Introduction

Crown-of-thorns starfish, Acanthaster spp. (hereafter CoTS), are corallivorous asteroids common across coral reefs of the Indo-Pacific. At low densities, CoTS play an important role in coral reef functioning and can enhance local coral diversity (Birkeland 1989a; Keesing 1990; Uthicke et al. 2009; Pratchett et al. 2014). However, CoTS populations are prone to population fluctuations resulting in periodic outbreaks. Localised adult CoTS densities can increase by several orders of magnitude within 1–2 yrs (Birkeland and Lucas 1990). Rampant CoTS population proliferation is often followed by rapid and extensive coral loss (Pratchett et al. 2014). As a result, CoTS outbreaks are considered a major contributor to persistent declines in coral cover across the Indo-Pacific (Osborne et al. 2011; Trapon et al. 2011; De'ath et al. 2012; Vercelloni et al. 2017; Bozec et al. 2021).

Several hypotheses have been proposed to explain the proximal causes of CoTS outbreaks, but they are widely debated and remain largely unresolved (Moran 1986; Birkeland and Lucas 1990; Pratchett et al. 2017). Outbreaks have been suggested as a natural consequence of exceptional but variable reproductive output (the ‘natural causes’ hypothesis) (Dana et al. 1972; Vine 1973; Babcock et al. 2016; Caballes et al. 2021). Concurrently, regional outbreaks may be facilitated by location-specific patterns in dispersal (connectivity) caused by inherent hydrodynamic and geographical properties (Hock et al. 2014). Alternatively, CoTS outbreaks may be caused (or exacerbated) by anthropogenic degradation of coral reef ecosystems, due to eutrophication and/or overfishing and modification of trophic networks (Pratchett et al. 2017, 2021).

Many hypothesised causes of CoTS outbreaks focus on factors affecting the persistence of early life-history stages, as processes that dictate larval and recruit success can drive adult abundances (Keesing and Halford 1992b; Gosselin and Qian 1997; Wilmes et al. 2018). Significant research effort has focused on the potential for variations in larval nutrition (the ‘nutrient enrichment’ hypothesis) to contribute to periodic CoTS population booms (Pearson and Endean 1969; Lucas 1973). Increased phytoplankton density as a result of high nutrient loads from agricultural land use may release CoTS larvae from food limitation in otherwise oligotrophic tropical waters (Brodie et al. 2005; Fabricius et al. 2010). As possible coping mechanisms to thrive in water with low particulate food, CoTS larvae have also demonstrated resilience to a range of nutrient-poor conditions (Wolfe et al. 2015a, 2017), through developmental phenotypic plasticity (Wolfe et al. 2015b) and associations with microbial phototrophs (Carrier et al. 2018). Similarly, persistence and growth plasticity of CoTS when food-limited have also been demonstrated in the juvenile stage (Deaker et al. 2020), suggesting that early life-history stages of CoTS may be inherently robust to fluctuations in environmental conditions.

The ‘predator-removal’ hypothesis postulates that a reduction in the number of CoTS predators through overfishing enhances CoTS survival and outbreak potential (Endean 1969; McCallum 1987, 1990). These claims followed the first documented outbreak on the Great Barrier Reef (GBR) in Australia, coinciding with exhaustive fishing-induced depletion of the giant triton, Charonia tritonis, a predator of adult CoTS (Endean 1973). However, the effectiveness of predation by C. tritonis on CoTS remains unresolved (Ormond et al. 1990; Motti et al. 2022), given that reported consumption rates are low, < 1 wk−1 (Endean 1969; Birkeland 1989b), and their historical (cf. pre-harvest) densities are unknown (Hall et al. 2017). The ‘predator-removal’ hypothesis has since expanded to consider the impact of large reef fishes, including many targeted by commercial fisheries (Ormond et al. 1990; Sweatman 1997; Kroon et al. 2020). Multiple studies have suggested that reefs exploited by fisheries, and therefore potentially lacking CoTS predators, may experience more severe and more frequent CoTS outbreaks relative to protected areas (Dulvy et al. 2004; Sweatman 2008; McCook et al. 2010; Mellin et al. 2016; Vanhatalo et al. 2017; Westcott et al. 2020; Kroon et al. 2021). This is supported by the increase in prevalence of injured and damaged CoTS within protected reef zones (Caballes et al. 2022). Despite the increasing number of fish species identified as potential CoTS predators (Cowan et al. 2017a; Kroon et al. 2020), mechanistic understanding of how fisheries-exploited species directly interact with CoTS is still lacking (Pratchett et al. 2021).

Until recently, research on the potential for predator regulation of CoTS has primarily focused on the adult life stage (Endean 1969; Pearson and Endean 1969; McCallum et al. 1989; Hall et al. 2017; Kroon et al. 2020, 2021). However, early life-history stages, including small planktonic larvae and benthic juveniles, are likely to be highly susceptible to predation relative to adults and represent a typical population bottleneck of most marine broadcast spawning invertebrates (Gosselin and Qian 1997). CoTS gametes and larvae may experience significant predation pressure from planktivorous fishes (Cowan et al. 2016a, 2017b, 2020), and predation-induced mortality rates of newly settled CoTS are suggested to be high (Keesing and Halford 1992a). Indeed, variation in outbreak intensity with fishing protection may be occurring through multiple and potentially indirect interactions with early life-history stages rather than adults (Sweatman 2008; Kroon et al. 2020). Importantly, CoTS juveniles are exposed to predators for much longer (i.e. months to years) in comparison with gametes and larvae (i.e. days to weeks) (Deaker et al. 2020; Wilmes et al. 2020a). Consequently, even small variations in predator-induced mortality of CoTS during this juvenile life stage can accumulate to substantially change the likelihood of outbreaks (Keesing and Halford 1992b; Morello et al. 2014; Keesing et al. 2018; Wilmes et al. 2018). Predation and variations in survival during the early life-history stages of the CoTS population cycle remain critical knowledge gaps but have the potential to significantly inform understanding of outbreak initiation mechanisms.

Coral rubble is the primary settlement habitat of CoTS (Zann et al. 1987; Wilmes et al. 2020b) and is home to a wide range of crustaceans, molluscs, echinoderms, fishes and worms, which occupy all trophic guilds (Glynn and Enochs 2011; Cortés et al. 2017). Trophodynamics within the rubble cryptofauna are largely unresolved, yet a variety of cryptic predators are common (reviewed in Wolfe et al. 2021). Rubble-dwelling predators play potentially important roles in the structuring of coral reef communities from the bottom-up, particularly through predation on vulnerable early life-history stages of organisms that recruit to rubble (Glynn 2006, 2013). Empirical data on relevant predation rates of these communities are sparse, making it difficult to quantify the magnitude of their role in population and recruitment bottlenecks, and thus, greater reef food webs.

Identification of CoTS predators common in rubble has thus far occurred opportunistically and less often on the juvenile stage (reviewed in Cowan et al. 2017a). Invertebrate communities in rubble habitats may contribute significantly to juvenile CoTS mortality (Keesing and Halford 1992a; Keesing et al. 1996, 2018), but there is poor understanding of the identity of specific predator species. Notably, the fireworm Pherecardia striata and harlequin shrimp Hymenocera picta have been observed feeding on juvenile CoTS across reefs of the Eastern Pacific (Glynn 1984). In Panama, where these predators are found in high abundance, CoTS are relatively scarce and have not exhibited outbreaks (Glynn 1982). In a laboratory context, the peppermint shrimp Lysmata vitatta has demonstrated capacity to consume CoTS juveniles (Balu et al. 2021), and the presence of polychaete worms and trapeziid crabs limited CoTS settlement and metamorphosis (Cowan et al. 2016b). In all, these initial findings provide strong evidence for the importance of invertebrate predation, yet a comprehensive evaluation of juvenile CoTS predators during this rubble-dwelling life stage has never been conducted (Wilmes et al. 2018).

Here, we assessed whether common rubble-dwelling taxa consume juvenile CoTS in a series of feeding trials. First, predators collected from coral rubble were offered one juvenile to determine whether they could indeed consume CoTS. Second, for two of the most likely predators in these initial trials, we examined their capacity to detect and consume CoTS in natural rubble mesocosms. From these experiments, we identify key novel predators of CoTS juveniles and discuss the implications of juvenile predation in context of CoTS population outbreaks on coral reefs.

Methods

Juvenile CoTS rearing

The taxonomy of the Pacific species of Acanthaster is uncertain (Haszprunar and Spies 2014), so we refer to the species on the GBR as Acanthaster sp. Adult CoTS were collected in the Cairns Region, Queensland, Australia and shipped live to the National Marine Science Centre in Coffs Harbour, New South Wales, Australia, where they were maintained in flow-through seawater at the approximate temperature of the collection habitat (25–27 °C). CoTS were spawned in December 2021 by dissecting the gonads from multiple males and females. Ovaries were rinsed in 1 μm filtered seawater (FSW) and steeped in 10–5 M 1-methyl adenine in FSW to induce ovulation. Sperm was collected directly from gonads using a pipette. Eggs and sperm that had been checked for quality microscopically were combined for fertilisation in a 1-L beaker at a sperm-to-egg ratio of approximately 100:1. Once greater than 95% fertilisation was confirmed, the sperm were rinsed away using FSW.

Larvae were reared in a 300-L culture container in FSW at 25–27 °C that was changed daily. The larvae were fed the cryptomonad algae Proteomonas sulcata at a density of approximately 1–5 × 104 cells mL−1 as needed. After 18–22 d, competent larvae were settled onto polycarbonate plates containing a culture of crustose coralline algae (CCA) and mixed algal biofilm. Juveniles were reared on these plates in flow-through seawater at approximately 25–27 °C. In February 2022, CoTS and CCA were transported to Heron Island Research Station where they were housed in two 6-L flow-through aquaria (mean 27.3 °C ± 0.1, n = 60) throughout the experiment.

Predator candidate collection

Potential CoTS predators were collected on SCUBA across coral rubble patches of Heron Reef, Australia. Collections were conducted between 2 and 12 m depth, as this covers the depth range wherein CoTS settlement and juvenile distributions have been documented in situ (Wilmes et al. 2020b; Doll et al. 2021). Rubble communities were sampled using multiple collection methods to obtain adequate replication and ensure representation of a variety of taxa. First, RUbble Biodiversity Samplers (RUBS), which are 3D-printed models used to standardise biodiversity sampling in rubble, were regularly deployed in rubble patches using standard protocol (Wolfe and Mumby 2020). Additionally, a series of plastic mesh baskets (4 L) were filled with rubble and buried in depressions at these sites. Both RUBS and rubble-filled baskets were collected after a minimum of 4 d and redeployed periodically to sustain rubble community collections. Upon retrieval, RUBS and rubble-filled baskets were lifted from their depressions in the rubble and sealed immediately in individual plastic bags underwater. Each sample was then returned to the laboratory and extensively searched for predator candidates by using probes, forceps and pressurised seawater delivered from wash bottles to dislodge individuals. In tandem with these passive collection methods, active searches for larger mobile taxa were conducted. This involved searches among rubble patches, manually overturning rubble pieces and collecting conspicuous individuals using small hand nets or by hand. Based on field collections, density of confirmed CoTS predators was approximated as either individuals per dive hour for manually collected species or individuals per L for species collected passively (see Table 1). We note that manual and passive collections were concentrated in rubble sites with known, and often specific, predator communities so the density estimates here may not reflect total abundance of these species across broader habitats and reef zones.

Table 1 Description of confirmed juvenile CoTS predators (n = 26), including taxonomical classification, body size and incidence of consumption and injury across feeding trials

The potential for collected individuals to be predatory was determined from extensive literature searches and reports of diet for each species. Specifically, species reported as herbivores in the literature or those with body sizes much smaller than the minimum CoTS size (0.8 mm) were not used in experiments, while those considered scavengers, generalists or predators were used. Selected species were housed in communal 6-L flow-through aquaria with natural rubble supplied with seawater to provide food and shelter until use in predation trials. Larger individuals were separated to ensure predation among predator candidates did not occur within housing tanks and were fed chopped bait prawns every few days. Individuals were starved for ~ 24 h before use in experiments.

Predation trials

To determine whether a predator candidate collected from rubble ate CoTS, single specimens were placed with an individual juvenile CoTS in initial feeding trials. Predator candidates were photographed (Olympus TG6) for identification and measured (nearest mm) before being placed in individual 800-mL containers with flow-through seawater (~ 0.3 L min−1). No shelter was provided. One juvenile CoTS was then randomly selected from the housing tank, photographed under a dissecting microscope (Olympus SZ61, Dino-Eye AM7025X) and placed into each predator tank. Juvenile size (maximum diameter, mm) was quantified using ImageJ and ranged from 0.8 to 3.8 mm in diameter across predation trials (mean 1.67 ± 0.02).

Feeding trials lasted for a maximum of 3 d with tanks checked once per day. When the juvenile CoTS was not readily visible, the tank was thoroughly searched before all water contents were filtered through a 200-µm mesh and further rinsed with freshwater. If the juvenile was still absent, the predator candidate was examined and rinsed with freshwater to ensure the CoTS juvenile was not on the predator itself, which occurred in several instances. CoTS were deemed consumed when not found after this extensive search process was repeated several times. At the end of the trial period, CoTS were scored as either not consumed, injured (partial predation) or consumed. CoTS that were found were scored as injured if they had visible portions of their body missing following the trial (see Fig. 1L, M).

Fig. 1
figure 1

Phylogenetic relatedness of 104 CoTS predator candidates—blue labels denote confirmed predator taxa. Featured predator candidates A the fireworm Eurythoe complanata, B Peristernia reincarnata, C the marbled shrimp Saron marmoratus, D the red decorator crab Schizophrys aspera, E Thalamita admete, F Thalamitoides tridens, G Chlorodiella nigra, H Actaeodes hirsutissimus, I Etisus anaglyptus and J Neoliomera insularis. K Shows healthy, intact juvenile CoTS, while L and M each show 1 juvenile CoTS damaged by predators with arrow indicating the site of injury. Scale bars in panels A–J equate to 10 mm, and 2 mm in panels K–M

Rubble mesocosms

Several species emerged as successful predators of CoTS in the initial experimental feeding trials. To address the potential of these predators to detect and consume juvenile CoTS among rubble, we conducted a series of feeding trials in tanks filled with natural coral rubble. Two decapods, Schizophrys aspera (Majidae) and Thalamita admete (Portunidae), were used given their consistency in initial feeding trials (see Results section) and their high abundance in our rubble sites (see Table 1).

To establish the rubble mesocosms, 1-L plastic mesh baskets were filled with rubble in situ and buried in rubble patches at ~ 4 m for at least 4 d to colonise a natural rubble community (rubble water displacement volume = 288 mL ± 10, n = 18). Baskets were collected on SCUBA and sealed in individual plastic bags for transportation back to the laboratory, as above. The contents of each rubble-filled basket were then placed in 6-L flow-through tanks (supplied with 1.15 L ± 0.06 seawater min−1) and left to stabilise for at least 6 h before trials commenced. Each tank was stocked with 30 juvenile CoTS (1.02–3.78 mm) and one predator, either S. aspera or T. admete (n = 6 per predator). Control tanks without predators were also conducted to capture any background CoTS mortality caused by the wider rubble community and to account for potential error in searching and handling procedures (n = 6). Rubble pieces in all treatments were covered with a range of sessile biota including CCA, turf algae, macroalgae, sponges and ascidians, as typical of shallow reef rubble (Wolfe et al. 2021). This microhabitat complexity served to amplify the cryptic nature of juvenile CoTS, which fed on CCA throughout the experiment, as evidenced by feeding scars on rubble pieces.

After 4 d, tanks were searched extensively for CoTS. Each rubble piece was removed and rinsed with freshwater at least 3 times to remove all visible CoTS and other rubble fauna, a common method for extracting rubble fauna (Takada et al. 2007; Stella et al. 2011; Wolfe et al. 2021). Given the complex morphology of some rubble pieces, it was necessary to break certain pieces apart to investigate crevices and holes within which CoTS may have been lodged. All CoTS were retained in a separate dish for each treatment, and all other individuals > 1 mm (the approximate size of CoTS used) from each rubble community were counted and classified to phylum under a dissecting microscope to indicate alternative prey and potential species’ interactions within mesocosms. All CoTS were photographed and measured, as above.

Data analysis

A phylogenetic tree of all predator candidates was constructed at the highest possible taxonomic resolution using the open-access interactive Tree Of Life (iTOL) tool (Letunic and Bork 2021). Fisher’s exact test was used to compare the incidence of feeding trial outcomes (relative proportion of not consumed, injured or consumed) among predator candidates with at least one observed instance of CoTS consumption or injury using the fisher.test function from the stats package in R version 4.0.0 (R Core Team 2020). Fisher’s exact test was selected to address small sample sizes, and the p value was computed using Monte Carlo simulation, given that the contingency table dimensions were larger than 2 × 2 (Patefield 1981). Pairwise comparisons of feeding trial results between species were extracted using the fisher.multcomp function in the RVAideMemoire package (Hervé 2022), and resulting p-values adjusted using Benjamini & Hochberg’s methods (Benjamini and Hochberg 1995). Inspection of pairwise comparisons was used to aggregate similar species into groups based on the relative frequencies of each outcome (I = incidental, C = consumers or P = partial consumers).

The proportions of CoTS that experienced predation (i.e. injured or consumed) in rubble mesocosm trials were compared across treatments using binomial regression implemented with the lme4 package (Bates et al. 2015). Here, we grouped the number of consumed and injured CoTS as partial predation was only observed in T. admete treatments, and both outcomes are ultimately representative of predatory interactions. Post hoc pairwise comparisons between treatments were conducted using Tukey’s HSD from the multcomp package (Hothorn et al. 2008).

Results

Predation trials

A total of 104 distinct taxa from 41 families and 74 genera (Fig. 1, Table S1) were collected and used across feeding trials (n = 404). Where possible, taxa were identified to species level, but 59 individuals were only identified to genus (Table S1). Feeding trials were replicated 1–19 times per species (mean = 3.9 ± 0.4 SE), with low replication for some owing to rarity of collection (Table 1, Table S1). Of the taxa tested, 26 species were found to have consumed or injured CoTS on at least one occasion (Fig. 1). Confirmed predators were overwhelmingly decapod crustaceans (85%), with greatest representation of species from the Portunidae and Xanthidae (Fig. 1A–J, Table 1). Two species of Annelida and of Mollusca were also found to consume juvenile CoTS. The remaining predator candidates (n = 78) that did not display capacity to consume CoTS across feeding trials (Table S1) were comprised of species from the Platyhelminthes (n = 7), Nemertea (n = 3), Annelida (n = 6), Mollusca (n = 12), Arthropoda (n = 43) and Chordata (n = 7).

Significant differences were found between relative proportions of feeding trial outcomes for confirmed predators (p < 0.001). Visual inspection of outcomes and pairwise comparisons revealed three distinct groupings. One class of predators, which we refer to as incidental, only partially or entirely consumed CoTS juveniles on 1 or 2 occasions (Table 1, Fig. 2). Predators in this class were comprised of worms, Eurythoe complanata (Fig. 1A) and Lepidonotus cristatus, gastropods, Latirus polygonus and Peristernia reincarnata (Fig. 1B), and several decapods, including species of the Alpheidae, Calcinidae and Hippolytidae (e.g. Saron marmoratus, Fig. 1C), as well as Xanthidae (e.g. Chlorodiella nigra, Actaeodes hirsutissimus and Neoliomera insularis, Fig. 1G, H and J, respectively). The incidental interactions of these predators stand in contrast to outcomes for Schizophrys aspera (Fig. 1D), a decorator crab that consumed CoTS in 89% of feeding trials (Fig. 2). We categorise this predator as an active consumer. A third group of predators, we refer to as partial consumers, less frequently consumed entire juvenile CoTS and most often inflicted arm and body damage of varying severity (Fig. 1K–M, Fig. 2). The partial predator archetype was most often associated with species of Portunidae (e.g. Thalamita admete and Thalamitoides tridens, Fig. 1E, F), as well as two xanthid crabs, Cyclodius ungulatus and Etisus anaglyptus (Fig. 1I, Table 1, Fig. 2).

Fig. 2
figure 2

Incidence of outcomes for feeding assays of confirmed juvenile CoTS predators (n = 26). Groups denote species classified as (I) incidental, (C) consumer, or (P) partial predators

Rubble mesocosms

Of the 30 juvenile CoTS added to control tanks (i.e. no added predator), we recovered > 98% (29–30 ind.). Both Schizophrys aspera (p < 0.001) and Thalamita admete (p < 0.05) consumed significantly more CoTS compared to the control, and these two predator treatments also differed (p < 0.001) (Fig. 3A, Table S2). When T. admete was present, we found ~ 93% CoTS intact, suggesting that 2 ± 0.5 CoTS were preyed on (Fig. 3A). Of these, one-third were injured rather than fully consumed. In S. aspera mesocosms, we recovered 21% of CoTS (i.e. 23.7 ± 1.6 ind. consumed). Taking background loss or mortality rates from control tanks into consideration (0.13 ± 0.06 CoTS d−1), T. admete and S. aspera results translate to daily consumption rates of 0.37 ± 0.14 and 5.78 ± 0.41 CoTS d−1, respectively. CoTS size did not vary substantially among treatments (mean 1.94–2.15 mm).

Fig. 3
figure 3

A proportion of CoTS (± SE) preyed on in experimental rubble mesocosms and B mean abundance (± SE) of rubble-dwelling taxa (> 1 mm) after 96-h exposure to either control conditions, a Thalamita admete or Schizophrys aspera individual (n = 6 per treatment). Significant differences between treatments are indicated as *p < 0.05 and ***p < 0.001

Post-trial rubble communities across all mesocosms treatments consisted primarily of individuals from the Arthropoda and Annelida (Fig. 3B). Fewer molluscs were found in control tanks, while lower total numbers of rubble fauna were documented following feeding trials with T. admete. The greatest abundance of rubble fauna after the experiment was found in tanks with S. aspera.

Discussion

A crucial first step in resolving the contribution of predator–prey interactions to population dynamics is identifying key predator species. Here, we investigated the capacity of 104 species to consume CoTS in their early post-settlement life stage. We focused our attention on predators found in coral rubble, as this is where juvenile CoTS predominantly settle and begin their early benthic life stage as herbivores (Zann et al. 1987; Wilmes et al. 2020b). There are currently > 90 species identified to consume CoTS across various life phases, most of them reef fishes and just 24 coral reef invertebrates (Cowan et al. 2017a; Kroon et al. 2020; Balu et al. 2021). Our results more than double this list with observations of 26 new invertebrate species that demonstrated varying capacity to consume CoTS in laboratory feeding assays.

By far the most consistent predator was the red decorator crab, Schizophrys aspera. Juvenile CoTS were consumed in their entirety in nearly all S. aspera feeding trials, including in the rubble mesocosm setting highlighting their ability to locate and consume CoTS among natural rubble and over a diversity of other prey options. Interestingly, rates of consumption by individual S. aspera in rubble mesocosm trials (approximately 5.8 CoTS d−1) are comparable to reported whole-of-community predation rates (5.05 CoTS d−1) for similar size–age cohorts (Keesing and Halford 1992a), although CoTS densities between studies potentially differed. The feeding rates of S. aspera documented here were lower than that measured of planktivores feeding on CoTS larvae (Cowan et al. 2016a), although larval duration of CoTS is short (days to weeks) compared to the juvenile phase (months to years), meaning the cumulative impact of predators on juveniles through time would be the critical bottleneck in CoTS ontogeny. Additionally, S. aspera consumption rates far exceed those reported on adult CoTS by Charonia tritonis (< 1 CoTS wk−1), a predator that has received considerable attention (Pearson and Endean 1969; Hall et al. 2017). These results help document how mortality rates of CoTS change during ontogeny (Wilmes et al. 2018) and emphasise the importance of predatory interactions during early life-history stages, when their susceptibility as prey is high.

The discovery of S. aspera as the foremost predator of juvenile CoTS in coral rubble at Heron Reef is not altogether unexpected. Decorator crabs have been implicated as important predators of echinoderms (Wicksten 1980; Mladenov 1983; Bonaviri et al. 2012; Clemente et al. 2013; Fagerli et al. 2014), including asteroids (Ling and Johnson 2013). While these predator–prey interactions occur between different species in different ecosystems, majoid crabs seem to be important echinoderm specialists. Decorator crabs are named given their propensity for covering their exoskeletons in a variety of taxa, including algae, sponges, ascidians and even hydroids, for camouflage (Wicksten 1980; Guinot and Wicksten 2015) and predator deterrence (Stachowicz and Hay 1999). We hypothesise that these decorating habits may confer some level of tolerance to toxins common in echinoderm species, including the saponins and plancitoxins found in CoTS (Howden et al. 1975; Shiomi et al. 1988), which make them unpalatable to many predators (Lucas et al. 1979; Shiomi et al. 1990, 2004). Four other decorator crab species, including the epialtids, Menaethius monoceros (n = 4, 1 observed incidence of consumption), Tiarinia cornigera (n = 8) and Tiarinia sp (n = 3), and the inachid, Camposcia retusa (n = 1), were also trialled in feeding assays. However, these species did not show the same consistency in consuming CoTS as S. aspera, a formidable predator deserving of further attention. In all, the demonstrated CoTS predation capacity of S. aspera should prompt further consideration of its distribution and abundance, especially relative to known outbreak initiation zones and between fished and unfished reefs.

A secondary group of predators, comprised primarily of portunid crabs, were distinct in their ability to cause injury more often than consume whole juvenile CoTS. While more opportunistic in nature, a larger number of species displayed this predation style as compared to S. aspera. Outcomes of rubble mesocosm trials with the portunid Thalamita admete exemplified this predation strategy, wherein predatory interactions occurred less frequently than in S. aspera trials and often led to injury rather than whole CoTS consumption. Lower overall abundance of other rubble taxa during these trials also suggests that CoTS are not the preferred food for this predator when other options are present. Conversely, S. aspera had the highest abundance of rubble fauna post-experiment, suggesting that CoTS may be a preferred food item.

Other predators have been noted to inflict injuries on CoTS, including the harlequin shrimp Hymenocera picta (Glynn 1982), and more recently in laboratory experiments with the peppermint shrimp Lysmata vittata (Balu et al. 2021). Indeed, a large proportion of CoTS juveniles and adults in situ are found to have injuries (Glynn 1984; McCallum et al. 1989; Messmer et al. 2017; Budden et al. 2019; Wilmes et al. 2019; Caballes et al. 2022). The likelihood of CoTS recovery from these dismembered states needs further attention as partial predation may, in some cases, result in the regeneration of multiple individuals (Lawrence and Vasquez 1996) or alternatively cause mortality depending on severity (Deaker et al. 2021). Surviving individuals with sustained injuries may display arrested growth trajectories, delayed transitions between life stages and reduced reproductive output, as recognised for other echinoderms (Lawrence and Larrain 1994; Zajac 1995; Lawrence and Vasquez 1996; Bingham et al. 2000; Maginnis 2006; Lindsay 2010; Budden et al. 2019; Deaker et al. 2021). Therefore, despite partial consumers not causing immediate CoTS mortality, the population-level consequences of partial predation may be substantial.

The remaining predators identified in this study were classified as incidental, given that CoTS were only consumed or injured on 1 or 2 occasions across feeding trials. For some of these species, such as the xanthid crabs Luniella spinipes, Actaeodes hirsutissimus, Atergatis floridus and Neoliomera insularis, this classification may be an artefact of low sample size, caused by constraints on our ability to collect individuals in the focal habitat of this study. Despite their potential to consume or injure CoTS, the low numbers of these xanthids found across our study sites suggest that their impacts on CoTS populations may be limited. Yet, 15 other xanthid species exhibited no predation (see Table S1), while two species, Etisus anaglyptus and Cyclodius ungulatus, grouped within the partial predators owing to a greater sample size and higher frequency of predator interactions with CoTS. Further research is required to demonstrate whether the species tested here can consistently consume or injure CoTS and to quantify their abundance in rubble communities in other locations.

Other incidental predators, including reef shrimp Athanas parvus and Saron marmoratus, did not generally demonstrate interest in consuming CoTS despite adequate replication of feeding trials. For these species, it is unlikely that they have meaningful interactions with CoTS in situ. This contrasts other reef shrimp (H. picta and L. vittata) known to consume CoTS (Glynn 1982; Balu et al. 2021), although we note these species are from distinct genera that were not found in rubble here. Similarly, the one incident of CoTS consumption by the fireworm Eurythoe complanata identified here contrasts the voracity of another identified amphinomid predator of CoTS, Pherecardia striata (Glynn 1984). However, P. striata predation occurs primarily on already injured CoTS, which may also motivate E. complanata but was not investigated here. Indeed, this would result in secondary predation of the partially consumed CoTS from interactions between CoTS and species of the Portunidae, which would, in turn, limit juvenile regeneration. These indirect predator interactions and the potential for facilitation among predatory species warrant further enquiry, especially given 75% of taxa (78 species) tested here did not show interest in whole starfish.

Future research could explore how the predator–prey interactions described here may change or persist across the broader CoTS ontogeny and age-size classes, including juveniles that have made the switch to a corallivorous diet and adults. We expect vulnerability to predation to change across these size and diet transitions, which can occur as early as 140–190 d post-settlement (Neil et al. 2022) but may be delayed when no coral is available (Deaker et al. 2020). Indeed, even within the herbivorous juvenile period, predation pressure may decrease significantly as body size increases, as documented in both experimental and field studies (Keesing et al. 2018; Balu et al. 2021).

We note that no juvenile CoTS were found in our searches. Given our predator collections were aligned with the time of year that CoTS juveniles are expected to be settled (Doll et al. 2021) and concentrated within the ~ 8 m depth range considered the “hotspot” of juvenile densities (Wilmes et al. 2020b), we expected to encounter some CoTS. However, the demonstrated ability of S. aspera, and to a lesser extent T. admete, to consume CoTS among natural rubble make it possible that predator populations had already reduced CoTS juvenile densities in their immediate surroundings. Heron Reef, a protected no-take zone, has experienced relatively little CoTS impact in comparison with other reefs on the GBR (Pratchett et al. 2014; Matthews et al. 2020). Indeed, protected reefs may have fewer or less severe CoTS outbreaks (Sweatman 2008; Kroon et al. 2021), which seems to be reflected at our study sites. This pattern for no-take zones is likely driven by higher predation from fishes and invertebrates in rubble (as found here) combined. Our ability to detect and collect high numbers of these predators in their cryptic habitats indicates they are indeed common in rubble on Heron Reef.

Broader information on the distribution of these novel predators remains limited but may begin to inform whether these predators have measurable influence over CoTS populations at broader reef scales, especially for S. aspera. It seems critical to incorporate the novel predators identified here in surveys on reefs open and closed to fishing to address mechanistically the potential for CoTS outbreaks to be suppressed in protected areas (Sweatman 2008; Kroon et al. 2021). Whether cryptic decapods benefit from no-take zones or are conversely at greater risk of predation themselves owing to higher fish densities requires attention, especially now in the context of trophic links to CoTS populations. In all, this study offers preliminary information that will contribute to resolving the role that predators play in regulating CoTS populations across life stages.