Functional biogeography of marine vertebrates in Atlantic Ocean reefs

Marine vertebrates play key functional roles on reef ecosystems. Despite their phylogenetic distance, different vertebrate lineages could play similar functions on reefs, which has been overlooked by current research on marine functional biogeography. We provide the first comprehensive assessment of the functional structure and inventory of ecosystem functions delivered by 224 vertebrates—marine mammals, sea turtles, sharks, rays and bony fish—in Atlantic Ocean reefs.


| INTRODUC TI ON
Human influence has reached almost every place on Earth (Lewis & Maslin, 2015;Steffen, Crutzen, & McNeill, 2007). In the oceans, intense and widespread anthropogenic impacts such as overfishing, pollution and habitat loss are threatening species and their functions (He & Silliman, 2019;Young et al., 2016). Marine biodiversity has been declining and changing considerably over the Anthropocene, with potentially greater losses of top predators (e.g. mainly sharks), large-bodied (e.g. marine megafauna), habitat specialists and species with terrestrial contact (e.g. sea turtles and pinnipeds; Ceretta et al., 2020;McCauley et al., 2015;Pimiento, Leprieur, et al., 2020). Nevertheless, the influence of declines and changes in biodiversity on ecosystem functions such as productivity, nutrient cycling and trophic regulation still needs to be better understood Fonseca & Ganade, 2001;Larsen et al., 2005;Levine, 2016;Mouillot et al., 2014). Within that context, the use of functional diversity can reveal the effects of biodiversity loss on ecosystem functioning, beyond the loss of taxonomic entities (Bellwood et al., 2004;Mouillot et al., 2014;Pimiento, Leprieur, et al., 2020;Tavares et al., 2019;Villéger et al., 2017). Our study aims to identify the potential overlap in functions of vertebrate species from distant lineages and assess the effect of simulated species loss on reef ecosystem functions.
The extinction of reef fauna has largely contributed to the declines and changes in species and functions observed in the oceans (Bellwood et al., 2004;Hammerschlag et al., 2019;Heithaus et al., 2008;Pimiento, Leprieur, et al., 2020). Top predators such as sharks consume large amounts of prey body mass and control their populations (Ruppert et al., 2013). Grazers, in their turn, can limit algae growing and ensure coral reef resilience (Adam et al., 2015;Christianen et al., 2019;Goatley et al., 2012). Species traits are assumed to be linked to these ecological functions (e.g. maximum body size, body mass, trophic group, schooling behaviour, metabolic rate and mobility; Bellwood et al., 2019;Tavares et al., 2019). The role of top predators is linked to high trophic levels, large body sizes and body mass (Roff et al., 2016;Tavares et al., 2019), while effective grazing appears related to eye diameter and position, gape position and shape, total gut length and body size (Bonaldo et al., 2014;Villéger et al., 2017). A single trait as body size, for example, is related to bioturbation (Bonaldo et al., 2014;Tavares et al., 2019), individual mobility (Villéger et al., 2017), nutrient cycling (Allgeier et al., 2014;Tavares et al., 2019), trophic regulation and community structuring in marine vertebrates (Tavares et al., 2019). Yet, the diversity of ecosystem functions performed by different taxonomic groups and their degree of functional redundancy in reef assemblages remains poorly known.
When ecological communities lose species, they do not necessarily lose functions and services (Mouillot et al., 2014) because these can be insured by species with similar traits relative to the species being lost (Mouillot et al., 2014;Pimiento, Bacon, et al., 2020). Such "functional redundancy" is an emergent property of ecological communities that depends on local species richness and trait similarity among co-occurring species-more species implies more functions, and more species per function implies more insurance (Fonseca & Ganade, 2001;Halpern & Floeter, 2008;Pimiento, Leprieur, et al., 2020;Rosenfeld, 2002). Also, functions and services may be insured by distantly related taxa (e.g. algae removal may be maintained by fishes after sea turtles become locally extinct; Goatley et al., 2012); however, we have just an incipient knowledge about functional redundancy across taxa (e.g. Pimiento, Leprieur, et al., 2020). For example, "functional uniqueness," which is an indicator of functional redundancy according to the overall isolation of each species in total trait space, highlights the irreplaceability of each species to perform unique ecosystem functions and services (Bellwood et al., 2003;Pimiento, Leprieur, et al., 2020); the extinction of functionally unique species implies direct loss of such ecosystem properties (Mouillot et al., 2014). "Functional specialization" represents the mean distance of a species from the total species pool in trait space, with specialist species exhibiting extreme trait combinations, and contributing for functional diversity .
Here, we provide a comprehensive assessment of the functional diversity of reef vertebrates from the Atlantic Ocean, including the Caribbean Sea. First, we compiled a taxonomically comprehensive database of the ecosystem functions of 224 species of bony fishes, sharks, rays, sea turtles and mammals.
Traits compiled were maximum body size, maximum body mass, trophic group (diet), maximum depth, caudal fin and body shape classifications and are linked to six reef species functions: herbivory pressure, bioturbation/bioerosion, coral reef resilience, mesopredation, top predation and trophic regulation Tavares et al., 2019;Villéger et al., 2017). Then, we spatialized such trait information at the regional and local assemblage levels to provide a geographically comprehensive assessment of the influence of marine vertebrate loss on Atlantic Ocean reefs. We simulated future extinction scenarios of vertebrate species extinction based on IUCN ranks and quantified the impact of potential species losses on functional diversity in Atlantic reefs (Leitão et al., 2016). If species richness safeguards functions, K E Y W O R D S coral reefs, cross-taxa, ecosystem functions, marine megafauna, multi-taxa, threatened species then future extinctions in species-rich regions should not influence functional richness. In contrast, future extinctions should erode functional richness and redundancy in species-poor areas, but should increase the uniqueness between closest species.
Additionally, we tested the hypothesis that the future extinction of sharks should cause decreases in community functional richness due to its combinations of functional traits (e.g. large size, body mass and trophic group) and importance in the regulation of trophic cascades. In this case, the ecosystem functions performed by this group, as that of top predators, should be under greater threat (Dulvy et al., 2014;Heithaus et al., 2008). Also, we predicted that herbivory would be severely compromised at local reef communities (Atwood et al., 2020;Bellwood et al., 2004), due to the low number of species that support this function in Atlantic Ocean reefs (Siqueira et al., 2019).

| Database
We compiled information on 224 marine vertebrate reef-associated species: four marine mammals, five sea turtles, 89 elasmobranchs and 126 bony fishes. The species considered here use rhodolith beds, coral and rocky reefs or coralline algae banks for sheltering, feeding and spawning (Pinheiro et al., 2018). The Teleostei families considered were Acanthuridae, Carangidae, Epinephelidae, Kyphosidae, Lutjanidae, Girellidae, Serranidae subfamily Epinephelinae and the Labridae subfamily Scarinae (parrotfishes), based on the recognized importance of these groups as predators, mesopredators or herbivores in reefs (Bonaldo et al., 2014;Ferreira & Gonçalves, 2006;Longo et al., 2014;Morais et al., 2017). The list of Atlantic reef fish species was obtained by combining (1) the most up-to-date available inventory of Elasmobranchii and Teleostei for the Southwestern Atlantic region (Pinheiro et al., 2018) with (2) the GASPAR Project database (General Approach to Species-Abundance Relationships) that provided data for Teleostei species found in the Atlantic Ocean (Bender et al., 2017;Kulbicki et al., 2013;Parravicini et al., 2013) and  Table S1.
To build a species presence matrix at the Atlantic Ocean scale and to capture regional variation on reefs' structure and diversity, we divided the Atlantic reef area into ten provinces (sensu Spalding et al., 2007)

| Species traits and ecosystem functions
We compiled traits that represent species performance and fitness in reef environments, as well as a consistent link with ecosystem functions Tavares et al., 2019;Villéger et al., 2017). Traits assigned to species in this study were maximum body size (cm), maximum depth of occurrence, trophic group (diet category), maximum body mass (g), caudal fin morphology and body shape classifications (Table 1). Trait data were compiled from FishBase (www.fishb ase.org), seaLiFeBase (www. seali febase.org), iucn (www.iucnr edlist.org), and also from a literature search. For Teleostei reef fish species, maximum body size (cm), inhabited depth category (maximum depth of occurrence) and trophic group (diet category) traits were obtained from the GASPAR Project database (Mouillot et al., 2014) and from the most up-to-date list of traits for Atlantic reef fishes (Quimbayo et al., 2021). Maximum body mass (M) was estimated for each species using weight-length relationships: M = aLt b , where Lt is the species' maximum recorded length and a and b were coefficient estimates for species, which were obtained from FishBase and seaLiFeBase references (Froese & Pauly, 2019;Palomares & Pauly, 2019). For species lacking specific length-weight parameters, we used averaged congener coefficients. The depth categories followed an ascending depth order (Pinheiro et al., 2018).
Marine mammals actually have tails, but we classified them as caudal fins so that this attribute would be comparable across different taxonomic groups. This classification was extended from bony fish species (Quimbayo et al., 2021) and applied cross-taxa because the caudal fin determines swimming capacity/ability and prey capture efficiency (Fish et al., 2008;Fu et al., 2016;Lingham-Soliar, 2005;Villéger et al., 2017), and is therefore associated with functions based on vertebrate species feeding and position in the water column. For species without caudal fin, as turtles and 20 ray species, "absent" was inserted for that trait.
Ecosystem functions of each reef vertebrate species were compiled from the literature through an online search using the following keyword combinations: "spp. + ecosystem function" and "spp.

| Data analysis
Analysis of the functional structure of Atlantic reef vertebrate assemblages was conducted at four spatial scales: (1) regional scale, which corresponds to the Atlantic Ocean; (2)

| Functional diversity at the Atlantic reef space
To assess the functional diversity of Atlantic reef vertebrates at the regional and province scales, quantitative traits as species' maximum body size (cm) and maximum body mass (g) were first standardized to avoid that traits with large variation cause a disproportional influence on results (scale function; "base" Package; R Core Team, 2018). We then measured the dissimilarity across the 224 species in our data set considering their six functional traits using the Gower distance (Gower & Legendre, 1986) and conducted a Principal Coordinate Analysis (PCoA) to ordinate species into a functional space (Villéger et al., 2008). To assess the quality of the functional space, we used the mean squared deviation (mSD; Maire et al., 2015) ( Figure S1). According to this metric (mSD), we selected the four main PCoA axes to build the functional trait space because a 4D space faithfully represents the original Gower's distances. The functional spaces occupied by vertebrate assemblages in Atlantic Ocean reefs, by different taxonomic groups and by threatened and non-threatened species were calculated using the convhulln function ("geometry" Package) (Habel et al., 2015). We also applied this function to measure the overlap between different taxonomic groups in functional space. To assess the unique and shared functions of reef vertebrate fauna, we grouped species according to their ecosystem functions categories (see Figure 1

| Mapping functional richness, uniqueness and specialization in Atlantic Ocean reefs
For each grid cell, we calculated species richness as well as a set of metrics describing the functional structure at the assemblage level.
We first calculated the functional richness (FRic) corresponding to the trait volume occupied by species co-occurring in each grid cell (Villéger et al., 2008). We then characterized the level of isolation of each species inside the functional space for each grid cell, which allows quantifying the level of species' uniqueness or redundancy at the assemblage level (Mouillot, Graham, et al., 2013;Pimiento, Leprieur, et al., 2020).
To do so, we calculated the mean functional uniqueness (FUn) per grid cell as follows. First, FUn was calculated based on Gower's distance as the mean distance between each species and the five nearest neighbouring species in the grid cell (see Pimiento, Leprieur, et al., 2020). Then, FUn was averaged across all species co-occurring in each grid cell. The mean functional uniqueness (or functional originality "FOri" sensu Mouillot, Graham, et al., 2013) increases as species contained in an assemblage share less traits with others.
We also calculated the mean functional uniqueness considering the three, ten and fifteen nearest neighbouring species (see Figure S2).
These regions were used because low species richness for "bioturbation/bioerosion" and "herbivory pressure" ecosystem functions at the assemblage scale (grid cell) precluded the calculation of functional richness. We used a null model approach to randomly remove Then, to evaluate whether the potential extinctions of mesopredators may lead to a disproportionate decrease in functional richness (i.e. greater loss of functional richness than that expected by species richness alone), we used a null model that keeps constant the species richness in each assemblage (grid cell) and randomly removes from 1% to 100% of species 1,000 times, without considering their ecosystem functions. The "observed loss of functional richness" (I obs ) is compared to 1,000 simulated values (I sim ) (Gotelli & McCabe, 2002) by extracting the standard deviation (σ sim ) and the average (I sim ), and subsequently by calculating a standardized effect size (SESn) as follow A positive value indicates a greater decline in functional richness than that expected by randomly removing species from local assemblages. All analyses were performed using the R software version 3.4.4 (R Core Team, 2018).

| The ecosystem functions of marine vertebrates in Atlantic Ocean reefs
Our literature search through 160 papers resulted in a list of seven ecosystem functions of reef vertebrates: 1. herbivory pressure, 2.

| Linking threatened species, traits and ecosystem functions
The first four axes of the Principal Coordinate Analysis (PCoA) accounted for 88% of variation in regional functional space (PC1 = 38%, PC2 = 21%, PC3 = 17 and PC4 = 12%). The Tropical Northwestern Atlantic province, located in the southern Caribbean, had both the highest species richness (n = 134) and functional diversity (FRic = 0.94). The functional structure found in this province, and in the Tropical Southwestern Atlantic, corresponds to ~90% of the regional functional space-that of the totality of species considered in our study. In contrast, Benguela, located in southern Africa, presented the lower species richness (n = 31) and functional volume (40.6% of the regional functional space; see Figure S3). The distribution of threatened species in functional space (n = 63, 60.1% of  Figure S4). ) and the light-orange polygon represents non-threatened species (92.5%) relative to the regional functional space (dashed line). CENTER: the volume filled by different taxonomic groups is shown in different coloured polygons: mammals (dark pink), sea turtles (green, 0.0156%), Elasmobranchii (light blue, 29.3%) and Teleostei (dark blue, 53.1%). BOTTOM: ecosystem functions of vertebrates are mapped into functional space: "Herbivory" (light green), "Top predators" (orange) and "Mesopredators" (pink). The solid line ellipse represents "Trophic regulation" and encompasses top predators, mesopredators and herbivores. The dotted ellipse represents "Coral reef resilience"; (b) the centroid of traits associated with ecosystem functions mapped in functional space; (c) RDA plot of the relationship between the trait categories and the ecosystem functions performed by vertebrates in Atlantic reefs R² = .87, p < .01, see Figure S5). Functional uniqueness displayed greater values in one grid cell of the Caribbean region but also in one cell of the Eastern Atlantic (e.g. Benguela). FUn and species richness were found to be negatively and strongly associated (R² = .74, p < .01; Figure S5), FUn showing the lowest values (i.e. lower level of functional redundancy) in species-poor assemblages ( Figure 3c).

| Functional diversity across Atlantic
Also, functional specialization showed a similar pattern, with higher value for one assemblage in the North-Eastern Caribbean region ( Figure 3d). FSpe and species richness were also found to be negatively related (R² = .63, p < .01; Figure S5).

| Influence of simulated extinctions on ecosystem functions
At the regional scale, the simulated removal of threatened and nonthreatened vertebrate species associated with specific ecosystem functions did not differ from the losses expected at random (Figure 4ac). The only exception was for mesopredators, whose simulated extinction shrunk 40% of the regional functional space. In this case, the loss of mesopredators followed by the random removal of vertebrate species was greater than expected by chance (Figure 4d). At the scale of marine regions ( Figure S6a-l), the removal of mesopredators also significantly impacts the functional space of the Caribbean and Southwestern Atlantic regions ( Figure S6j,k). In the Eastern Atlantic, herbivorous' species removal compromises almost 20% of functional richness ( Figure S6f). At the assemblage scale, Atlantic reefs might be largely compromised by the loss of mesopredators species, which in certain communities reaches up to 90% of functional loss. The largest proportion of species loss in reef communities through the removal of mesopredators was identified for the Caribbean (60%), followed by the Brazilian (50%) and the western African coasts (45%) (Figure 5a). The greatest impacts in functional space following the removal of mesopredators occurred in southern Caribbean reef communities (94%) and in the southern part of the Southwestern Atlantic (70%-90%), followed by northern Africa (75%-90% of functional loss; Figure 5b). The null model has revealed that the observed mesopredators' functional loss is higher than expected at random mainly for assemblages at the Eastern Atlantic, but it does not differ from random expectations in almost all assemblages of Caribbean reefs (Figure 5c). Sites with greater mesopredator species richness have higher functional redundancy-i.e. lower functional losses ( Figure S7).

| D ISCUSS I ON
We have found that it is the loss of mesopredator species that will severely compromise the functional structure of Atlantic Ocean reefs, from regional to local scales. This is possibly an outcome of the variety of traits found in mesopredators, which imply a broad distribution in functional space, high functional richness and greater functional losses when extinct. Despite the recognized importance of large-bodied sharks as top predators in marine ecosystems Heithaus et al., 2008), their (simulated) extinction did not affect the functional diversity of Atlantic reef assemblages as predicted. At the global scale, mammals, sharks, rays and bony fishes together occupy most of the functional space of the marine megafauna . At this scale, the potential extinction of threatened elasmobranchs will cause the most severe changes in functional richness and uniqueness .
Our analysis revealed that the considered traits were consistently associated with particular ecosystem functions and that marine mammals, sea turtles and fish species in Atlantic Ocean reefs shared numerous functions. The importance of these four taxonomic groups and their functions in reef ecosystems have been reported and categorized Hammerschlag et al., 2019;Luiz et al., 2019;Pimiento, Leprieur, et al., 2020;Tavares et al., 2019;Villéger et al., 2017), but there have been few attempts to identify similarities in functions delivered by species belonging to such distant lineages, in a cross-taxa approach (but see Pimiento, Leprieur, et al., 2020).
When we simulated the loss of threatened vertebrate species (n = 63; 28.12%), we found that such loss may compromise the regional functional space (60.1%). Most of these threatened species share a macrocarnivore diet and large body sizes, which have been investigated in previous studies as predictors of extinction risk for fish and marine mammal species (Bender et al., 2013;Ceretta et al., 2020;Dulvy & Reynolds, 2002;Dulvy et al., 2003). Sharks are the majority of top predator species in Atlantic reefs (87.5%) and are largely threatened (46%). In the Atlantic, these species suffer from impacts as bycatch (Oliver et al., 2015), trade and food consumption (Barreto et al., 2017) and overexploitation (Luiz & Edwards, 2011).
Further, sharks share life-history traits, as slow growth and late maturity, which increase their vulnerability and hamper stock recovery (Dulvy et al., 2014;Stevens et al., 2000). Despite the important role of sharks on ecosystems, the redundancy among top predators is questionable since most shark species are now considered mesopredators given reductions of their size and body mass (Roff et al., 2016). This result reinforces threats to marine mesopredators since the scarcity of top predators makes this functional group the next to be depleted from marine food webs (Ferretti et al., 2010;Myers et al., 2007;Roff et al., 2016).
Contrary to our hypothesis, our results show that despite the low herbivore species richness (n = 36), trait diversity in functional space (and functional uniqueness) ensures herbivory in the Caribbean and Southwestern Atlantic. In these reefs, herbivory is mainly performed by bony fishes through diverse feeding modes as browsing, excavating or scraping on the reef substrate (Ferreira & Gonçalves, 2006;Francini-Filho et al., 2010;Mantyka & Bellwood, 2007). Yet, different herbivorous species, as bony fishes, marine turtles and mammals may target distinct algal resources (Cardona et al., 2020;Castelblanco-Martínez et al., 2009;Tebbett et al., 2020). For example, the herbivory pressure exerted by a single Chelonia mydas equals many fish individuals (Goatley et al., 2012), and green turtles have a major contribution in sheltered reefs with low rugosity, low coral cover and high algal cover (Cardona et al., 2020). Unfortunately, this essential reef function delivered by threatened herbivorous reptiles and mammals, as Chelonia mydas and Trichechus spp., is at global risk (Atwood et al., 2020). Thus, functional complementarity is needed to maintain ecosystem functioning (Cardona et al., 2020).
As expected, being the centre of marine biodiversity in the Atlantic Ocean (Bellwood et al., 2004;Floeter et al., 2008), the Caribbean is a "hot spot" of reef vertebrate species richness and functional richness.
Such uniqueness is possibly caused by the local presence of different taxonomic groups, which are expected to be more distinct physically and functionally .
Despite the similar patterns of functional uniqueness and specialization of vertebrates in Atlantic Ocean reefs, the loss of mesopredators will modify the functional structure of assemblages, especially those with lower species richness. The importance of mesopredators to reef functioning is well known (Roff et al., 2016). Their removal from reef ecosystems can alter not only patterns of nutrient cycling between reef habitats but also the behaviour of prey species (McCauley et al., 2012;Rizzari et al., 2014).
Furthermore, there is evidence for limited redundancy between small sharks and bony fish based on maximum prey size and gape width analysis, suggesting the unique role of these reef sharks as mesopredators (Barley et al., 2020). In the Caribbean, the most diverse vertebrate reef assemblage in the Atlantic, changes in reefs have been associated with the absence of mesopredator species as groupers and sharks, caused by fishing pressure, habitat degradation and pollution (Cheung et al., 2010;Ward-Paige et al., 2010). In the Southwestern Atlantic coast, groupers have suffered marked population declines in recent decades Zapelini et al., 2019). Overall, the rarity and limited distribution of sharks and large-bodied groupers to few reefs suggests that their ecological function as mesopredators may be compromised in the Brazilian province (Morais et al., 2017).
Reef ecosystems are home to a fascinating diversity of spe- reinforcing that these species deserve attention in future conservation planning. Our study does not provide a mechanistic understanding between traits and functions but evaluates whether some well-known functions in reef ecosystems can be related to particular or a set of traits. In that context, more experimental and fieldwork are needed to evaluate how species traits affect ecosystem functioning and to define the traits that are more strongly related to a "realized" function. As demonstrated in our analysis, studies on the ecosystem functions of marine species offer great opportunities to improve the roadmap for saving reefs from future degradation .

ACK N OWLED G EM ENTS
We thank the many volunteers who helped compile information and

CO N FLI C T O F I NTE R E S T
The authors do not have a conflict of interest statement.

PEER R E V I E W
The peer review history for this article is available at https://publo ns.com/publo n/10.1111/ddi.13430.

DATA AVA I L A B I L I T Y S TAT E M E N T
Data available at: https://github.com/luiza waech ter/Atlan tic-marin e-verte brate -species.