellular respiration, energy derivation by oxidation of organic compounds, Golgi vesicle transport, aerobic respiration), and 38 under-represented with several related to negative regulation of metabolic processes (Supplementary Table S1). Nineteen Cellular Component GO IDs were seen with four over-represented (Golgi apparatus, organelle sub compartment, respirasome, Golgi apparatus sub compartment) and 15 under-represented (Supplementary Table S2). Additionally, 16 Molecular Function GO IDs were enriched, seven over-represented (three involving receptor activity, three with gated ion channel activity and one electron transfer activity) and nine under-represented (Supplementary Table S1).
Three GO terms that were significantly over-represented and are of interest for this study are neurotransmitter receptor activity (GO:0030594), ionotropic glutamate receptor activity (GO:0004970) and glutamate receptor activity (GO:0008066). A set of 12 transcripts contribute directly to the enrichment of these GO terms in aesthetasc-bearing antennule tissue ( Table 2). Several of these sequences have high identity to two genes characterised in P. argus, IR4 (KC603903.1) and IR7 (KC603904.1). Others have high sequence identity with various ionotropic glutamate receptors.  (Table 3, Kruskal-Wallis test, H(2) = 0.8, P = 0.7). Water current preference during the 1 h before upstream lobsters moult (1 h before mean moult time for control replicates) was statistically similar between treatments and controls (Kruskal-Wallis test, H(2) = 2.9, P = 0.2), although the average time of non-ablated treatments is increased compared to the baseline (Table 3). Lobsters with non-ablated aesthetascs spend a greater amount of time in the moulting cue current than lobsters with ablated aesthetascs during the 10 min after a lobster moults upstream (Table 3, Mann-Whitney U test, U = 4, P = 0.04). This difference indicates the aesthetasc facilitates a preference response to conspecific moulting cues, which is visualised by a density plot mapping the coordinate points of lobsters within the choice arena during the 10 min moult cue exposure period (Fig. 4).

Discussion
Investigation of the P. ornatus transcriptome has revealed significant upregulation of 70 ionotropic receptor transcripts within the aesthetasc-bearing region of the lateral antennular flagellum. Additionally, deafferentation of chemoreceptors in this region of the antennules prevents P. ornatus juveniles from responding to conspecific moulting cues. This demonstrates that chemoreception via the distal region of the antennular lateral flagellum is required to mediate a behavioural response to conspecific moulting cues. The aesthetasc setae in this region of the antennules are responsible for olfaction in spiny lobsters 9,13,27 , therefore we propose olfaction plays at least a partial role in recognition of conspecific moult cues. This finding is critical for research aimed to minimise  Within the chemical rich environment of marine crustaceans, chemical communication is vital in facilitating many social behaviours, such as aggregation 28 , identifying familiar and naïve conspecifics 55 , differentiating between dominant and subordinate individuals 56,57 , as well as between inter-moult and moulting lobsters 10,58 . Spiny lobsters feature sensory sensilla over much of their body, particularly the antennules, mouthparts, and dactyls, innervated with both chemoreceptor and mechanoreceptor neurons, however, olfactory receptor neurons are only present in aesthetasc chemosensilla 9,13,27 . Chemical sensory pathways are responsible for receiving several forms of non-contact conspecific communication, such as courtship pheromones 12 , blood-borne injury cues 29 , and urine-borne cues responsible for mediating conspecific aggregation and aggressive behaviours 28,56 . Aiming to understand the mechanisms driving the detection, attack, and cannibalism of moulting and postmoult juveniles, we have examined the role of chemoreception via the lateral antennular flagellum, including the olfactory aesthetasc, in recognising conspecific moulting cues without physical or visual contact.
We can conclude from differential expression analysis and enrichment analysis that IR expression is enriched within the aesthetasc-bearing tuft of the antennules. Enrichment analysis found GO terms of key interest, neurotransmitter receptor activity, ionotropic receptor activity, and glutamate receptor activity, to be over-represented in the antennules. The genes associated with these GO terms have moderate to high sequence identity with several ionotropic receptors (IR) and ionotropic glutamate receptors (iGluR) in the Caribbean spiny lobster (P. argus), giant tiger prawn (Penaeus monodon) and the Pacific white shrimp (Litopenaeus vannamei). Phylogenetic conservation has been observed for various IRs with IR25a conserved in all protostomes studied to date 17,18 . The similarity in sequence, and more significantly, the phylogenetic similarity seen between the predicted co-IRs, 25a and 93a-like homologues, identified here in P. ornatus and confirmed homologues in P. argus and H. americanus supports our identification. Four co-IRs have been described in crustaceans, IR25a, IR93a, IR8a and IR76b. Single cell ORN transcriptomes produced by Kozma 54 , found P. argus ORNs only express two of the four co-IRs, IR25a and IR93a, the same two identified in the current study. IR25a and IR93a are obligate co-receptors expressed in essentially all ORNs 20 , and IR25a is also expressed in most CRNs of P. argus antennules, antennae, and dactyls 16,19 . The combination of obligate co-IRs with phyla conserved divergent IRs enables binding specificity for the receptor 17,19,20 . Divergent IRs, IR4, IR7 and IR21a-like, present upregulated expression in P. ornatus aesthetasc-bearing antennule tissue. Several transcripts for IR4 and IR7 contribute to the over-representation of key GO terms in the aesthetasc-bearing antennule tissue studied here.
In P. argus, both aesthetasc and non-aesthetasc chemoreceptor sensilla mediate lobster detection of, identification of, and orientation towards food sources 59 , however temporary deafferentation of aesthetasc chemoreceptor neurons does somewhat decrease success in finding food 25 . The use of dual antennular chemosensory pathways to mediating odorant activation for locating food in spiny lobsters enables a functional redundancy; if one type  www.nature.com/scientificreports/ of chemoreceptor is inhibited, the other still fills the role. Our expression analysis demonstrates the significant presence of IRs in the lateral flagellum of the antennules of P. ornatus, however, to functionally annotate these receptors in the ORNs requires a cellular level loss-of-function investigation. This would be required to determine if aesthetasc sensilla alone detect conspecific moulting cues, and if so, present a new line of investigation to limit cannibalism in P. ornatus culture. Transcriptomic analysis of single ORN cells from P. ornatus aesthetascs, as previously done by Kozma 54 for P. argus, would differentiate between IRs expressed in ORNs and CRNs in this region of the antennules. Functional assessment of IRs in chemosensory behaviour has been achieved through RNA interference (RNAi) knock-out of IR25a and IR93a in several Daphnia species 60 . In the case of P. ornatus, reverse genetic studies with RNAi have proven to be ineffective, due in part to animal-wide low expression of the enzymes and receptors required for functional and systemic gene silencing 61 . Despite multiple attempts to generate clear phenotypes with RNAi in P. ornatus and other Palinurid lobsters, no significant silencing has ever been observed which limits the functional annotation of ORNs in spiny lobsters. Additionally, the development and functionality of chemosensory systems in crustaceans with drastically altered morphology across life-stages is understudied 12 , making it difficult to select appropriate targets for RNAi and develop assays to test for altered olfactory capacity. Continued research contributing to the transcriptomic database for P. ornatus across multiple life-stages may reveal a difference in IR expression patterns in the antennules, which may then inform on optimal genes to silence with RNAi when the technology becomes more accessible in P. ornatus.
With currently available knowledge we assessed the function of chemoreception via the aesthetasc-bearing region of the lateral antennular flagellum, in mediating a behavioural response to conspecific moulting cues. A two-current choice flume was used in a previous study with juvenile P. ornatus to identify preference for or against conspecific moulting cues, with a focus on the moult stage of the responding lobster and its relationship to the moulting lobster. This research found inter-moult lobsters are attracted to moulting cues from lobsters whom they are naïve to 10 . This bioassay was applied to the current study, pairing inter-moult lobsters with naïve, pre-moult lobsters, and tracking the time inter-moult lobsters spent in the conspecific cue current for 10 min following upstream ecdysis. Lobsters whose antennules were exposed to seawater only (non-ablated), responded as expected based on the aforementioned bioassay, spending an average 91.0 ± 3% of the 10 min in the moulting cue current. Alternatively, lobsters whose antennules were temporarily ablated with deionised water exposure spent an average 64.6 ± 6% of the 10 min in the moulting cue current, statistically similar to the baseline results for all treatment and control replicates. This indicates deafferentation of chemoreceptor neurons, including ORNs, in the aesthetasc-bearing region of the lateral antennular flagellum removes the ability to perceive a moulting conspecific via chemical cues alone. Flume assays have been highly effective in identifying preference behaviour in crustaceans, in response to various conspecific odours and feed attractants 58,62,63 . The current study is applicable to further aquaculture research and development, but also has great potential for fisheries and ecology as we need to understand how aquatic animals interact with each other and their rapidly changing habitats.
Cannibalism is one of the most prevalent limiting factors in the culture of several commercially significant species 1,6 , including P. ornatus 5 . Lobsters with temporarily ablated aesthetascs do not display preference for or against moulting cues, indicating they are incapable of responding to such cues without functional aesthetascs. The knowledge gained here indicates that antennule chemoreception, and possibly olfaction, are a link in cannibal-prey recognition. The identity of the chemical cue being received by cannibals remains a missing link in this research. Ecdysis is regulated by a multitude of hormones responsible for separating the external cuticle from the epidermis, and simultaneously forming a new cuticle. Numerous studies have indicated chemical cues released by moulting crustaceans influence conspecific behaviours associated with survival, aggression, and cannibalism 11,64 . An investigation into blood-borne and urine-borne cues, targeting molecules in high abundance at ecdysis, may elucidate the type of molecules responsible for indicating the biological state of lobsters at ecdysis to others.

Conclusion
Here we endeavoured to uncover the chemosensory pathways responsible for receiving conspecific moulting cues and mediating P. ornatus response using transcriptomic analysis coupled with a functional bioassay. We successfully implemented the bioassay with functional ablation of the aesthetasc-bearing region of the lateral antennular flagellum, demonstrating the role of chemoreception in this region in detecting conspecific ecdysis. The accompanying transcriptomic analysis of this antennular region revealed upregulation of 70 ionotropic receptor transcripts, adding to the growing genomic and transcriptomic knowledge for this key aquaculture species. This study indicates that chemoreception via the aesthetasc-bearing region of the lateral antennular flagellum facilitates recognition of moulting conspecifics in the tropical rock lobster, P. ornatus. Furthermore, we propose this cue recognition plays a role in mediating cannibalism of post-moult juveniles in communal culture. Continued research in this area may aim to identify chemicals released into the environment at ecdysis and received by potential cannibals. As commercial culture for this species continues to develop, limiting cannibalism in communally cultured juveniles is essential to sustaining productivity.

Data availability
RNA sequence data for this study has been submitted to the NCBI Sequence Read Archive (http:// www. ncbi. nlm. nih. gov/ sra under BioProject: PRJNA903480). All other data presented in this study is available on request from the corresponding author.