The earliest diverging extant scleractinian corals recovered by mitochondrial genomes

Evolutionary reconstructions of scleractinian corals have a discrepant proportion of zooxanthellate reef-building species in relation to their azooxanthellate deep-sea counterparts. In particular, the earliest diverging “Basal” lineage remains poorly studied compared to “Robust” and “Complex” corals. The lack of data from corals other than reef-building species impairs a broader understanding of scleractinian evolution. Here, based on complete mitogenomes, the early onset of azooxanthellate corals is explored focusing on one of the most morphologically distinct families, Micrabaciidae. Sequenced on both Illumina and Sanger platforms, mitogenomes of four micrabaciids range from 19,048 to 19,542 bp and have gene content and order similar to the majority of scleractinians. Phylogenies containing all mitochondrial genes confirm the monophyly of Micrabaciidae as a sister group to the rest of Scleractinia. This topology not only corroborates the hypothesis of a solitary and azooxanthellate ancestor for the order, but also agrees with the unique skeletal microstructure previously found in the family. Moreover, the early-diverging position of micrabaciids followed by gardineriids reinforces the previously observed macromorphological similarities between micrabaciids and Corallimorpharia as well as its microstructural differences with Gardineriidae. The fact that both families share features with family Kilbuchophylliidae ultimately points towards a Middle Ordovician origin for Scleractinia.

Scleractinian corals are renowned for their capacity to create spectacular shallow-water calcium carbonate reef structures. Azooxanthellate scleractinians-corals that do not establish a symbiotic relationship with dinoflagellates of the family Symbiodiniaceae 1,2 -total approximately the same number of extant zooxanthellate species of the order, but are not restricted to tropical, shallow-waters as the latter. The early evolutionary history of the order Scleractinia has been the subject of intense scientific debate. For example, the monophyly of this order has been challenged since the order Corallimorpharia was previously recovered as a clade nested within the main scleractinian lineages ("naked coral" hypothesis 3 ). Nevertheless, it has been shown that such a corallimorpharian position was an artefact of the use of amino acid sequences in phylogenetic analyses 4 . Furthermore, the evolutionary history of the main reef builders has also attracted great attention as the "molecular revolution" challenged the long-established morphological systematics especially for the higher taxonomic ranks [5][6][7][8][9][10][11] . Originally, based on gross morphology, the order Scleractinia was divided into five 12 , or thirteen 13 suborders. However, more recently, molecular data pointed to three main clades: "Basal"; "Complex"; and "Robust" corals 4,11,[14][15][16] .
In contrast to widely accepted Triassic emergence of Scleractinia, divergence of the earliest scleractinian clade with extant representatives (families Micrabaciidae and Gardineriidae) was suggested to have occurred between the Ordovician and Silurian, around 425 million years ago (mya) 14 , or in the Silurian, 407 mya 16 . Although recovering a slightly later onset, a recent study by Quattrini and collaborators 17 has also pointed to a Paleozoic origin for the order. Composed of exclusively azooxanthellate taxa 18 , representatives of Micrabaciidae share some morphological skeletal characters (septal bifurcations) with the Ordovician Kilbuchophylliidae (~ 460 mya), but otherwise represent a morphologically unique coral group 19 . On the other hand, gardineriids develop usually a thick, exclusively epithecal wall 20 typical of some of the oldest known solitary Mesozoic (Middle Triassic) corals 14 (~ 230 mya; Fig. 1a-c,f,g). Gardineriidae is also composed exclusively of azooxanthellate solitary corals, occurring  (Moseley, 1876) in distal, basal and lateral views, respectively), gardineriids have very robust coralla (f-g, Gardineria hawaiiensis Vaughan, 1907) in distal and lateral views, respectively). In micrabaciids (here L. formosissima) Thickening Deposits (TDs) are composed of an irregular meshwork of fiber bundles oriented sub-parallel to the skeleton surface (d,e), whereas in gardineriids (G. hawaiiensis) TDs are arranged in small bundles of fibers oriented approximately perpendicular to the skeleton surface. Consequently, micrabaciid TDs show variable crystallographic orientation (d, seen as lack of larger areas of similar vivid interference colors in polarized light), whereas in gardineriids TDs are crystallographically ordered and larger areas of similar vivid interference colors are visible in polarized light (h). Thin-sections in polarized microscope views (d,h), and polished and lightly etched sections in Scanning Electron Microscopy views (e,i).

Results
Generated raw reads ranged from 3,037,202 to 5,605,634 (MiSeq run) and corresponded to 26 Supplementary Table S1 online), although that from L. formosissima lacks a part of the 16S ribosomal gene (probably around ~ 30 bp; Table 2). Sanger data from Letepsammia franki was assembled into five contigs comprising 18,206 bp in total and the only absent gene was trnW. Micrabaciidae mitogenomes have 37.8-37.9% GC levels, values slightly lower than that observed for G. hawaiiensis and Corallimorpharia (both with 39.7%). Nevertheless, "Complex" corals include a wide range of values, from 36.2% in Porites lobata to 40.5% in Pavona clavus (see Table 1).

Discussion
The phylogenetic analysis presented here indicates a paraphyly of the previously thought earliest diverging scleractinian "Basal" group 11,[14][15][16] , represented by the two families Micrabaciidae and Gardineriidae. Based on all mitochondrial genes, the recovered phylogeny suggests that the micrabaciids were the first to diverge, being a sister group to all other scleractinians, including gardineriids. These results add further evidence for the hypothesis of solitary azooxanthellate corals as origin for the group.  www.nature.com/scientificreports/ Despite its crucial relevance for improving our understanding of the deep evolutionary patterns in Scleractinia, phylogenetic analyses have been based on limited data from a few partial genes (majoritarily cox1, 16S rDNA, 18S rDNA and 28S rDNA [14][15][16] ). Only recently, using 933 loci (278,819 bp) captured with a targetedenrichment approach, Quattrini et al. 17 recovered the family Micrabaciidae as the first diverging lineage within "Robust" corals. Nevertheless, they lack representatives of Gardineriidae or other exclusively deep-water azooxanthellate families, such as Deltocyathidae and Anthemiphylliidae. On one hand, Anthozoa mitochondrial genes tend to evolve at slower rates in comparison with other metazoans 47,48 , thus favouring their application to investigate Scleractinia relationships through deep time or among taxonomic ranks above genus. On the other hand, mitogenomes may be genetically saturated, making them problematic for studying deeper levels of relationship inside Cnidaria (e.g. class level 44 ). Therefore, sequencing of more mitogenomes and tests for saturation will elucidate the utility of mitochondrial genes in clarifying deep phylogenies within Scleractinia.
In general, micrabaciid mitogenomes have similar sizes to those of G. hawaiiensis and some "Complex" corals (e.g. Tubastraea coccinea, Dendrophyllia arbuscula, Fungiacyathus stephanus, and Pseudosiderastrea tayamai; Table 1). Among the mitogenomes sequenced herein, differences in length were mainly due to IGR sizes rather than among genes. A similar length variation pattern has previously been observed for the entire order and supports our findings for this family 35 . At 19,542 bp (Fig. 2), R. niphada holds the longest Scleractinia mitogenome known to date, followed by two Pseudosiderastrea species with 19,475 bp 49 . Micrabaciids were expected to bear long mitogenomes (> 19 kbp) based on their phylogenetic placement since the mitogenome of Gardineria hawaiiensis is 19,429 bp long 4 and the overall size of scleractinian mitogenomes appears to be shrinking in relation to Corallimorpharia (between 1 and 6 kbp) and also within the order (i.e. "Robust" corals have 2-3 kbp shorter mitogenomes than "Complex" corals). Nevertheless, a clear pattern is still not evident in the class Anthozoa as a whole. While octocorals seem to present a narrower range of variation regarding mitogenome size (18 to 19.8 kbp 50,51 ), hexacorals show a broader range. Actiniarian mitogenomes vary between 16 and 20 kbp, which is close to that observed in scleractinians 52,53 . Antipatharia ranges from 18 to 20 kbp [54][55][56] and Zoantharia includes 20-kbp long mitogenomes, similar to Corallimorpharia 57,58 . www.nature.com/scientificreports/ Different from size correlations, the recovered topology does not match previous reconstructions that included "Basal" lineages 11,14,15,59 (Fig. 3), with a few exceptions such as reconstructions based on 28S rDNA and mitochondrial 16S rDNA that recovered Micrabaciidae and Gardineriidae as a polytomy and a paraphyly, respectively 14 . Here, micrabaciids and gardineriids were recovered in a paraphyly, the former being the earliest to diverge. Notably, Micrabaciidae shares anatomical features in common with corallimorpharians, including the presence of a thick mesoglea 60 and acrospheres positioned around and above the polyps 61 . The molecular discrepancy between Gardineriidae and Micrabaciidae is strongly supported by differences in microstructural organization of their skeleton. For example, although the presence of an epithecal wall is common in many fossil and extant scleractinians, its occurrence as the only wall of the corallum, like in gardineriids, is exclusive in modern corals, but seems to have been a more common feature in early Mesozoic corals 20,62 . In contrast to the Gardineriidae skeletal thickening deposits (TD) (i.e. bundles of fibers arranged perpendicularly to the growing surfaces), micrabaciid TD are shaped in form of chip-like fiber bundles, sub-parallel to the skeletal surfaces, creating an irregular meshwork within the skeleton, which is not comparable with any microstructural organization from other modern or fossil scleractinian 14,19 . Because distinct patterns of TD organization are highly conservative traits in the evolution of scleractinian corals 63 , a unique micrabaciid fine-scale skeletal organization clearly suggests a long-period of independent evolutionary history in relation to gardineriids. On the other hand, although microstructural organization of Ordovician kilbuchophyllids is unknown (these fossils occur as moulds), these Paleozoic corals with scleractinian pattern of septal insertion had an epithecal wall (somewhat similar to gardineriids), and a pattern of bifurcations of higher septal cycles similar to micrabaciids. Together, these morphological characteristics allied to the molecular based phylogeny point towards a common and deep Paleozoic root for the order Scleractinia.
The early divergence of the azooxanthellate, solitary, deep-water micrabaciids and gardineriids (also supported by Stolarski et al. 14 and Kitahara et al. 15 ) contrasts with some hypotheses for whether first scleractinians were symbiotic and if they inhabited shallow or deep water environments 64 . In fact, Campoy and colleagues 16 used four markers (18S rDNA, 28S rDNA, 16S rDNA and cox1) and 513 scleractinian coral species from almost all extant families and hypothesized that the first scleractinian would have been azooxanthellate and solitary. Nonetheless, symbiosis with zooxanthellae was widespread in Triassic corals 65 and there is some degree of disagreement about it being lost and reappearing a few times 64,66 or being gained only once during scleractinian evolution 16 . In contrast, it appears that coloniality was the first one to be gained and there is an agreement that it was lost and gained more than once 16,64 and even the presence of multiple mouths in one polyp seems to be a labile trait in some families of this order as well (e.g. Dendrophylliidae 66 and Fungiidae 67,68 ).
Interestingly, all Paleozoic purported scleractiniamorph corals (i.e. Kilbuchophyllia, Houchangocyathus, and possibly Numidiaphyllum) were solitary polyps or had a loosely constructed phaceloid growth form (see Scrutton 69 ; Ezaki [70][71][72] ). Due to the several similarities with living solitary azooxanthellate scleractinians 69,72 , these aforementioned corals were purported to be azooxanthellate. Also, coral-zooxanthellae symbiosis has been established around 14 mya after the P/T boundary 65,73 , and although diagenetic conditions preclude unambiguous determination of a symbiotic or asymbiotic condition in Paleozoic corals, azooxanthellate and solitary lineages of living corals are the first to diverge in recent evolutionary reconstructions of the order Scleractinia. Such pattern is not limited to Micrabaciidae and Gardineriidae, but includes some lineages within "Complex" and "Robust" corals 11,14,15,74 . Examples are the families Anthemiphylliidae and Deltocyathidae in "Robust" and Fungiacyathidae, Turbinoliidae and Flabellidae in "Complex" 11,[14][15][16]74 , showing that both clades present azooxanthellate deep-sea corals as first lineages to diverge. Hence, it would be more parsimonious to assume that the origin of the order is more likely azooxanthellate and solitary. Considering the coral fossil gap observed during the Lower Triassic, corals may have survived as azooxanthellate taxa living in the deep sea, as proposed by Ezaki 70 , and are too rare to be detected in the fossil record after the end-Permian extinction, as suggested by Stanley and Fautin 75 .
In terms of evolution inside Micrabaciidae, the recovered topology reinforces the observations made by Owens 76 that R. niphada may be an intermediate species between Rhombopsammia and Letepsammia and, therefore, the first cycle septal solidity in the former and the total number of septa might not grant the split of both genera. Additional data from the remaining micrabaciid genera (i.e. Leptopenus and Stephanophyllia) will help to clarify such relationship and may shed light on deep-sea adaptations among scleractinian corals.  Owens, 1986: IK-2012-3832) were sourced from the Muséum national d'Histoire naturelle (Paris, France) Cnidaria collection. Total genomic DNA extraction was performed using the DNeasy Blood and Tissue kit (Qiagen) and libraries for Next Generation Sequencing (NGS) were prepared using TruSeq DNA Nano library preparation kit (Illumina; one library per species, based on one sample each). Since DNA from R. niphada was particularly degraded, Covaris shearing parameters were changed for this sample according to manufacturer suggestions (duty cycle: 5%; duration: 70 s). Moreover, in order to avoid adapter-dimer formation, adapters were diluted (3×) and the number of cycles at the PCR step was set to 12 cycles, following Illumina recommendations. Library concentrations were quantified on a Qubit 2.0 fluorometer and size distributions were assessed on a Bioanalyzer (Agilent). Samples were pooled with other libraries and sequenced on two different MiSeq v3 2 × 300 bp runs (L. formosissima in a run with other seven libraries and L. superstes and R. niphada in a different run with other eight). The same libraries from species L. superstes and R. niphada were also included on a NextSeq v2 High Output 2 × 75 bp pooled run with 16 samples in total. Illumina sequencing was performed at the Genome Investigation and Analysis Laboratory of the Centro de Facilidades para a Pesquisa, University of São Paulo. Raw sequences were trimmed using Trimmomatic 77 under default settings, and the trimmed reads Scientific Reports | (2020) 10:20714 | https://doi.org/10.1038/s41598-020-77763-y www.nature.com/scientificreports/ were used to assemble mitogenomes using MITObim 78 . Complete assembly of mitogenomes was ascertained by a circular mitogenome recovered, as determined by circules.py 78 . Assembled mitogenomes were then annotated by MITOS2 79 . Protein coding genes with start or stop codons that did not match the mold/coelenterate mitochondrial genetic code were inspected for the presence of suitable codons before the beginning or after the end of their annotations and re annotated accordingly. Data from previous attempts to sequence Micrabaciidae mitogenomes by primer walking (using the same samples; primer sequences and PCR settings from Lin et al. 74 ) were used to refine R. niphada and L. formosissima assemblies. Data for L. franki were generated by Sanger sequencing followed by editing and assembling in Sequencher 80 . A selection of published mitogenomes (Table 1) was downloaded from GenBank and re-annotated on MITOS2. Nucleotide sequences were aligned by gene (11 PCGss, 2 rRNAs and 2 tRNAS) or exon for multi-exon genes (i.e. nad5 and cox1) in MAFFT v7 using L-INS-i algorithm 81 . Alignments were tested for substitution saturation on DAMBE v7.0.12 82 . In saturated alignments, the third codon position nucleotide was removed and tested for saturation again. Only non-saturated alignments were concatenated using catsequences (https ://githu b.com/ Chris Creev ey/catse quenc es) and the final matrix is available at Zenodo (https ://doi.org/10.5281/zenod o.41338 05). For phylogenetic reconstructions, both maximum likelihood (ML) and Bayesian inference (BI) methods were used. The matrix was partitioned by either gene or exon in multi-exon genes. The former was carried out in RAxML v8.2.12 83 with 1000 bootstrap replicates and 100 random starting trees (GTR + G model). The latter was performed in MrBayes v3.2.7 84 on CIPRES portal 85 , after searching for the best substitution model for each alignment using Bayesian Information Criterion on jModelTest2 86 (run on CIPRES portal). Two Markov chain Monte Carlo runs with four chains each were run for 20 million generations, sampling once every 1000 trees, and discarding the first 30% of them as burn-in, following run convergence check in Tracer v1.7.1 87 .

Data availability
The data underlying this article are available in the GenBank Nucleotide Database at https ://www.ncbi.nlm. nih.gov/nucco re/ and can be accessed with accessions MT705247, MT706034, MT706035 and MT706036-MT706040. The alignment used for phylogenetic reconstructions is available in Zenodo at https ://doi. org/10.5281/zenod o.41338 05.