Introduction
Traditionally, identification of coral species is based on skeletal macrostructures, mainly on the characteristics attributed to the corallum and corallites, being the interspecific limits defined by the discontinuities of the diagnostic traits (Foster, Reference Foster1977; Budd and Stolarski, Reference Budd and Stolarski2009; Benzoni et al., Reference Benzoni, Arrigoni, Stefani and Pichon2011; Budd et al., Reference Budd, Fukami, Smith and Knowlton2012a). However, challenging the Scleractinia taxonomy, morphology may be quite inconsistent, varying intra- and interspecifically, leading to the overlapping of skeletal features and the misidentification of the morphotypes (Brown and Navin, Reference Brown and Navin1992; Veron, Reference Veron1995; Bruno and Edmunds, Reference Bruno and Edmunds1997; Benzoni et al., Reference Benzoni, Arrigoni, Stefani and Pichon2011; Menezes et al., Reference Menezes, Neves, Barros, Kikuchi and Johnsson2013).
Incongruences on coral morphologies have been highlighted by molecular phylogenies, emphasizing the relevance of an integrative approach (Romano and Cairns, Reference Romano and Cairns2000; Budd and Stolarski, Reference Budd and Stolarski2009; Benzoni et al., Reference Benzoni, Arrigoni, Stefani and Pichon2011; Arrigoni et al., Reference Arrigoni, Stefani, Pichon, Galli and Benzoni2012; Budd et al., Reference Budd, Fukami, Smith and Knowlton2012a, Reference Budd, Romano, Smith and Barbeitos2012b). These studies have impacted the phylogeny of two important families from the Atlantic and Indo-Pacific Oceans: Faviidae and Mussidae (Fukami et al., Reference Fukami, Budd, Paulay, Solé-Cava, Chen, Iwao and Knowlton2004, Reference Fukami, Chen, Budd, Collins, Wallace, Chuang, Chen, Dai, Iwao, Sheppard, Knowlton and Ahmed2008; Budd et al., Reference Budd, Romano, Smith and Barbeitos2012b). Using nuclear and mitochondrial markers, Nunes et al. (Reference Nunes, Fukami, Vollmer, Norris and Knowlton2008) have suggested a close relationship between Favia leptophylla Verrill, Reference Verrill1868 and the Brazilian endemic genus Mussismilia Ortmann, 1890, the species renaming as Mussismilia leptophylla.
Following this novelty, Budd and Stolarski (Reference Budd and Stolarski2009) presented a thorough skeletal analysis into three distinct ‘scales’: macromorphology, micromorphology, and microstructure, defining the family Mussidae as polyphyletic, with most members distributed into two major molecular clades, one Atlantic, including Mussidae and Faviidae, and the other represented by Pacific Mussidae and Pectiniidae.
Supported by recent literature, Mussidae is expected to comprehend two distinct subfamilies: Mussinae Ortmann, 1890 and Faviinae Gregory, 1890 – among other genera, the latter being represented by the Atlantic members of Favia Milne Edwards, 1857 and Mussismilia (Fukami et al., Reference Fukami, Budd, Paulay, Solé-Cava, Chen, Iwao and Knowlton2004; Budd et al., Reference Budd, Fukami, Smith and Knowlton2012a, Reference Budd, Romano, Smith and Barbeitos2012b), while Scolymia Haime, 1852 remained in Mussinae. By its turn, Faviidae has been characterized as polyphyletic within the Robust clade – although closely related to Mussidae, it has been ‘lowered to the rank of subfamily’ (Budd et al., Reference Budd, Romano, Smith and Barbeitos2012b, p. 472). In fact, apparent similarities between Atlantic and Indo-Pacific faviids have been attributed to morphological convergence (Fukami et al., Reference Fukami, Budd, Paulay, Solé-Cava, Chen, Iwao and Knowlton2004).
Mussismilia fossils from Mio-Pliocene have been found across the Caribbean and the Mediterranean Sea (Veron, Reference Veron1995; Riegl and Piller, Reference Riegl and Piller2000). Currently endemic to Southwestern Atlantic, modern Mussismilia corals (the ‘brain corals’) present a wide distribution range along the coastal Brazilian shallow water environments, occurring from Maranhão to São Paulo State (0°S/44°W to 23°S/45°W). Although represented by a few species, the genus plays an important role as reef builders. The massive Mussismilia braziliensis (Verrill, Reference Verrill1868), for instance, stands out as the major builder of the mushroom-shaped coral pinnacles, the ‘Chapeirões’ from Abrolhos reefs at Southern Bahia (16°S/45°W) (Castro and Pires, Reference Castro and Pires2001). True coral reefs in the South Atlantic occur mostly along the northeastern section of the Brazilian coast, where the influence of the warm waters of the Guyana and Brazilian currents supports favourable conditions for the carbonate deposition, allowing reef flourishment (Maida and Ferreira, Reference Maida and Ferreira1997).
For long unchangeable, and represented by three congeners, Mussismilia genus was modified in the last decade by a new addition: M. leptophylla (Nunes et al., Reference Nunes, Fukami, Vollmer, Norris and Knowlton2008; Budd et al., Reference Budd, Fukami, Smith and Knowlton2012a, Reference Budd, Romano, Smith and Barbeitos2012b). Originally, the subplocoid M. braziliensis (Verrill, Reference Verrill1868) and Mussismilia hispida (Verrill, Reference Verrill1902), and the phaceloid Mussismilia harttii (Verrill, Reference Verrill1868) were described as Acanthastrea Milne Edwards & Haime, 1848 (M. braziliensis) and Mussa (Oken, 1815) (M. harttii and M. hispida) (Verrill, Reference Verrill1868, Reference Verrill1901, Reference Verrill1902). Being distinguishable from other genera because of exclusive micromorphological characteristics of the septa teeth, Mussismilia species have been separated by the colony form (and pattern of corallite development), corallite wall, calice size, septal thickness, and number of septa per cycle (Budd et al., Reference Budd, Fukami, Smith and Knowlton2012a, Reference Budd, Romano, Smith and Barbeitos2012b). According to Budd and Stolarski (Reference Budd and Stolarski2009) and Budd et al. (Reference Budd, Fukami, Smith and Knowlton2012a), major micromorphological aspects (e.g. septa marginal teeth, ornamentation of the septal face) have also shed light on the controversies related to the skeleton interspecific variation, and have become an important tool for consolidating the genus taxonomy.
Regarding the species currently accepted as Mussimilia, M. harttii singularly displays a phaceloid development (‘corallites walls of adjacent corallites separated by void space; each corallite forms a branch’, sensu Budd et al., Reference Budd, Fukami, Smith and Knowlton2012a, p. 481, Reference Budd, Romano, Smith and Barbeitos2012b). Without tissue lying between the corallites, ‘true’ phaceloids may be promptly recognized by the widely spaced units, lacking costa, exotheca, coenosteum, and/or coenosarc. However, even pointed out as a strong diagnostical character, a few colonies may show some degree of fusion between adjacent polyps, forming closely spaced corallites with deposition of exothecal elements. Considering Verrill's morphotypes, phaceloid variants were described as follows: (1) conferta with corallites joined by a vesicular exotheca (very similar to M. hispida); (2) laxa with dichotomous corallum without exotheca (typical phaceloid) (3); intermedia with partially free corallites joined by an exotheca at the base, and finally (4) confertifolia with corallites separated by deep grooves at short distances (Table 1).
Characterization of the Mussismilia species, including morphotypes, macro- and microstructures with special inclusion of Mussismilia leptophylla
Based on Amaral (Reference Amaral1992), Budd et al. (Reference Budd, Fukami, Smith and Knowlton2012a), Verrill (Reference Verrill1868, Reference Verrill1901, Reference Verrill1902), and data from the present study.
Most M. harttii morphotypes may be observed across their entire range of geographical distribution. However, studies by Laborel (Reference Laborel1969/70) have not supported ‘conferta’, probably because the huge development of the exothecal elements – an unusual pattern, which bias the recognition of the species in the field, leading M. harttii be misidentified as M. hispida.
Two geographic subspecies have been described for M. hispida (Laborel, Reference Laborel1969/70): M. hispida tenuisepta and M. hispida hispida. The former is distributed northwards São Francisco River (10°S), and the latter occurring southwards (Laborel, Reference Laborel1969/70). Contrasting with M. hispida hispida, Verrill (Reference Verrill1901) has originally classified Mussa (Symphyllia) tenuisepta as those colonies with moderately broad, irregular calyces, polystomodeal, with numerous and thinner septa. As similarly proposed by Amaral et al. (Reference Amaral, Ramos, Leão, Kikuchi, Lima, Longo, Cordeiro, Lira and Vasconcelos2009), colonies with meandroid corallites fit in the 'hispida tenuisepta' pattern, while colonies with regular corallites are 'hispida hispida' (Table 1).
Anthropogenic impacts on the Mussismilia corals have been the focus of intense debate, due to the overall risks of the productivity impoverishment of the coastal seas, and the biodiversity losses along the Tropical South Atlantic (Dutra et al., Reference Dutra, Kikuchi and Leão2006; Francini-Filho et al., Reference Francini-Filho, Moura, Thompson, Reis, Kaufman, Kikuchi and Leão2008; Miranda et al., Reference Miranda, Cruz and Leão2013). Despite major worldwide concern on global climate changes, several other factors (e.g. chemical, biological and solid pollution, deforestation, urbanization, unsustainable exploration of the natural resources) are rapidly depleting local coral communities (Leão et al., Reference Leão, Kikuchi, Ferreira, Neves, Sovierzoski, Oliveira, Maida, Correia and Jonhsson2016; Kubicek et al., Reference Kubicek, Breckling, Hoegh-Guldberg and Reuter2019). Thus, baseline surveys represent an unprecedented strategy to support reef resilience and its conservation. Alternatively, studies dealing with natural morphological variation within and among populations may also provide answers to how a changing world is affecting the species adaptation and the ecological interactions as well.
Therefore, intending to corroborate the definition of the interspecific limits and the species identities as well, the present study aims to analyse qualitatively and quantitatively the macro- and microstructures of the endemic brain corals M. hispida, M. harttii, and M. braziliensis through a population approach.
Materials and methods
The Bahia State has one of the longest coastlines of Brazil (~1100 km), located between 11°27′26.70″S and 18°20′9.35″S of latitude, being characterized by pristine natural environments, including rocky shores, coral reefs, sandy beaches, mangroves, and estuaries (Tessler and Goya, Reference Tessler and Goya2005). The study area comprises two distinct geographical sections: the Todos-os-Santos Bay and the South Littoral (SL), including the following true reefs: Caramuanas, Boa Viagem, Moreré, and Abrolhos Archipelago (Figure 1). Sampling was carried out by snorkeling and scuba diving at 1.0–5.0 m depth. Colonies of Mussismilia hispida, Mussismilia braziliensis, and Mussismilia harttii ranging from 10.0 to 30.0 cm were randomly selected and removed using a hammer and a chisel. A minimum distance of 3.0 m between neighbouring colonies was adopted to cover a greater morphological diversity. In the laboratory, the corals were bleached in a solution of sodium hypochlorite (2%), washed, and dried at ambient temperature. Samples from the Abrolhos Archipelago were donated by the ‘Laboratório de Recifes de Corais e Mudanças Globais’ (RECOR/IGEO/UFBA). Testimonies previously deposited in the Cnidaria Collection of the Museu de História Natural da Universidade Federal da Bahia (acronym: UFBA) were also examined. All collected samples were deposited in the UFBA.
Map of the study area and sampling sites, including Boa Viagem reef in the Todos-os-Santos Bay, and reefs from the South Littoral: Caramuanas, Moreré (Boipeba Island), and Abrolhos Archipelago.
Taxonomy and morphometric analysis
Skeleton structures, including corallum and corallites, were qualitatively and quantitatively evaluated (Table 1). Species identification and morphometric variables were supported by the specialized literature (Verrill, Reference Verrill1868, Reference Verrill1901, Reference Verrill1902; Laborel, Reference Laborel1969/70; Foster, Reference Foster1977, Reference Foster1979; Neves, Reference Neves2004; Budd et al., Reference Budd, Fukami, Smith and Knowlton2012a). Morphometric measurements were developed under a Nikon SMZ 1000 stereomicroscope with an eyepiece micrometre and a Nikon Coolpix 995 digital camera attached. Skeleton fragments were also mounted with aluminium pin stubs for septal teeth and spine analysis, being previously covered with a double-sided sticky tape, sputter-coated with 35 nm of gold in a Denton Vacuum Desk V ion coater, and examined through a Jeol JSAA-6610LV. Scanning electron microscope (SEM) images supported the microstructure analysis.
Statistics
A total of six corallites per colony were randomly selected and examined using a grid of numbers (Foster, Reference Foster1985). Colony diameter (D col = mean of the two major axes of the colony [mm]), which is influenced by the age, and meandroid and/or polystomodeal corallites (irregular patterns) were not considered; for the statistical analysis we included: N cor = corallite number in an area of 5 cm2; D ica = inner diameter of callice, mean of the two major inner axes of theca cavity margins (mm), D icor = corallite diameter, mean of the two major axes of the outer margins of corallite (mm); N sep = number of septal elements; C sep = mean length from septa first cycle to columella (mm); P col = mean columellar fossa depth (mm), D cor = mean distance between calice, based on the inner theca margin between two close corallites (mm). Due to heteroscedasticity and the distinct number of samples per population, a non-parametric Kruskal–Wallis test (Theodorsson-Norheim, Reference Theodorsson-Norheim1986) was developed as well to evaluate which characteristics differed significantly among the populations and species. To visualize all traits together per locality and per species, a principal component analysis (PCA) was adopted (Jolliffe, Reference Jolliffe2002). For interspecific analysis, only populations where the three species occurred sympatrically were used (i.e. Caramuanas and Moreré reefs). Packages of Vegan (Oksanen et al., Reference Oksanen, Blanchet, Friendly, Kindt, Legendre, Mcglinn, Minchin, O'hara, Simpson, Solymos, Stevens, Szoecs and Wagner2017) and ggplot2 (Wickham, Reference Wickham2009) in R environment, and Excel 2010 (Microsoft©) were used for the statistical analysis.
Results
A total of 132 corallums was examined (Mussismilia hispida n = 53; Mussismilia harttii n = 41; Mussismilia braziliensis n = 38), comprising 792 corallites, and resulting in 5544 data analysed; all means, standard deviation, and total range are summarized in Table 2. In M. hispida colonies, only the number of septa showed statistical difference among populations with higher values in Moreré population (Tables 2 and 3). On the other hand, the marginal teeth and the spines did not show interpopulational variation. Septal ornamentation is composed by conical and elongated spines. Bi- or trifurcated conical spines are regularly distributed along the septal faces (Figure 2A, E, I), while those elongate, with regular tips (not bifurcated), are distributed linearly along the septa (Figure 2B, F).
Mean, standard deviation (SD), and range of M. hispida, M. harttii, and M. braziliensis morphological characters measured in the present study
N cor, corallite number in an area of 5 cm2; D ica, inner diameter of callice, mean of the two major inner axes of theca cavity margins (mm), D icor, corallite diameter, mean of the two major axes of the outer margins of corallite (mm); N sep, number of septal elements; C sep, mean length from septa first cycle to columella (mm); P col, mean columellar fossa depth (mm); D cor, mean distance between calice, based on the inner theca margin between two close corallites (mm).
Results of Kruskal–Wallis test among the populations of each Mussismilia species
N cor, corallite number in an area of 5 cm2; D ica, inner diameter of callice, mean of the two major inner axes of theca cavity margins (mm), D icor, corallite diameter, mean of the two major axes of the outer margins of corallite (mm); N sep, number of septal elements; C sep, mean length from septa first cycle to columella (mm); P col, mean columellar fossa depth (mm); D cor, mean distance between calice, based on the inner theca margin between two close corallites (mm).
SEM images of micromorphological characters of Mussismilia hispida. Variation of septal teeth and spines in colonies from three reefs: Caramuanas (A–C), Moreré (D, E), and Boa Viagem (F). Septal teeth, bi/trifurcated, and distribution of spines along the septal face and margins (G, H). Details of septal spine (I).
For M. harttii colonies, corallite diameter, number of septal elements, mean length from septa first cycle to columella, and mean distance between calice were statistically different among populations with higher values for colonies from Moreré; corallite number and mean distance between calice showed higher values in the Caramuanas population (Tables 2 and 3). Contrasting with M. hispida, teeth and septal spines were highly inconspicuous among the populations (Figure 3). On the other hand, M. braziliensis showed significant difference among the populations for the number of septal elements (higher in the Moreré population) and the mean columellar fossa depth (mm) (higher values in the Abrolhos population) (Table 3). As supported by SEM images, septal spines vary among the populations (Figure 4). In the Moreré population, spines are conical with rounded edges, forming a perpendicular line along the septal teeth (Figure 4A, B). This line is a continuous deposition of aragonite fibres composed by two larger granules separated by a smaller one (Figure 4E). The spines from Abrolhos colonies are conic and wide at the base, being distally bi- or trifurcated (Figure 4B, E). This population has an exquisite sequence of three spines forming a slight curvature at an obtuse angle below the septal tooth. In the Caramuanas population, bi- or trifurcated conical spines are intercalated and sparsely distributed along the lateral septum surface (Figure 4D, F). Similarly, the spines of septal teeth are multidirectional and bifurcated, being irregularly arranged (Figure 4G, H).
SEM images of micromorphological characters of Mussismilia harttii. Variation of septal teeth and spines in colonies from two reefs. Moreré (A, B) and Caramuanas (C, D). Distribution of spines on the septa face and margins (E).
SEM images of micromorphological characters of Mussismilia braziliensis. Variation of septal teeth and spines in colonies from three reefs. Moreré (A, B), Abrolhos Archipelago (C, D), and Caramuanas (E, F). Septal teeth and distribution of spines on the septa face and margins (G, H).
Comparing M. hispida, M. harttii, and M. braziliensis, significant differences were observed in the non-parametric analysis for all the analysed characters (N cor: P = 0.0000; D ica: P = 0.0000; D icor: P = 0.0000; N sep: P = 0.0000; C sep: P = 0.0000; P col: P = 0.0000; D cor: P = 0.0000). The PCA shows 76.10% in axis 1 and 12.21% in axis 2 (Figure 5). The results showed differentiation between the species; colonies of M. braziliensis from both areas showed a distinct group from other species populations. Mussismilia harttii populations from Caramuanas and Moreré showed a trend in cluster in different groups, suggesting that for this species, the environment may play an important role in the morphological pattern of colonies. There is an intersection between individuals of M. harttii from Moreré with M. hispida; the overlapping of the characteristics in M. harttii is strongly marked in the variant ‘conferta’ in which the phaceloid development is feebly distinguishable (Figure 5). In the PCA, colonies of M. harttii from Moreré showed high convergence with M. hispida morphology. In contrast, the microstructures are conspicuously different, being a valuable tool for defining the interspecific limits (Figure 3). The septal spines in M. hispida and M. harttii occur in greater densities than those in M. braziliensis. The spines are slender in M. hispida; curved on the top in M. harttii; and thicker, bi- and trifurcated, with a grainy appearance in M. braziliensis.
Graph and result of PCA with morphological variables of the species of M. hispida (MI.C; MI.BP), M. harttii (MH.C; MH.BP), and M. braziliensis (MB.C; MB.BP). Distribution of characters in the axes and factors that most influenced the analysis (MI.C, Caramuanas; MI.BP, Moreré; MH.C, Caramuanas; MH.BP, Moreré; MB.C, Caramuanas; MB.BP, Moreré).
Discussion
The results obtained in this study corroborate previous studies regarding the establishment of morphotypes within the Mussismilia genus, with strong evidence of macro- and micromorphological variation, majorly in Mussismilia hispida and Mussismilia harttii (Verrill, Reference Verrill1868, Reference Verrill1901; Laborel, 1967, Reference Laborel1969/70). As observed, Mussismilia species are well-defined, and the morphological patterns may be attributed to natural variation and the influence of fine-scale environmental elements, as well. Moreover, Nogueira et al. (Reference Nogueira, Neves and Johnsson2015) demonstrate that variable morphologies of Mussismilia related to the structural complexity of the colony may act on the richness and abundance of the associated fauna. This makes the approach also important for understanding the composition and structure of the local benthic community – indicating that studies dealing with coral morphologies may have strong implications on the knowledge of the biodiversity.
Before the present study, data dealing with morphological variation among Brazilian corals were restricted to four species (all of them also distributed in the Caribbean): Siderastrea stellata, Siderastrea radians, Favia gravida, and Montastraea cavernosa (Amaral, Reference Amaral1994; Neves, Reference Neves2004; Santos et al., Reference Santos, Amaral, Hernández, Knowlton and Jara2004; Amaral and Ramos, Reference Amaral and Ramos2007; Menezes et al., Reference Menezes, Neves, Barros, Kikuchi and Johnsson2013). The results obtained with the endemic Mussismilia support the relevance of the corallite analysis and the micromorphology for the scleractinian taxonomy.
The close congeners, S. stellata and S. radians, are brooding corals with panmictic populations that occur sympatrically along the Brazilian coast – the gene flow across long geographical distances has been also attributed to levels of intraspecific variation in the species (Neves et al., Reference Neves, Andrade, Lang and Solferini2008). Indeed, the morphotypes of S. stellata and S. radians show a high overlap of diagnostic characteristics, being a source of misinterpretations, and thus challenging researchers worldwide. However, as seen to Mussismilia, micromorphology of the theca and radial structures has also proved to be an essential tool for Siderastrea taxonomy (Neves, Reference Neves2004).
The results indicate that skeleton variation occurs on two scales in Mussismilia: macro and micromorphological. According to the literature, teeth and septal spines are diagnostic characters for Mussidae corals (Neves et al., Reference Neves, Johnsson, Sampaio and Pichon2006; Budd and Stolarski, Reference Budd and Stolarski2009; Budd et al., Reference Budd, Fukami, Smith and Knowlton2012a, Reference Budd, Romano, Smith and Barbeitos2012b). Indeed, our SEM images reveal that the characteristics and arrangement of microstructures differ interspecifically, supporting the species' identities.
Although consistent differences in the macromorphology were not statically supported among localities (except for two characters: diameter of corallites and the number of septa), all the four morphs described by Verrill (Reference Verrill1901) for M. harttii were found and analysed in Bahia State – namely, ‘conferta’, ‘confertifolia’, ‘laxa’, and ‘intermedia’. It is irrefutable that spaced corallites with phaceloid development represent a natural tendency of this species. However, some degree of morphological overlapping may occur between the morph ‘conferta’ and M. hispida, particularly because of the fusion of the adjacent exotheca (joining the nearby corallites superficially). Only by checking the lower development zone of the corallites, the phaceloid development of M. harttii is evident. Hence, the morph ‘conferta’ is not easily distinguishable from M. hispida in situ through visual identification of the colony top (which makes the method of identification based on images from video transects unreliable).
In this study, the morphotypes of M. harttii and M. hispida, occurred sympatrically in three reefs Caramuanas, Moreré, and Boa Viagem, and biotic and abiotic factors are expected to vary in some scale at each locality. Similarly, Amaral and Ramos (Reference Amaral and Ramos2007) found distinguishable morphological patterns in Favia corals at two different spatial scales, within different environments of a single reef and between distinct reefs. Additionally, budding may also act as a trigger for morphological divergences, as observed in the morphs ‘hispida hispida’ (with regular corallites) and ‘hispida tenuisepta’ (with irregular corallites) (Amaral et al., Reference Amaral, Ramos, Leão, Kikuchi, Lima, Longo, Cordeiro, Lira and Vasconcelos2009).
The development of new polyps in coral colonies (extracalicinal vs intracalicinal budding) is an important taxonomical attribute, being the intracalicinal budding advantageous for colonial growth, once adding new mature elements it may improve the success of reproductive activity. Mixed budding patterns in a single species have been observed in S. stellata and F. gravida (Neves, Reference Neves2004, authors' personal communication). However, the development of Mussismilia colonies has been strictly attributed to the intracalicinal budding (Laborel, Reference Laborel1969/70), with irregular polyps (or ‘lobed’) of M. hispida evidence of the poliestomodeal pattern. Indeed, mixed budding is uncommon among most scleractinian corals, and extracalicinal budding is unknown for M. hispida and M. harttii. Therefore, a new colonial development strategy is here reported for the genus.
Finally, the characters selected in this study were consistent with the definition of the interspecific limits of Mussismilia corals, ensuring Verrill's morphotypes, and the relevance of the traditional skeleton analysis for the taxonomy of reef-building genera.
Acknowledgements
We thank the LABIMAR team for outstanding help in the field; the LAMUME/Phisics Institute (Universidade Federal da Bahia) by the SEM support; Saulo Freitas for the maps; the RECOR/UFBA for the Mussismilia colonies from Abrolhos. This study is part of the project ‘Assessment and research of sun corals in the Todos-os-Santos Bay’, a cooperation agreement between UFBA and CENPES/PETROBRAS (grant number: 5850.0107361.18.9).
