Revision of Oligocene Mediterranean meandroid corals in the scleractinian families Mussidae, Merulinidae and Lobophylliidae

Traditional morphology-based systematics indicates close evolutionary relationships between Caribbean and Indo-Pacific ‘faviid’ and ‘mussid’ reef corals. However, molecular phylogenies reveal three distinct family-level clades, which diverged by middle Eocene time: (1) Caribbean faviids + mussids; (2) Indo-Pacific faviids; and (3) Indo-Pacific mussids. During the early Cenozoic, members of these clades also occurred in a third geographical region, the Mediterranean, but became extinct in that region during the Miocene, as the Tethys broke up. We perform morphological phylogenetic analyses including Caribbean, Indo-Pacific and Mediterranean Oligocene and Recent taxa to reconstruct the pattern of divergence between the three regions, and examine how it was related to biogeography. First, fossil specimens were selected from museum collections, and a total of 13 species (three of which are new) were distinguished using nine morphological features. These 13 species were then added to a dataset with taxa consisting of 62 Recent plus one additional extinct species, and with 50 characters. In addition to traditional macromorphology, the characters include new micromorphological and microstructural features observed using electron microscopy and transverse thin sections. Phylogenetic analysis was performed on the dataset using parsimony. The results show that, contrary to traditional systematics, 11 of the 13 Mediterranean extinct coral species group more closely with Indo-Pacific taxa than they do with Caribbean taxa. Recent Caribbean taxa and Indo-Pacific ‘mussids’ form distinct clades; but Indo-Pacific ‘faviids’ form four poorly resolved subclades basal to the Caribbean clade. These results suggest that Mediterranean meandroid corals belong to a cosmopolitan pantropical fauna, from which modern Caribbean meandroid corals diverged as the Caribbean became isolated. Phylogenetic analyses including fossils have higher resolution than analyses including only modern corals. The systematics of the 13 extinct species are formally revised. Two new species – Variabilifavia ausuganensis sp. nov. and Echinophyllia sassellensis sp. nov. – and one new genus – Paraleptoria gen. nov. – are named, and one undescribed species is left in open nomenclature. Two previously synonymized genera are resurrected. http://zoobank.org/urn:lsid:zoobank.org:pub:A5731291-9792-464D-87CF-E1DED0C9057E


Introduction
The biodiversity of modern reef corals differs significantly between the Indo-Pacific and Caribbean regions. In the Indo-Pacific, biodiversity has been estimated as 17 families, 92 genera and >700 species based on traditional taxonomy, whereas in the Caribbean it has been estimated at 11 families, 27 genera and »60 species (Veron 2000). Traditional taxonomy indicates that only 17% of all modern reef coral genera and none of the 17 modern reef coral families have distributions limited to the Caribbean, whereas 76% of all genera and 39% of families are restricted to the Indo-Pacific (Veron 1995(Veron , 2000. These numbers suggest that the Caribbean fauna represents merely a depauperate extension of the Indo-Pacific fauna. However, recent molecular analyses (Fukami et al. 2004(Fukami et al. , 2008 and studies of micromorphology and microstructure (Budd & Stolarski 2009Budd et al. 2010) have found this not to be true. Instead, much of the modern Caribbean fauna appears to be only distantly related to the Indo-Pacific fauna, with the faunas of the two regions diverging during the Palaeogene (Schwartz et al. 2012). These results indicate that the evolutionary histories of the two regions are significantly more complex than previously recognized, and that the differences in biodiversity may have developed in response to different geographical and environmental conditions in the two regions.
In order to trace the evolutionary histories of the two modern regions, we have been performing analyses of the fossil record, not only from Caribbean and Indo-Pacific locations, but also at other intermediate locations within the westward-flowing Tethyan Seaway which extended around the globe during the Palaeogene (Berggren & Hollister 1974). One diverse intermediate region is the Mediterranean, where reef corals thrived and constructed large reefs during the Oligocene and Miocene (Bosellini & Perrin 2008;Perrin & Bosellini 2012) but are extinct today (Perrin & Bosellini 2013;Vertino et al. 2014). This region has a long history of palaeontological research dating back to the early 1800s, when the names of species were originally established. Much of our research has therefore concentrated on museum collections. In addition, due to the long history of research on the region, a plethora of scientific names has been assigned to both genera and species, many of which are synonyms or are incorrectly and inconsistently applied. The problems in name usage have necessitated re-examining type specimens and coding sets of morphological characters following recent advances in scleractinian morphological analysis.
In the present study, we focus on a subset of the Palaeogene Mediterranean reef coral fauna consisting of Oligocene meandroid reef corals within three closely related families À Mussidae Ortmann 1890, Merulinidae Verrill 1865 and Lobophylliidae Dai & Horng 2009 À and formally revise their systematics. We focus on meandroid corals because they complement ongoing research comparing the molecular and morphological data in modern members of this group described below. Our eventual goal is to: (1) expand this work to include the Paleocene, Eocene and Miocene, as well as non-meandroid genera (and Hydnophora) in the three families; (2) compare taxa in the three families with their Caribbean, Middle Eastern and Indo-Pacific counterparts; and (3) reconstruct the phylogeny of the group. This phylogeny will then be analysed to determine how origination and extinction events correspond with long-term changes in climate and biogeography.
This new system is used in the Corallosphere database (http://corallosphere.org) and in WoRMS (http://www. marinespecies.org). Our approach to the study of fossil corals performs morphological phylogenetic analyses including both modern and fossil corals to assign taxa to families and genera (see also Schwartz et al. 2012;Santodomingo et al. 2014). These analyses are specifically designed to address the following question: are Mediterranean corals more closely related to Caribbean or Indo-Pacific corals? First, we examined fossil meandroid corals in classic and contemporary museum collections, selecting 112 specimens that belong to the families Merulinidae, Lobophylliidae and Mussidae for analysis. Secondly, we grouped the specimens into 13 extinct species by assessing nine different morphological features, including both macromorphology and microstructure. Thirdly, we added one additional extinct species, coded the extinct species using a set of characters modified from Budd et al. (2012), and added 62 living species to create a combined extinct C Recent morphological character matrix (76 taxa £ 50 characters). Fourthly, we performed phylogenetic analyses using the character matrix. Finally, we formally revised the systematics of the extinct species based on our results.
Over the past 200 years, Oligocene Mediterranean meandroid corals have been assigned to 20 different genera (Supplemental Appendix A) in seven currently recognized scleractinian families (Agariciidae, Calamophylliidae, Euphylliidae, Meandrinidae, Lobophylliidae, Merulinidae, Mussidae). As explained earlier, our revision focuses on only three of these families: Mussidae, Merulinidae and Lobophylliidae. However, we first examined all available museum specimens of meandroid corals in the collections described above and used the macromorphological criteria in Supplemental Appendix A to determine which specimens belonged to three families under consideration. Within the three selected families, we excluded solitary taxa (e.g. Leptomussa, Cricocyathus, Petrophylliella) and members of seven genera: Favia, Dipsastraea, Caulastraea, Lobophyllia, Orbicella, Cyphastrea and Hydnophora. The first six genera were excluded because they are not meandroid (i.e. their corallite series contain 3 centres), and Hydnophora was excluded because it was recently revised by Bosellini (1999). Examples of species that were excluded at this initial stage are shown in Figure 1. A list of all included 112 specimens is provided in Supplemental Appendix B.
Next we qualitatively grouped the included specimens into species (a total of 13 species, Supplemental Appendix C) using the following macromorphological and microstructural criteria: corallum (colony shape; development of epitheca), corallite (length, width, and structure of corallite series, 'valleys'), development of coenosteum, costa continuity, septa (number, relative development), columella (size, structure, linkage), development of paliform lobes, relative abundance of dissepiments, and wall structure. Observations were made on calical surfaces using a stereoscope and on transverse thin-sections using transmitted light. The resulting species comprise the extinct taxa that were analysed in the phylogenetic analysis and are formally described in the systematic account.

Localities
Oligocene fossil specimens used for this study come from seven localities: five are from Italy, one from Slovenia and one from France (Fig. 2). Their updated biostratigraphical ages and palaeoenvironments, together with most recent references, are summarized in Table 1. Of the seven sites, four include some of the most significant coral-bearing units in the Oligocene of the north-western Tethys, and have been known since the nineteenth century from famous monographs and related museum collections: Gornji Grad in Slovenia, the Marostica area and the Eastern Lessini Mountains in the Vicentin Southern Alps, and Sassello in Liguria (Catullo 1856;Reuss 1864;d'Achiardi 1866d'Achiardi , 1867d'Achiardi , 1868aReuss 1868Reuss , 1869Michelotti in Sismoda 1871;de Angelis 1894;Osasco 1898Osasco , 1902Prever 1921Prever , 1922. A brief overview of the localities is provided below following the order in Table 1. Gornji Grad (N Slovenia) (Locality 1). The coral fauna belongs to the Gornji Grad Beds (named Oberburg in the old literature) and was first described by Reuss (1864) with the collection reposited at the NHMW. The Gornji Grad Beds were deposited within the so-called 'Slovenian Corridor', located between the northernmost extension of the Tethys to the south and the developing Paratethys. They are represented by brackish to marine marls, sandstones and marine carbonates, with corals found mainly within rudstone carbonates and marly levels (Nebelsick et al. 2000). An early Oligocene age (Rupelian) has been assigned based on the larger benthic foraminiferal association (Biozone SB21 of Cahuzac & Poignant 1997) (Nebelsick et al. 2000). For the coral-rich facies, nearshore turbid-water conditions dominated by sedimentresistant corals have been recently interpreted (Silvestri et al. 2011).
Localities of Veneto and Trentino regions (northern Italy) (Localities 2, 3, 4). All of these localities belong to the Lessini Shelf, a major palaeogeographical unit of the Southern Alps characterized by a well-exposed succession of Palaeogene shallow-water carbonates (Bosellini 1989). The areas of the Eastern Lessini Mountains and Marostica are punctuated by a large number of coral outcrops that represent some of the best-known Cenozoic type localities in the scleractinian systematic literature (Catullo 1856;d'Achiardi 1866d'Achiardi , 1867d'Achiardi , 1868aReuss 1868Reuss , 1869. The Lessini outcrops belong to the Rupelian Castelgomberto Limestone, a 200 m thick unit organized in a number of cycles where well-bedded grainstone units alternate with marly horizons extremely rich in corals (Bosellini & Trevisani 1992). Palaeoenvironmental reconstructions of these sites suggest that scattered coral patches colonized the muddy-unstable substrate of the shallow internal platform or 'lagoon-like' setting in low-energy hydrodynamic conditions (Frost 1981;Bosellini & Trevisani 1992). A very similar depositional environment has been proposed for the coeval coral facies of the Marostica area (Pfister 1980;Frost 1981). Towards the northern margin of the Lessini Shelf, near Borgo Valsugana (Valsugana valley, eastern Trentino Province), the Oligocene shallow-water sedimentary succession involves a series of Rupelian cycles (parasequences) similar to those observed in the Vicentin Lessini Mountains (Luciani & Trevisani 1992), with a rich coral assemblage occurring within marly levels (Boschele et al. 2011).

Sassello (Liguria, north-western Italy) (Locality 5).
Sassello is another classic locality of the Italian Oligocene, well known from the studies of Michelotti in Sismonda (1871), de Angelis (1894) and Prever (1921,1922). This area belongs to the southern part of the western Tertiary Piedmont Basin, a thrust-top basin developed during the Cenozoic over the suture zone between the Alps and Apennines and generated by post-collisional subsidence (Mosca et al. 2010). The area is rich in coral localities where small coral build-ups and scattered coral assemblages are associated with the mixed siliciclasticcarbonate sediments of the Molare Formation and  developed in a conglomeratic fan delta to clay-rich prodelta setting (Pfister 1985;Fravega et al. 1987;Silvestri et al. 2008;Quaranta et al. 2009). The age of the Sassello coral facies, determined through the analysis of the larger benthic foraminiferal assemblage, is Upper Rupe-lianÀLower Chattian (Biozone SB22A-22B of Cahuzac & Poignant 1997) (Quaranta et al. 2009;unpublished data).
Abesse (Saint-Paul-l es-Dax, Aquitaine, SW France) (Locality 6). Corals from the Aquitaine Basin are also well known since the nineteenth century and have been extensively studied by Chevalier (1962) and revised more recently by Cahuzac and Chaix (1996), who recognized at the locality of Abesse (near Saint-Paul-l es-Dax) an extraordinary rich coral fauna (up to 100 species). A marginal-coastal depositional setting associated with a mixed carbonate-siliciclastic sedimentation has been suggested for these coral facies, characterized by 'faluns' (i.e. shelly sands) and littoral/lagoonal marly sands (Cahuzac & Janssen 2010;unpublished data). These corals were not forming reefs or any sort of framework. As regards their age, the coral deposits have been ascribed to the uppermost Chattian (Cahuzac & Poignant 2002;Cahuzac & Janssen 2010; unpublished data).  Fringing reef fore reef Bosellini & Russo (1992); Bosellini (2006) Castro (Salento Peninsula, southern Italy) (Locality 7). Corals analysed from Castro come from the Castro Limestone, one of the rare Oligocene true large reef complexes of the Mediterranean region. This carbonate unit is widely exposed along the eastern coast of the Salento Peninsula and has been interpreted as a fringing reef complex disconformably overlying the tectonically deformed Cre-taceousÀEocene eastern margin of the Apulia Platform (Bosellini & Russo 1992;Bosellini et al. 1999). Palaeoenvironmental reconstructions suggest a very well-preserved lateral zonation of reef facies from the back reef towards the reef slope, and reveal changes in coral composition, growth form and fabric within the reef framework (Bosellini & Russo 1992;Bosellini & Perrin 1994;Bosellini 2006). Coral richness is moderately high, consisting of approximately 21 genera and 30 species (Bosellini 2006). The age of the Castro Limestone has been attributed to the early Chattian according to the larger benthic foraminiferal assemblages (Biozone SB22B of Cahuzac & Poignant 1997) (Parente 1994).

Systematic palaeontology
Remarks. In the systematic account that follows only new species names are considered in synonymies. Question marks are indicated for species whose type specimens are lost and which were figured inadequately or not at all. Only Mediterranean Oligocene species are included in synonymies. Non-Oligocene and non-Mediterranean synonyms are considered in the Remarks sections. Genera are determined based on the phylogenetic analysis shown in the results section; species are based on characters in Supplemental Appendix C. The term 'non-type' refers to specimens that are not primary types. For the two species in which we have no thin sections (Variabilifavia ausuganensis, Paraleptoria polygonalis), wall type has been estimated by study of the colony surface.
Type species. Favia perrandi Prever, 1922, p. 29, pl & Haime, 1848 (double-wall) and Hydnophyllia (no coenosteum) by having a continuous spongy columella with trabecular linkage. Finally, it differs from Platygyra Ehrenberg, 1834, which lacks coenosteum. Variabilifavia therefore appears to be a distinct genus, as confirmed by the separate clade it forms shown in the phylogenetic analysis section. We assign the name Variabilifavia to this genus, because we consider the type species of Variabilifavia, Favia perrandi Prever, 1922, to be a synonym of Favia confertissima Reuss, 1868 as described below.
Derivation of name. After 'Valle di Ausugum'; Ausugum was the old Roman name of the village Borgo Valsugana.
Michelotti described Meandrina cerebriformis for the first time in 1838 (Michelotti 1838, p. 154) without figuring it, and indicated that it contained both modern Caribbean and fossil (Verona) specimens. He redescribed it as Symphyllia crebriformis in 1861 (Michelotti 1861, p. 39), again without figuring it, and indicated that it occurred in Dego and Sassello. The specific name 'crebriformis' appears to be a misspelling of 'cerebriformis'. The extinct species differs morphologically from Lamarck's (1816) Meandrina cerebriformis, which is modern Caribbean and synonymous with Diploria labyrinthiformis (Linneaus, 1758). Lamarck's species has larger valleys, more extensive coenosteum, and septothecal walls. Pfister (1980) erected the lectotype for Diploria crebriformis (MPUR 3700) based on Michelotti's collection in Rome.
Michelotti in Sismonda (1871) described and figured Diploria intermedia for the first time on p. 324 and redescribed Symphyllia crebriformis without figuring it on p. 326. The holotypes for both species are lost; however, D. intermedia and S. crebriformis are both from Sassello and appear to be the same species.
Reuss' (1868) Favia confertissima from Castelogomerto is very similar to the Sassello species and differs primarily by having a narrower coenosteum; we therefore synonymize the two species. Frost (1981) considered Favia confertissima to be a Goniastrea, presumably because he believed it to lack coenosteum.
Remarks. Hydnophyllia has long been considered to be a synonym of Colpophyllia (Vaughan & Wells, 1943, p. 171;Wells, 1956, p. F403); however, it differs from Colpophyllia: (1) by lacking a coenosteum and not forming a distinctive 'double-wall' on the calical surface; (2) by sometimes being multiserial and/or having direct linkage (two or more lamellae) between centres; (3) by having a trabeculothecal (not parathecal) wall; and (4) by having little or no epitheca. Its septal cycles are subequal, instead of equal as in Colpophyllia, and it lacks the small septal lobes that are characteristic of Colpophyllia, instead having paliform lobes. We therefore resurrect Hydnophyllia, removing it from synonymy with Colpophyllia. Hydnophyllia is actually more similar to Oulophyllia Milne Edwards & Haime, 1848, as shown by the phylogenetic analysis below, but differs from Oulophyllia by forming longer series and lamellar (not trabecular) linkage between centres. Barta-Calmus (1973, p. 384) reported the holotype of Leptoria eocaenica to be lost. She renamed the species Hydnophyllia oligocenica and assigned a specimen in the Reis (1889)  . Mycetophyllia has <3 equal septal cycles, a reduced columella, reduced epitheca, and high, widely spaced, spine-shaped septal teeth. Symphyllia has very thick and unequal septa with large lobate teeth.
Remarks. Hydnophyllia costata has larger and more variable valley widths than any other species of Hydnophyllia; its series are V-shaped and uniserial, and their margins usually splay outward.
Diagnosis. Long, straight uniserial valleys (>5 centres), which are constant and narrow in width, and V-shaped.
This species appears to be the same as Meandrina meandrinoides Michelin, 1842 (p. 57, pl. 11, fig. 9), from the Miocene of Turin (Rivalba). ( Diagnosis. Usually one uniserial valley that forms a continuous series; well-developed paliform lobes and a spongy columella.
Remarks. Hydnophyllia serpentinoides is very similar to H. sublabyrinthica in valley width and calice shape (U-shaped), but its valleys are not as sinuous, its columella is spongy, and it has more dissepiments. Its colonies are also smaller and they usually form one continuous series. Barta-Calmus (1973)  Diagnosis. Mostly mono-to tricentric corallites (short series), which are narrow in width; closely spaced septa; well-developed medial lines within costosepta.
Remarks. Hydnophyllia stellata is most similar to H. fimbriata. It is distinguished from all other species of Hydnophyllia by having small, short valleys and paliform lobes.
Hydnophyllia stellata is similar to the Miocene species Coeloria siciliae Chevalier, 1962, which may be a synonym.
Remarks. Fossil Merulina isseli differs from modern M. ampliata (the type species) by having larger valleys that predominantly fork laterally (not terminally), and a discontinuous columella. Its abortive septa are less apparent, and it has transverse septal crosses. These differences suggest that it may represent a separate but closely related new genus or subgenus. Merulina is superficially similar to Variabilifavia but is distinguished from Variabilifavia by its long series, its lack of coenosteum and epitheca, its sparse endotheca, and the presence of abortive septa.
Remarks. Paraleptoria is distinguished from other meandroid genera in this study by its lamellar columella, T-shaped septal margins, and extensive costate coenosteum. Modern Leptoria phrygia (the type species) lacks coenosteum; whereas coenosteum is especially well developed in Paraleptoria. As a result, P. polygonalis does not group with modern Leptoria in the phylogenetic analysis, warranting the description of a new genus. In addition to P. polygonalis, the new genus contains the common early Eocene species P. flexuoissima.
Remarks. This species differs from the Eocene species, Paraleptoria flexuosissima, which has a valley width of 1.5À2 mm and 24À26 major septa per cm. Paraleptoria polygonalis was not reported by Pfister (1980), but Frost (1981) considered it to be a valid species. This species differs from Meandrina phyrgia Michelin, 1842 (p. 55, pl. 11, fig. 5), from the Miocene of Turin (Rivalba), which lacks a coenosteum and appears to be more similar to Merulina isseli. It differs from Leptoria bithecata Schuster, 2002b from the Oligocene of NW Greece, which has a valley width of 4.6À6 mm and a distinctive double-wall. It is also not the same as the species identified as Leptoria cf. concentrica (Duncan, 1880) by Schuster (2002a;Oligocene, Iran)  Diagnosis. Meandroid, long uniserial or multiserial valleys (>5 centres); no coenosteum; unequal costosepta; discontinuous trabecular columella with direct linkage (2 or more lamellae); no lobes or epitheca; parathecal wall; abundant dissepiments; extensive thickening deposits.
Remarks. Among Palaeogene corals, this genus is most frequently confused with Cyathoseris, which is also multiserial and has columellae with direct linkage (2 or more lamellae). However, Cyathoseris has synapticulae and little or no dissepiments; its septa are confluent and its septal cycles are equal.
Remarks. Echinophyllia differs from other members of Lobophylliidae by having an extensive vesicular coenosteum and lacking corallite walls. It is most frequently confused with the merulinid genus Pectinia Blainville, 1825, which also forms series and lacks corallite walls. However, in Pectinia, the coenosteum forms high, acute collines.

Analytical methods
Characters Following Budd & Stolarski (2009 and Budd et al. (2012), we selected 50 characters for morphological phylogenetic analysis (Supplemental Appendix E), and coded the 14 extinct species (13 species treated herein plus one Oligocene Hydnophora). The characters consist of three sets of features: (1) macromorphology (using a stereoscope, at magnifications <50£); (2) micromorphology (using a scanning electron microscope, at magnifications ranging from 50À200£); and (3) microstructure (using transverse thin sections, at magnifications <100£). Micromorphological and microstructural analyses are relatively new to morphological phylogenetic analysis. A glossary of morphological terms that are used in this monograph is provided in Supplemental Appendix D.
Macromorphological features serve as the primary diagnostic characters in traditional classification (Vaughan & Wells 1943;Wells 1956), e.g. Supplemental Appendix A and C. They are architectural in nature, including colony form (corallite budding and integration, the length and shape of calical series); the size and shape of the calice; the structure and development of the septa (number, spacing, relative thickness and length), the columella (and associated internal lobes), the corallite wall, endo-and exotheca, and the coenosteum.
Micromorphological features, most notably septal structure, were also included in the traditional definition of families and higher taxonomic levels, but only in a cursory way that did not involve the use of electron microscopy. They focus on the 3D geometry of teeth (dentation) along the upper margins of the costosepta and columella (the septal growing edge), as well as on granulation on septal faces and the sides of teeth. Teeth and granules are surficial projections, which reflect the underlying calcification axes that build the internal structure and framework of the costoseptum. In the present study, micromorphological features could not be examined in extinct taxa due to inadequate preservation, and therefore were coded as missing in the character matrix (Supplemental Appendix F).
With the exception of corallite wall structure, microstructural features were not used in the traditional classification of Vaughan & Wells (1943) and Wells (1956), although they were included in the classification systems of Alloiteau (1952Alloiteau ( , 1957 and Chevalier & Beauvais (1987). They involve the internal structure (i.e. the arrangement of calcification centres and fibres) within the wall, septa, and columella, and consist of 2D observations made primarily using transmitted light on petrographic thin sections. As in Budd et al. (2012), we focus both on the corallite wall (the skeletal structure uniting the outer edges of septa in a corallite) using the wall types defined by Budd & Stolarski (2011), as well as on the costosepta and columella. In the latter observations, we consider the degree to which calcification centres are clustered, the distinctiveness of costoseptum medial lines, and the presence of transverse structures or clusters of centres crossing medial lines. Despite recrystallization, these features could be observed in extinct taxa for which we have thin sections.
These modifications were made to facilitate inclusion of the extinct taxa in the character matrix, and to provide more precise definitions of characters.

Phylogenetic analysis
The phylogenetic dataset was constructed by adding the 14 extinct species (13 species treated herein plus one Oligocene Hydnophora) to a character matrix for 62 living species, making a total of 76 taxa (Supplemental Appendix F). The living species consist of 62 of the 67 species analysed in Budd et al. (2012, figs 2, 8 Parsimony analysis was performed on the dataset (76 taxa £ 50 characters) using PAUP Ã 4.0b10 (Swofford 2002). Heuristic searches were conducted with 10,000 random addition replicates on the 76-taxon by 50-character data matrix. Tree bisection and reconnection (TBR) was used as the branch-swapping algorithm. Zero-length branches were collapsed if they lacked support under any of the most parsimonious reconstructions. To assess clade support, heuristic searches were performed on 1000 bootstrap pseudoreplicate datasets (Felsenstein 1985), with 100 random addition sequence replicates for each bootstrap search. In addition, Bremer support values (Bremer 1988) were calculated for each node in the strict consensus of MPTs resulting from each analysis using TreeRot 3 (Sorenson & Franzosa 2007). The number of random addition sequence replicates performed for each constraint analysis was set to 100.

Results
The phylogenetic analysis performed on the character matrix (67 taxa £ 50 characters, Supplemental Appendix F) recovered 86 most parsimonious trees (MPTs) of length 93 steps, with a consistency index (CI) of 0.226 and a retention index (RI) of 0.750. The strict consensus tree of the MPTs is shown in Figure 11. Six clades (AÀF) are recovered as monophyletic, albeit with low support. Three modern species did not group with any of the six clades: Montastraea cavernosa, Goniastrea stelligera and Astrea curta. In general, the six clades agree with previously defined molecular clades. Clade A corresponds with molecular clade XXI (Mussidae) of Fukami et al. (2008); and clade F corresponds with molecular clades XVIIICXIXCXX (Lobophylliidae) of Fukami et al. (2008), with the exception of two members of the former family Pectiniidae (Oxypora and Echinophyllia). Clades BCCCDCE correspond with molecular clade XVII (Merulinidae) of Fukami et al. (2008), such that clade B corresponds with molecular subclade XVII-C (as defined by Stolarski 2011 andHuang et al. 2014), clade C with molecular subclades XVII-ACGCH, clade D with molecular subclades XVII-BCDCF, and clade E with molecular subclade XVII-F.
The 14 extinct species group with four of the six clades shown in Figure 11. The two Variabilifavia species group with clade A. Species of Paraleptoria and Merulina and Oligocene Hydnophora group with clade C. The seven Hydnophyllia species group with clade D, and the fossil Echinophyllia groups with the former family Pectiniidae. The seven Hydnophyllia species form a separate subclade within clade D. Assignments of species to families and genera detailed in the Systematic palaeontology section above are based in part on these results.

Discussion
The results of the present study show considerably higher diversity in Oligocene Mediterranean meandroid corals (13 species) than that reported most recently (two species) by Frost (1981) and Pfister (1980), resulting in the resurrection of many of the species originally described by Catullo (1856). Our examination of nineteenth century type species (Catullo 1856;d'Achiardi 1868b;Reuss 1864Reuss , 1868Reuss , 1869Reuss , 1874 indicates that species names correspond among these classic works as shown in Figure 12. Our phylogenetic analysis (Fig. 11) shows that only two of the 13 species group with a modern Caribbean family (Mussidae-XXI), and that 11 species group with modern Indo-Pacific families (Merulinidae-XVII and Lobophylliidae-XVIIICXIXCXX). The overall results reveal six main clades labelled AÀF: one (A) consisting of the family Mussidae, four (BÀE) consisting of the family Merulinidae and one (F) consisting of the family Lobophylliidae. The one exception is Clade D, which includes a subclade (OxyporaCEchinophyllia, members of the former family Pectiniidae) belonging to the Lobophylliidae. At the genus level, three distinct fossil genus-level 'subclades' occur within the six main clades (Fig. 11), resulting in the resurrection of two fossil genera, Variabilifavia (two species) in clade A (Mussidae) and Hydnophyllia (seven species) in clade D (Merulinidae), and the description of one new fossil genus, Paraleptoria (one species), in clade C (Merulinidae). The remaining three species group with the modern genera Merulina (clade C, Merulinidae), Echinophyllia (clade D, Lobophylliidae) and Symphyllia (clade F, Lobophylliidae). Variabilifavia, Hydnophyllia and Paraleptoria all have a distinctive trabeculothecal wall, similar to modern Indo-Pacific Leptoria, Platygyra and Hydnophora. Variabilifavia is morphologically most similar to modern Caribbean Diploria (septothecal) in that it has a well-developed ambulacrum; however, its series are shorter, and its coenosteum is not as well developed. In addition, it has a discontinuous columella and more septal cycles, and differs in wall structure. Variabilifavia is also similar to modern Caribbean Favia (septothecal), but has longer series and a trabeculothecate wall. By contrast, Hydnophyllia is most similar to modern Indo-Pacific Oulophyllia (parathecal) in its series dimensions, lack of coenosteum and discontinuous columella. However, it too differs in wall structure. It has frequently been confused with modern Caribbean Colpophyllia (parathecal), which unlike Hydnophyllia has coenosteum, a double-wall, septal lobes and reduced thickening deposits. Paraleptoria is most similar to modern Indo-Pacific Leptoria, both genera having trabeculothecal walls and a lamellar columella. However, unlike Leptoria, Paraleptoria has a well-developed coenosteum. Clearly the patterns in wall structure are noteworthy and require further investigation.
The grouping of Mediterranean extinct species with modern family-level clades in the present study disagrees with previous interpretations that Oligocene Mediterranean and modern Caribbean meandroid corals are more closely related to each other than they are to Indo-Pacific corals (Pfister 1980;Frost 1981). Instead, they suggest that most Mediterranean Oligocene meandroid corals were part of a more cosmopolitan pan-tropical Tethyan fauna, which extended from the Indo-Pacific into the Mediterranean. The seaway between the Eastern Mediterranean Basin and the Mesopotamian Basin narrowed during the Oligocene, but nevertheless marine circulation from the Indian Ocean to the Atlantic was maintained until the early Miocene (Perrin 2002;Harzhauser et al. 2007). However, few corals with possible Indo-Pacific affinity have been reported to date from Iran and elsewhere on the Arabian shelf (Schuster 2002a, c;Schuster & Wielandt 1999). The close relationship that we have observed between Mediterranean and Indo-Pacific meandroid corals concurs with the 'hopping hotspot' theory of Renema et al. (2008), which states that diversity was highest in the Mediterranean during the Eocene and the hotspot shifted in a south-west direction toward South-East Asia, where it has remained since the Miocene (Johnson et al. 2015a, b).
On the western side, our analyses indicate that the Caribbean region may have been more geographically isolated by the widening Atlantic than previously interpreted. Chevalier (1977) and Perrin (2002) have noted that gradual isolation began in the late Oligocene in association with a major change in oceanographic circulation in the central Atlantic. Our results suggest that isolation occurred even earlier, in the early Oligocene or Eocene. Additional comparative work between Caribbean and Mediterranean EoceneÀOligocene reef coral faunas using up-to-date systematics is underway to test this hypothesis. Similarly, previous authors have suggested that many of the Mediterranean reef-building genera and species may have been geographically restricted (Perrin & Bosellini 2012). Further work comparing Caribbean, Mediterranean and Indo-Pacific Eoce-neÀOligocene reef coral faunas using up-to-date systematics is needed to reassess these geographical ranges.
The addition of fossils to the morphological phylogenetic analysis in the present study, albeit with missing data, has resulted in a significant improvement not only in tree resolution over constructions based only on modern taxa, but also in the agreement between morphological and molecular trees. In a previous modern-only morphological analysis (Budd et al. 2012, fig. 8; 68 taxa £ 38 characters), 243 maximum parsimony trees were recovered with a tree length of 200. The Adams consensus tree showed that Indo-Pacific Lobophylliidae (molecular clade XIX) formed a distinct group, and Caribbean Mussidae (molecular clade XXI) formed three distinct groups (two Faviinae, one Mussinae). However, Indo-Pacific Merulinidae (molecular clade XVII) remained for the most part unresolved. In a phylogenetic analysis of modern-only corals based on exactly the same morphological character set as the present study (62 taxa £ 50 characters analysed using PAUP Ã ), the analysis found 33,136 maximum parsimony trees with a tree length of 284, compared with 86 trees of length 93 when the 14 fossils are included. Caribbean Mussidae (molecular clade XXI) form a distinct group, and with the exception of Oxypora-CEchinophyllia, former members of the Pectiniidae, Indo-Pacific Lobophylliidae (molecular clades XVIIIC XIXCXX) form a distinct group. However, Indo-Pacific Merulinidae do not. By contrast, when fossils are included in the morphological analysis, no longer are modern Caribbean meandroid corals (Mussidae: molecular clade XXI) intermixed with modern Indo-Pacific taxa (Merulinidae: molecular clade XVII) as in Budd et al. (2012), but instead the Caribbean taxa form a separate clade. Moreover, as described above, modern Indo-Pacific merulinids form four distinct clades (labelled BÀE in Fig. 11), which are congruent with the molecular merulinid clades described by Budd & Stolarski (2011) and Huang et al. (2014).
The improved agreement between morphological and molecular data when fossils are included in the analysis has been found elsewhere in analyses of corals (Santodomingo et al. 2014), as well as in numerous other organisms, including seed plants and amniotes (Donoghue et al. 1989). Improved agreement has been found even when incomplete taxa are added to an analysis (Wiens & Tiu 2012). These results support the notion that limited taxon sampling is equally as problematic as limited character sampling in phylogenetic analysis. Adding extinct taxa has the potential to improve significantly phylogenetic accuracy in groups, such as corals, in which numbers of morphological characters are low, convergences and reversals are high, and branches are relatively long. Fossils have been hypothesized to have a unique combination of primitive and derived character states that can strengthen certain groupings, especially if the included extinct corals are temporally closer to the ancestor (Gauthier et al. 1988;Huelsenbeck 1991;Smith 2010). In the present study, this is the case within the merulinids (clades BÀE), as shown in Figure 11. Resolving higherlevel relationships among mussids, merulinids and lobophylliids would require including even older fossils in future analyses, dating back to the divergence of the merulinids and lobophylliids in the Cretaceous (Santodomingo et al. 2014). Improved phylogenies in scleractinian systematics will require not only integration of molecular and morphological data, but also the inclusion of extinct taxa.

Conclusions
We have shown that an up-to-date approach to systematics is essential for uncovering evolutionary patterns in reef corals that have long been masked by inadequate phylogenetic analysis. Our work has discovered previously unrecognized faunal differences that are clearly related to biogeography. Most importantly, Oligocene Mediterranean meandroid reef corals are more closely related to modern Indo-Pacific corals than they are to modern Caribbean corals. Mediterranean meandroid corals may have been part of a more cosmopolitan Tethyan fauna, whereas Caribbean meandroid corals were more isolated, resulting in the development of the genetically unique fauna that occupies the region today. The systematics of additional coral families needs to be similarly analysed to determine how widespread these patterns are and to understand better the role that biogeography has played in the evolution of the modern reef coral fauna.