Introduction

Recent evidence indicates that trophic niche partitioning between syntopic scarine parrotfishes is more extensive than previously described (Nicholson and Clements 2020, 2021). Niche differences contribute to the maintenance of biodiversity by stabilizing species coexistence (Leibold and McPeek 2006; HilleRisLambers et al. 2012). Trophic niche partitioning had been predicted in studies which highlighted disparity in parrotfish cranial trophic anatomy (Bellwood and Choat 1990; Wainwright et al. 2004; Wainwright and Price 2018), implying trophic partitioning in feeding ecology (Price et al. 2010; Clements et al. 2017; Wainwright and Price 2018). Ecomorphological relationships were identified in a co-occurring assemblage of Indo-Pacific parrotfishes, where variation in trophic cranial anatomy corresponded to the successional status of carbonate feeding substrata (Nicholson and Clements 2021). This disparity in trophic cranial anatomy, combined with new evidence for feeding selectively over very fine-grained spatial scales (Nicholson and Clements 2020, 2021), points to complex trophic niche partitioning within the parrotfishes, especially in the genus Scarus.

Perhaps the most striking example of this trophic partitioning is seen in Scarus spinus, which scrapes crustose coralline algae (CCA) (Nicholson and Clements 2020, 2021). Scarus spinus was found to bite CCA significantly more than fourteen other species of syntopic parrotfish (Nicholson and Clements 2021). Scarus spinus has evolved specialized cranial trophic morphology, including a large adductor mandibulae muscle mass (Nicholson and Clements 2021), to facilitate scraping this very hard reefal substrate. CCA are non-geniculate members of the subclass Corallinophycidae (Class: Florideophyceae, Phylum: Rhodophyta) (Le Gall and Saunders 2007) and are considered less susceptible than other macroalgae to herbivory due to their encrusting nature and low nutritional quality (Montgomery and Gerking 1980; Steneck 1983, 1985, 1986; Bruggemann et al. 1994; Teichert et al. 2020). CCA are important ecosystem engineers (Nelson 2009; Schubert et al. 2020), contributing calcareous sediment to reefs and stabilizing reef frameworks by binding loose substratum (Littler and Littler 2013; Diaz-Pulido et al. 2014).

While other parrotfish have been observed biting CCA, the proportion of bites on CCA from other parrotfish species is low, ~ 3–20% (Bruggemann et al. 1994; Bellwood 1995; Hoey and Bellwood 2008; Ong and Holland 2010; Hamilton et al. 2014; Neal et al. 2020; Nicholson and Clements 2020, 2021). This is lower than expected, as CCA occupy on average a third of the benthic substratum on coral reefs (Littler 1973; Iryu et al. 1995; Dean et al. 2015), indicating that most parrotfish avoid CCA. In contrast, S. spinus appears to feed predominantly on CCA substrata (Nicholson and Clements 2020, 2021).

The “microphage feeding hypothesis” (Clements et al. 2017; Clements and Choat 2018) suggests that microphotoautotrophs associated with CCA, particularly protein-rich cyanobacteria, are the likely nutritional targets of S. spinus. CCA host microscopic biota as biofilm (Johnson et al. 1991; Adey, 1998), endophytes and endoliths (Ghirardelli 1998; Tribollet and Payri 2001). CCA biofilms (Vermeij et al. 2011; Gomez-Lemos and Diaz-Pulido 2017) and epithallial sloughing (Johnson and Mann 1986; Littler and Littler 1999) deter settlement of macroalgal spores, yet CCA promote settlement of invertebrate larvae including coral propagules (Tebben et al. 2015; Whitman et al. 2020). CCA biofilms contain cyanobacteria (Sneed et al. 2015; Nicholson and Clements 2020; Petersen et al. 2021), but biofilm composition varies between CCA species and this variation can influence the probability of coral settlement (Webster et al. 2004; Jorissen et al 2021). Species delimitation of CCA has been problematic due to a reliance on morpho-anatomical methods, and so molecular barcoding has been increasingly advocated (Gabrielson et al. 2018; Twist et al. 2019; 2020).

Various microphotoautotrophs that live as endoliths (microborers) within CCA may also serve as dietary targets for S. spinus. These include three filamentous cyanobacteria: Mastigocoleus testarum (Nostocales), Hyella spp. (Pleurocapsales) and Plectonema terebrans, the latter being the dominant endophyte in live CCA (Tribollet and Payri 2001). This cryptic cyanobacterium has been reassigned to the genus Leptolyngbya (Komárek 2007; Roush and Garcia-Pichel 2020), a taxon containing endolithic members (Horath and Bachofen 2009). Cyanobacterial taxonomy is in a state of flux (Komárek 2018; Engene et al. 2018; Dvořák et al. 2021). For example, molecular sequencing reassigned Lyngbya majuscula to at least six new genera (Nuryadi et al. 2020). Therefore, it is essential that polyphasic methods are used for cyanobacteria identification (Komárek 2016).

Dinoflagellates have also been observed as endophytes within CCA (Krayesky-Self et al. 2017; Fredericq et al. 2019). The euendolithic (true-borer), siphonous, microchlorophyte Ostreobium spp. is found under CCA crust (Tribollet and Payri 2001; Nicholson and Clements 2020) and is also a potential nutritional target of excavating parrotfishes (Clements et al. 2017; Clements and Choat 2018).

The concept that scarine parrotfishes target microphotoautotrophs (Clements et al. 2017; Clements and Choat 2018) led us to hypothesize that high-protein microphotoautotrophs, particularly filamentous cyanobacteria, are consistently associated with substrata targeted by Scarus spinus. The present study tests this hypothesis by extracting reef core samples of substrata grazed by these parrotfish at distinct mid-shelf and outer-shelf locations and uses a combination of microhistology and DNA next-generation sequencing to determine which biota are consistently associated with targeted substrata. We note that microscopic dietary targets cannot be readily identified on the reef a priori; therefore, our methodology is designed to bring targeted reef samples to the microscope to ensure high dietary resolution.

Materials and methods

Core sampling protocol

Sampling was conducted at the Lizard Island Complex (Map: Supplementary Fig. S1), Great Barrier Reef, Australia, over 10 days in March 2018 at the front of South Island (a mid-shelf site) and 10 days in February 2019 at Day Reef (an outer-shelf site). Individual terminal phase Scarus spinus were followed on snorkel until feeding was observed. The targeted bite site was immediately photographed in situ, and the feeding substratum surrounding the bite site was then extracted using a handheld brace with a 22-mm-diameter hole saw attachment. This produced a sample (bite core) of approximately 2 cm depth. Each bite core was returned immediately to the boat and stored on ice. Bite cores were photographed and then fixed in 70–80% ethanol within 1–3 h of extraction. Bite cores were taken of the first bites observed for each individual fish that was followed. Forty S. spinus bitecores were extracted, 20 each for the mid-shelf and the outer-shelf.

Bite core indices

Three indices were parameterized for all 40 bitecores following Nicholson and Clements (2020): (1) epilithic biota percentage surface cover using Coral Point Count (CPCe v 4.1, Kohler and Gill, 2006), (2) density of filamentous cyanobacteria (Supplementary Fig. S2), and (3) density of diatoms.

Statistical analysis for bite core indices

To compare similarities and differences between the mid-shelf and outer-shelf samples in bite core epilithic composition, the CPCe dataset was used to perform non-metric multidimensional scaling (nMDS) using a Bray–Curtis distance matrix. The function metaMDS in the “vegan” package (Oksanen et al. 2019) in R (R Core Team 2021) was used for the nMDS ordination. A minimum stress value of 0.107 was produced after 25 iterations, and the scree plot indicated two dimensions were appropriate, producing a nonmetric R2 = 0.989. Ellipses representing 95% confidence intervals were drawn around centroids for the two samples using the function ordiellipse in the “vegan” package. Significant differences between the mid-shelf and outer-shelf samples in bitecore epilithic community composition from the CPCe analysis were examined using a permutational multivariate analysis of variance (PERMANOVA) test in R using the “vegan” package function adonis2, after ensuring the homogeneity of variances using the betadisper function (McArdle and Anderson 2001). To observe which bitecore biota produced the variability between the samples, we used vector fitting by performing the “vegan” envfit function, and significant variables (p < 0.01) were fitted to the NMDS plot. To test whether the samples differed significantly in the density of filamentous cyanobacteria and diatom cells, we used one-way analysis of variance (ANOVA).

16S/18S next-generation sequencing

Fragments of the variable V3-V4 region of the 16S SSU ribosomal RNA gene (464 bp) were amplified to further resolve cyanobacterial taxonomy. We used 18S DNA sequencing to determine CCA taxa. Following the 1cm2 scrape for microscopy, the other half of the bitecore surface was scraped to a depth of 1 mm for DNA analysis. These bitecore surface scrapings were then ground using a pestle and mortar and DNA extracted using the DNeasy Blood and Tissue Kit (QIAGEN, Germany). Our gene amplification protocol is described in supplementary methods. Bioinformatic analysis used QIIME2 2021.4 (Bolyen et al. 2019) with the DADA2 q2-plugin (Callahan et al. 2016; see supplementary methods). Nearest-neighbor 16S reference sequences for filamentous cyanobacteria were downloaded from CYDRASIL v.3 (Roush et al. 2021) and combined with the bitecore cyanobacteria ASVs to construct a phylogenetic tree. To construct the phylogenic tree for CCA, 18S reference sequences for coralline algae were collected, including non-geniculate and geniculate genera. Since CCA are polyphyletic, these references sequences were combined with the bite core ASVs classified as Corallinophyidae (see supplementary methods).

Bite penetration

To determine whether Scarus spinus could access the Ostreobium-rich layer under live CCA crusts, we measured two variables: bite depth and CCA crust depth. Prior to scraping the bitecores for microscopy and DNA sampling, photographs of bite marks were taken with a Nikon Stereoscopic microscope (SMZ 745 T) with Infinity2 camera. Length and width of the bite scars were measured using Infinity Analyze software (Lumenera®, Ottawa, Canada) or ImageJ (NIH). We used the Z-stacking function in the Leica Application Suite (Version 4.13.0, Switzerland) to measure depth of bite scars using a Leica M205 C (Leica Microsystems, Heerbrugg, Switzerland). For CCA crust thickness, a cross-section photograph was taken of all forty bitecores at 2.5 × magnification and the thickest part of the CCA crust was measured (Supplementary Fig. S3).

Results

Densities of filamentous cyanobacteria and diatoms did not differ between mid-shelf and outer-shelf locations (cyanobacteria lengths: F1,38 = 0.013, p = 0.908; diatoms cells: F1,38 = 1.562, p = 0.219; Fig. 1, Table 1). The mean percentage cover of CCA on the outer-shelf bitecores was significantly higher at 91.4% than the 76.8% CCA coverage on the mid-shelf bitecores (F1,38 = 23.046, p = 0.046) (Fig. 1).

Fig. 1
figure 1

Raincloud plots comparing four bite core parameters between the mid- and outer-shelf samples: CCA % cover, CCA crust depth (mm), cyanobacterial filament lengths (um) per cm2, number of diatom cells per cm2. Lighter shades: 2018 mid-reef shelf (n = 20); dark shades: 2019 outer reef shelf (n = 20)

Full size image
Table 1 Means and SE for twelve bite core parameters for Scarus spinus mid-shelf and outer-shelf
Full size table

Frequency occurrence for the bitecore epilithic biota observed in the microscopy is listed in Supplementary Table S1, and surface coverage from CPCe for each individual bite core is presented in Supplementary Fig. S4. Overlap in epilithic community composition between samples was revealed in the 95% confidence interval ellipses on the NMDS ordination plot (Fig. 2). The bitecore epilithic communities were more uniform in the outer-reef 2019 samples than in the mid-shelf 2018 samples. Epilithic bitecore community composition using the CPCe dataset was more similar within samples than between samples (PERMANOVA p = 0.005; betadisper p = 0.060) due to the mid-shelf samples displaying greater variation. Vector fitting using the ENVFIT function indicated that four categories, i.e., CCA, cyanobacterial tufts (cyanobacteria > 1 mm, visible as 5 × magnification), Gigartinales and Ceramiales, were responsible for the variation in bitecores between samples (p < 0.01), and these intrinsic variables were fitted to the NMDS plot with red arrows to indicate the direction of their variability.

Fig. 2
figure 2

Non-metric multidimensional scaling (NMDS) ordination plot comparing community composition of the epilithic bitecore biota from the CPCe analysis using Bray–Curtis dissimilarity matrix. Triangles represent the individual bitecores (n = 40), multiple bitecores are listed using the blue arrows. Ellipses are 95% confidence intervals around the centroids for the mid-shelf and outer-shelf bitecores. Blue arrows indicate core numbers for overlapping data points. Red arrows indicate effect and direction of epilithic biota significantly correlated with community dissimilarity (ENVFIT, p < 0.01)

Full size image

Cyanobacteria taxonomy

Overall, the average 16S sequence fragment length obtained was 422 bp (see Supplemental Table S3 for summary statistics). Cyanobacteria (excluding chloroplasts) represented 15.7% of all 16S reads mid-shelf and 13.1% for the outer-shelf bitecores. Taxonomic assignment produced 126 ASVs for filamentous cyanobacteria. The most frequently observed filamentous cyanobacteria by microscopy were the heterocystous, tapered, filamentous cyanobacteria Calothrix spp. (Nostocales), which were observed on 100% of the S. spinus bitecores (Fig. 3; Supplementary Table S1). Short Calothrix trichomes with basal heterocysts were present, resembling Calothrix fusca and Calothrix confervicola (Komárek and Anagnostidis 1989), as well as longer Calothrix trichomes with intercalary heterocysts (Supplementary Fig. S5). Longer Calothrix trichomes are most likely the same species as heterocyst differentiation in Calothrix varies with culture medium (Rai et al. 1978). While most of the Calothrix occurred as biofilm, the shorter Calothrix trichomes were observed just under the surface of the CCA thallus, suggesting an endophytic lifestyle similar to the endolithic capacity described from dolomite (Sigler et al. 2003). Sequence data confirmed the presence of Calothrix in mid- and outer-shelf samples, with 27 of the 126 ASVs falling within Calothrix/Rivularia based on Cydrasil placement (Fig. 4). Another heterocystous filamentous cyanobacterium, the euendolith Mastigocoleus testarum (Nostocales), was observed microscopically on 55% of the bite cores in the mid-shelf samples and 50% in the 2019 outer-reef samples (Fig. 4; Supplementary Table S1). The molecular data confirmed the presence of M. testarum in both mid- and outer-shelf samples. Morphotypes of Lyngbya majuscula (Oscillatoriales) were observed microscopically on 85% of the mid-shelf bitecores and 75% of the outer-shelf bitecores (Supplemental Table S1). This microscopic identification represents a range of morphologically similar Oscillatoriales (Engene et al. 2012, 2013, 2015, 2018; Nuryadi et al. 2020), a finding supported by eight ASVs falling within the Lyngbya-like Oscillatoriales (Fig. 4). The non-heterocystous Spirulina subsalsa (Spirulinales) was observed microscopically, and its presence confirmed in mid-shelf samples, while the morphologically similar Halospirulina (Spirulinales) was detected at both mid- and outer-shelf locations (Fig. 4). Sixty ASVs fell within the genus Leptolyngbya (Fig. 4), which have thin isopolar trichomes with isodiametric cells. Some Leptolyngbya were photographed in the same plane as the euendolith Mastigocoleus testarum, indicating that these may be the endophyte Plectonema (Leptolyngbya) terebrans (Fig. 3; Supplementary Fig. S5).

Fig. 3
figure 3

Microscope photographs of the most common filamentous cyanobacteria found on Scarus spinus bite cores. Black arrows indicate cyanobacteria taxa Ci: Long trichomes of Calothrix (Nostocales) with intercalary heterocysts, Cb: Short Calothrix (Nostocales) with only basal heterocysts, M: Mastigocoleus testarum (Nostocales) euendolith, showing lateral heterocyst and true-branching, H: Hyella patelloides (Pleurocapsales) an endolithic and pseudo-filamentous rare biota, S: Spirulina subsalsa (Spirulinales) with spiral non-heterocystous trichomes, L: Leptolyngbya spp., (Synechococcales) with very, thin trichomes and isodiametric cells, possibly representing the cryptic Plectonema terebrans. LM: Morphotype of Lyngbya majuscula (Oscillatoriales) non-heterocystous, non-branching (Photos GN)

Full size image
Fig. 4
figure 4

Circular maximum-likelihood 16S tree of the bite core cyanobacteria ASVs and their closet relative reference sequences from Cydrasil. Phylogenetic tree was constructed with FASTTREE alignment and generated with iTOL (https://itol.embl.de). Clade support (bootstrap) values greater than 80 are indicated with a black triangle. Stars indicate presence in molecular dataset in mid-shelf (light blue), outer-shelf (dark blue). Coloured bands indicate characteristics of the clade/ order

Full size image

Ostreobium as food?

The depth of CCA crusts ranged from 0.20 to 2.68 mm on the mid-shelf bitecores, and 60% of the mid-shelf bitecores had crusts less than 1 mm thick. The thickness of crusts ranged from 0.42 to 4.61 mm on the outer-shelf bitecores, with 35% of crusts less than 1 mm thick (Fig. 1). The mean thickness of CCA crusts was greater on the outer-reef bitecores (1.51 mm) than the mid-reef bitecores (1.03 mm); however, this difference was not significant (F1,38 = 2.794, p = 1.03; Table 1). 16S sequencing detected Ostreobium only from the mid-shelf sample, with nine ASVs assigned to Ostreobium spp. (Supplemental Fig. S7); however, in the 18S dataset it was present in both mid- and outer-shelf samples. The Ostreobium green band, which occurs directly under live CCA, was visible in some in situ bite photographs (Fig. 5). The mean S. spinus bite area was 5.95 mm2 on mid-shelf bitecores and 5.4 mm2 on outer-shelf bitecores (Supplementary Fig. S9). We were only able to determine bite scar depth for four of the outer-shelf bite cores, as we had limited access to the required equipment. Mean S. spinus bite depth for was 0.13 mm (Table S2).

Fig. 5
figure 5

Six examples of Scarus spinus in situ bite site photographs from Lizard Island GBR, mid-shelf reef sites and outer-reef sites, showing bite marks on CCA (Photos KDC). Photographs were taken immediately after each bite was observed, prior to bite core extraction

Full size image

18S and CCA taxonomy

The average 18S sequence fragment length was 351 bp (see Supplemental Table S3 for summary statistics). The 18S SILVA taxonomic analysis assigned 14 ASVs to the subclass Corallinophycidae. Three ASVs were assigned to the geniculate genera Jania and Amphiroa. Eleven ASVs corresponded to CCA (Supplementary Fig. S10), four within the family Porolithaceae and seven within Hydrolithaceae. The remaining CCA ASV was assigned to the genus Neogoniolithon, present only in our mid-shelf sample. 18S sequences of the early successional euendolithic microchlorophyte Phaeophila dendroides (Order Ulvales) were detected in both mid- and outer-shelf samples. One Symbiodiniaceae ASV (Cladocopium) was detected in both mid-shelf and outer-shelf samples (Supplemental Fig. S7). Sequences corresponding to the epiphytic Ceramiales genera Polysiphonia and Ceramium were detected in both mid- and outer-shelf samples; however, the filamentous rhodophyte Herposiphonia was only detected in mid-shelf samples. The turfing rhodophyte Wurdemannia miniata (Gigartinales) was also only detected in mid-shelf samples, supporting the results from the microscopy NMDS.

Discussion

The microscopy and molecular results both supported our hypothesis that the substrata grazed by Scarus spinus consistently involved CCA associated with endophytic, endolithic and epiphytic high-protein microphotoautotrophs including filamentous cyanobacteria.

CCA were present on every bitecore; the in situ photographs, taken immediately after feeding, show that S. spinus bites were exclusively on CCA. The mean surface coverage of CCA on the bitecores was significantly higher on the outer-reef than the mid-shelf bitecores (91% and 77%, respectively). This variation in CCA cover is consistent with published benthic transect surveys, which report higher CCA abundance at outer reef sites than mid-shelf sites (Scott and Russ 1987; Fabricius and De’ath 2001; Wismer et al. 2009; Dean et al. 2015). Our DNA sequencing detected Hydrolithaceae and Porolithaceae in both mid- and outer-shelf samples, whereas Neogoniolithon was only detected in mid-shelf samples. Neogoniolithon has been recorded previously in low abundance on both mid- and outer-shelves, while Porolithon onkodes is described as the dominant species of CCA at both mid- and outer-shelf reefs on the GBR (Dean et al. 2015). The presence of CCA on all S. spinus bitecores indicates an active preference for CCA in both locations sampled. Scarus spinus seems to target CCA irrespective of habitat or geographic location. In addition to our observations within the Lizard Island Complex, we observed the same specialized feeding in other locations, including Fiji (Supplemental Fig. S8A, B and video S3). Additionally, Scarus viridifucatus, the Indian Ocean sister species to S. spinus, appears similar in trophic anatomy and habitat, suggesting a similar diet (as seen in Supplemental Fig. S8C). Our study revealed that it is microphotoautotrophs living epiphytically on the CCA and as shallow endophytes within CCA that are the main nutritional targets for S. spinus.

Filamentous cyanobacteria were present on all 40 bite cores, consistent with the hypothesis that microphotoautotrophs including cyanobacteria are targeted by parrotfish (Clements et al. 2017; Clements and Choat 2018). The density of filamentous cyanobacteria on the substrata where S. spinus fed was consistent between locations. This was of particular interest as the mid-shelf samples were closer to terrigenous inputs which can boost cyanobacterial densities (Huisman et al. 2018; Ford et al. 2018; Zubia et al. 2019). Furthermore, the mid-shelf samples were collected closer in time to four consecutive major disturbances (bleaching events and cyclones) at the Lizard Island Complex (Madin et al. 2018; Hughes et al. 2019; Zawada et al. 2019). Such disturbances can also boost cyanobacterial densities (Larkum 1988; Beltram et al. 2019; Wismer et al. 2019; Ford et al. 2021), and accordingly, we had anticipated a higher density of cyanobacteria in the mid-shelf samples. We note that the CCA substrata comprising S. spinus bitecores had the lowest density of filamentous cyanobacteria of the five syntopic parrotfishes examined by Nicholson and Clements (2020). This, and the low volume of material likely ingested per bite, may explain why S. spinus display among the highest bite rates (42–45 bites/min) recorded in parrotfishes (Bellwood and Choat 1990; Hamilton et al. 2014; Supplemental videos S1, S2).

Calothrix (Nostocales) was observed on 100% of the S. spinus bitecores. The presence of Calothrix on coral reefs is reported in early studies (Randall 1955; Scott and Russ 1987; Larkum 1988; Larkum et al. 1988), and CCA and Calothrix are considered primary colonizers of bare substratum following grazing (Hixon and Brostoff 1996; Wilkinson and Sammarco 1982). Calothrix as a nutritional target is supported by early observations where Calothrix was identified as a food for parrotfishes: “During higher stages of the tide, herbivorous reef fish, notably several species of parrotfish and surgeonfish graze the intertidal reef flat Calothrix community” (Wiebe et al. 1975). Calothrix is consumed by other fish including the surgeonfish Acanthurus guttatus (Randall 1955; Chartock 1983), rabbitfish Siganus spp. (Bryan 1975; Paul et al. 1990), minnows Compostoma spp. (Power et al. 1988), rock-dwelling African cichlids (Reinthal 1990; Genner et al. 1999) and wild ayu Plecoglossus altivelis (Nakagawa et al. 2002). Calothrix is used in aquaculture feeds due to its high fatty acid and protein content (Olvera-Ramírez et al. 2000; Ruiz-Ramírez et al. 2005).

Although Calothrix was the most frequently observed filamentous cyanobacteria on the bite cores, our molecular data revealed a lower proportion of Calothrix sequences than expected based on microscopy. This echoes previous studies where heterocystous cyanobacteria were conspicuous in microscopy but undetected in sequencing (Garcia-Pichel et al. 2001; Ford et al. 2021). The problem with Calothrix appears to lie with DNA extraction due to their thick, tough, extracellular, polysaccharide sheaths (Tillet and Neilan 2000; Urrejola et al. 2019; Garcia-Pichel et al. 2001; Sihvonen et al. 2007). An amplification bias against filamentous cyanobacteria toward coccoid cyanobacteria in general 16S primers has also been noted (Lorenzi et al. 2019). These biases in DNA extraction and PCR amplification emphasize the need for a polyphasic approach when studying cyanobacteria (Garcia-Pichel et al. 2001; Komárek 2016). The underrepresentation of abundant filamentous cyanobacteria in standard 16S sequencing surveys highlights the point that relative read abundance does not equate to biomass, and so quantitative methods must be used in parallel when diet tracing (Nielsen et al. 2017; Deagle et al. 2019).

Lyngbya-like morphotypes were observed microscopically in low abundance on 80% of bite cores, and our molecular data included eight Oscillatoriales ASVs. Five of these ASVs were Lyngbya-like Lyngbya sp., Okeania plumata and Lyngbya aestuarii. Three were nearest neighbors to Lyngbya aestuarii, a cyanobacterium used as fish food in Philippine aquaculture (Nweze 2009). Three ASVs were a close match to Arthrospira platensis (Oscillatoriales), which is morphologically similar to Spirulina subsalsa. ASVs of the latter species were also present. Arthrospira platensis is commercially important, being rich in protein (60–70%) and essential fatty acids (Cohen 1997). Arthrospira platensis has long been used as a food for humans (Ciferri 1983) and a food supplement for animals (Becker 2007). Due to its helically coiled trichomes, it is sold under the name SPIRULINA, which is misleading because it belongs to the Order Oscillatoriales rather than Spirulinales (Nowicka-Krawczyk et al. 2019).

The third most frequently occurring cyanobacterium observed microscopically was the heterocystous euendolith Mastigocoleus testarum (Nostocales), a bioeroder involved in the dissolution of marine calcium carbonate (Ramírez-Reinat and Garcia-Pichel 2012; Guida and Garcia-Pichel 2016) that lives endolithically within both live and dead CCA (Tribollet and Payri 2001; Diaz-Pulido et al. 2014). The pseudofilamentous euendolith Hyella patelloides (Pleurocapsales) was also detected among the 16S sequences, but was rarely observed on the bitecores and is reported rare in endolithic assemblages (Grange et al. 2015). Mastigocoleus and Hyella are photophiles and bore only to shallow depths (Tribollet 2008), so are easy prey to grazers (Grange et al. 2015; Roush and Garcia-Pichel 2020). The identification of Leptolyngbya by microscopy and 16S ASVs is of interest, as Leptolyngbya (Plectonema) terebrans has been reported to comprise 80% of the endolithic assemblage within live CCA (Tribollet and Payri 2001) and is a dominant pioneering microborer in dead carbonate (Roush and Garcia-Pichel 2020).

The microscopic chlorophyte Ostreobium is at its densest under live CCA crust (Tribollet and Payri 2001). Our bite depth measurements showed that S. spinus bites were too shallow to penetrate through the majority of CCA crusts; however, repetitive bites on thinner CCA, such as observed on the mid-shelf, could expose the Ostreobium. Epiphytic Ceramiales were observed microscopically in low abundance on the mid-shelf bitecores and lower still on the outer-shelf bitecores, indicating that these are not nutritional targets for S. spinus. The presence of Symbiodiniaceae is consistent with studies showing that parrotfishes are vectors of dispersal for viable Symbiodinium (Porto et al. 2008; Grupstra et al. 2021). Diatoms, present on 92% of the bitecores, are reported to have a symbiotic relationship with Calothrix (Foster et al. 2011).

Why target CCA as a food source? Oculomotor data indicate that parrotfish locate prey visually (Rice and Westneat 2005) and CCA would be easy to find using visual cues. We propose that S. spinus targets CCA as a reliable source of high-protein microphotoautotrophs, including filamentous cyanobacteria. Ingestion of calcium carbonate in CCA may aid in trituration of food by the pharyngeal jaws (Carr et al. 2006), releasing cells from the thick mucilaginous sheaths of these filamentous cyanobacteria. If some cyanobacteria are toxic, such as Lyngbya majuscula (Albert et al. 2005; Engene et al. 2012, 2013; Taylor et al. 2014), ingestion of CCA might serve to reduce toxicity as calcium has been shown to protect Daphnia from toxin of the cyanobacterium Microcystis (Akbar et al. 2017).

Our data show that filamentous cyanobacteria, in particular endophytic and epiphytic Nostocales, were consistently associated with the CCA grazed by S. spinus. Cyanobacteria are rich in protein, but lack essential long chain polyunsaturated fatty acids (PUFAs) (Ahlgren et al. 1990; 1992). Diatoms, present on almost all bitecores, contain the PUFAs eicosapentaenoic acid (EPA, 20:5n-3) and docosahexanoic acid (DHA, 22:6n-3) (Dunstan et al. 1993; Dalsgaard et al. 2003; Chen 2012). Ostreobium spp. also contain these PUFAs and in addition contain docosapentanoic acid (DPA, 22:5n6) (Massé et al. 2020). The biotic assemblage associated with CCA therefore appears to provide S. spinus with a dilute but nutritionally complete food source.

The co-evolution of CCA and herbivory has been discussed elsewhere (Steneck 1985; Teichert et al. 2020); however, this is the first detailed investigation of the dietary composition of a CCA specialist. Parrotfish are identified as priority species for coral reef functioning (Wolfe et al. 2020), and the present study and others (Russ et al. 2015; Clements et al. 2017) highlight the complex dynamics of the parrotfish–coral reef relationship. Bites on CCA may aid coral recruitment, and since coral recruits are more successful in microcrevices (Mallela 2018), shallow scraping bites may provide microtopographical furrows for coral seeding. However, grazing may also remove coral recruits (Doropoulos et al. 2016; Jorissen et al. 2020). Coral propagules settle preferentially on climax biofilms (Webster et al. 2004), suggesting that freshly grazed CCA may hinder coral settlement. We are unaware of another parrotfish within the GBR system that exhibits this particular feeding behavior, highlighting the need for species-specific data on parrotfish herbivory.