The South American and Australian percichthyids and perciliids. What is new about them?

ABSTRACT A study including morphological characters and mitogenomics of South American and Australian fishes previously assigned to Percichthyidae was conducted. Results generated from these different data sets reveal major disagreements concerning the content and interpretation of the so-called percichthyids. A phylogenetic analysis based on 54 morphological characters suggests the existence of two major clades: (1) Percichthyidae, including the South American Percichthys and the Australian taxa Macquaria australasica, Macquaria (= Plectroplites), and Maccullochella; (2) Perciliidae with the South American genus Percilia at the base plus more advanced Australian genera Nannoperca, Nannatherina, Bostockia, and Gadopsis. In contrast, molecular and mitogenomic evidence suggests only one clade (Percichthyidae), with the exclusion of species of Macquaria (= Percalates). Additionally, the results reveal the existence of various taxonomic problems, such as the current interpretation of only one species of Percichthys in Argentina, an interpretation that is not supported by the present study; the existence of cryptic species of Percilia as well as of Gadopsis, Nannoperca, and Macquaria that will increase the diversity of the genera; and the need for an extensive revision of species previously assigned to Percalates versus Macquaria. Disagreements point to the need to develop further research on the so-called percichthyids and perciliids.


Introduction
Gondwanan continents are characterized, in general, by a notable endemism due to their isolation, exemplified by there being only a few fish taxa that are common to two or more continents. For example, dipnoans have representatives in Africa, Australia, and South America, with each con-tinent having its distinct family, and osteoglossomorphs are broadly diversified in Africa, but have one monotypic family inhabiting South America. The so-called temperate perches belonging to the family Percichthyidae are currently interpreted as inhabitants of southern South America and Australia, including Tasmania, while Perciliidae are only in South America (Arratia, 2003a(Arratia, , 2003bFricke et al., 2018). that Johnson's (1984) percichthyids "forms a monophyletic group". Later, it was suggested that Percilia and the Australian genera Edelia, Nannoperca, and Nannatherina are more closely related to each other than with Percichthys and other Australian forms (Arratia, 2013b). Nelson et al. (2016) recognized the family Percichthyidae, while the genus Percilia was excluded and kept in Perciliidae.
Starting in 2001, molecular studies (see Tab. 2, Figs. 2 and supplementary S1), including mitochondrial, nuclear and genomic markers, gave different interpretations of the content of Percichthyidae sensu Johnson (1984). The first molecular study (Jerry et al., 2001) included 20 Australian temperate perches, but not Percichthys and Percilia (Fig. supplementary S1a), so the monophyly of Percichthyidae was not evaluated. An important finding was that Macquaria (= Percalates) and Macquaria sensu stricto were found to be paraphyletic. Smith, Craig (2007) addressed the limits and relationships of Serranidae and Percidae families, considering Bostockia, Gadopsis, Howella, Macquaria (= Percalates) and Nannoperca as representatives of Percichthyidae. In their results, Smith, Craig (2007) found a polyphyletic Percichthyidae with three distinct lineages (Fig. supplementary S1b). Near et al. (2012) investigated the phylogenetic relationships of Elassoma using nuclear markers and included 68 species across a diversity of percomorph fishes. They found a paraphyletic Percichthyidae, where the catadromous Percalates colonorum and Percalates novemaculeata were distantly related to the other freshwater percichthyids (Fig. supplementary S1c), and given that Percilia irwini was nested within this group, they suggested that Perciliidae should be synonymized with Percichthyidae (Near et al., 2012). Later, both Betancur-R et al. (2013) and Near et al. (2013) found again a paraphyletic Percichthyidae because of Percalates, whereas the phylogenetic placement for Percichthyidae sensu stricto was incertae sedis for the former and within Centrarchiformes for the latter (Fig. supplementary S1d-e). Similar results regarding the paraphyletic condition of Percichthyidae and the phylogenetic position within Centrarchiformes was found by Chen et al. (2014;Fig. supplementary S1f) and Lavoué et al. (2014) using nuclear and mitogenomic markers, and Percichthys trucha and Percilia irwini were resolved as allied taxa (Figs. 2a,b). Finally, Sanciangco et al. (2016;Fig. 2c) and Betancur-R et al. (2017;Fig. 2d) using a combined supermatrix of mitochondrial and nuclear genes largely confirmed the paraphyletic condition of Percichthyidae sensu Johnson (1984).

Fig. 2.
Phylogenetic hypotheses based on mitochondrial, nuclear and genomics markers concerning the content of Percichthyidae sensu Johnson (1984). a, Mitogenomic maximum likelihood [ML] tree of Lavoué et al. (2014), b, total evidence ML of Lavoué et al. (2014), c, Sanciangco et al. (2016), and d, Betancur-R et al. (2017). Representatives of Percichthyidae are shown in pink, Percilia irwini (Perciliidae) is shown in light blue and Macquaria (= Percalates) is shown in green. The purposes of this contribution are: 1) to study and interpret certain morphological characters of the South American genera Percichthys and Percilia; 2) to understand South American percichthyids in their historical and biogeographic scenarios; 3) to investigate their phylogenetic relationships; and 4) to analyze and discuss current interpretations of percichthyids and perciliids. A brief presentation of the South American and Australian temperate perches follows.
The first mention of percichthyids or "Chilean trout" is from Molina (1782) in his description of the natural history of Chile. About 50 years later, Perca (= Percichthys) trucha, was formally described from freshwaters of southern Argentina. Such taxonomic assignment was accepted also for the Chilean percichthyids (Guichenot in Gay, 1848;Jordan, Eigenmann, 1890;Duarte et al., 1971;Arratia et al., 1981;Arratia, 1982;Dyer, 2000). In contrast, López-Arbarello (2004) restricted P. trucha to Argentina and recognized Percichthys chilensis Girard, 1854 (commonly accepted as a synonym of P. trucha, but see Fricke et al., 2018), which was described, however, based on specimens from the Maipo Basin, Chile. For the purpose of this contribution, we consider P. trucha as an inhabitant of Argentina and P. chilensis of Chile (Fig. 3).
Percichthys chilensis (Figs. 3a, 4a-b), P. melanops (Fig.  4c), and P. laevis are characterized as having a small mouth, with the articulation between lower jaw and quadrate positioned below the mid-region of the orbit (Figs. 5a-c), whereas the mouth is large in P. colhuapiensis, with the articulation of the lower jaw-quadrate positioned behind the posterior margin of the orbit (Fig. 5d) so that the fish is identified with the common name "perca bocona" (= "large-mouth perch"). In addition, there are differences in the musculature associated with feeding mechanism among these species (see Results below). However, differences in the upper jaw length from larval through juvenile to adult stages and the causal relationship between diet and shape of the head of the Argentinian species were interpreted as the result of high morphological plasticity of P. trucha by Crichigno et al. (2014). Those authors follow Ruzzante et al. (2006) whose results based on molecular evidence suggested that only one species, Percichthys trucha, inhabits both sides of the Andean Range, whereas P. melanops (Fig. 4c) lives on its west side.
In general, comparisons through the years show that in the past Percichthys was abundant, forming large schools in the main rivers of central-south Chile. Because of its fine taste, the fishes were used as food, and they were caught by artisanal fishermen using dynamite (a common practice at the beginning of the last century). This kind of fishing produced a big fish mortality until the use of explosives was prohibited in the last century. Thus, Percichthys had a broader distribution in Chile in the past, from Aconcagua Basin (ca. 33º S) in the north reaching up to Punta Arenas in the south (ca. 54º S, where it is now locally extinct) as demonstrated by specimens collected by J. B. Hatcher during his expedition to Patagonia at the end of the 19th century (Haller, 2004;Fig. 4a herein). The records cited in Eigenmann (1927;collected by the Irwin expedition during 1918collected by the Irwin expedition during -1919 are deposited in the California Academy of Sciences . Currently, species of Percichthys live in southern Argentina and central-southern Chile (Fig. 3b).
The fossil record of South American percichthyids is known from three genera- †Santosius Arratia, 1982, Percichthys (Fig. 6b), and †Plesiopercichthys Agnolin et al., 2014; one of them ( †Santosius) was living in the Neotropical Region and the other two in the Austral region of South America. The oldest known percichthyid is †Santosius antiquus (Woodward, 1898) from the Tremembé Formation (in the lacustrine unit of the Taubaté Basin) in the eastern part of São Paulo (Oligocene to early Miocene, ca. 30-25 Ma;. Later, an incomplete specimen assigned to †Santosius? sp. was described and illustrated from the Miocene of the Cura-Mallín Formation, Lonquimay, Bío-Bío Region (Rubilar, 1994); however, this assignment needs revision. Fossil Percichthys are known from four species: †P. hondoensis Schaeffer, 1947 from the ?lower Eocene of Cañadon Hondo, Chubut Province, Argentina, †Percichthys lonquimayensis Chang et al., 1978 (Fig. 6a), †Percichthys sandovali Arratia, 1982, and †Percichthys sylviae Rubilar, Abad, 1990 from the middle Miocene of Cura-Mallín (= Río Pedregoso) Formation (ca. 12.6 Ma), Lonquimay, Bío-Bío region, Chile (Pedroza et al., 2017). In addition, Percichthys sp. specimens from the same locality and formation were described and illustrated by Rubilar (1994:pl. 2, figs. 7-10). All these fossils are known from almost complete specimens, whereas †Plesiopercichthys dimartinoi is known from disarticulated bones that were recovered in the Pliocene of Farola Monte Hermoso, southern region of Buenos Aires Province, Argentina. Additionally, fragmentary bones (e.g., spines and preopercles) assigned to Percichthys sp. and undetermined percichthyids from the Upper Cretaceous of Bolivia were cited by Gayet in de Muizon et al. (1983), in Marshall et al. (1985), and in the Maastrichtian-Paleocene of Santa Lucía Formation, Bolivia, by Gayet et al. (1991).  Schaeffer, 1947, an assignment that is not supported by the preservation of the Bolivian material (Gayet, Meunier, 1998:fig. 25). An indeterminate percichthyid from the Upper Cretaceous of Brazil was mentioned by Gayet, Brito (1989). To the best of our best knowledge, the South American percichthyid spines and preopercles are not diagnostic to the familial or generic levels. Therefore, there is not well--supported evidence for the presence of percichthyids in the Upper Cretaceous of South America-their fossils are only confidently known from the Cenozoic.  (Fricke et al., 2018), following Jordan (1923) and Arratia (1982Arratia ( , 2003b The two species are represented by small fishes reaching a maximum length of 80 mm, Percilia irwini being a little smaller. They have an oblong or slightly elongate body covered with large ctenoid scales, relatively larger than those present in Percichthys, and with differences in scale characters, e.g., the position of the focus of the scales and length of ctenii. Percilia species lack scales on the frontal cranial region, infraorbital 1 and upper jaw; however, cycloid scales cover the opercular and cheek regions. The head profile is smoothly rounded with the snout region projecting slightly anteriorly to the lower jaw (compare figs. 95 and 108 in Arratia, 1982; see below). Their infraorbital bones, opercle, subopercle, interopercle, and posttemporal have smooth margins, whereas the preopercle is poorly serrated. These features make their identification easy when comparing with species of Percichthys. For a description of morphological features of Percilia gillisi and P. irwini see Arratia (1982).
Information concerning the biology and behavior of both species is incompletely known. There are a few reports about the reproduction and behavior of Percilia gillisi (Kilian, Campos, 1969;Chiang et al., 2011;Chiang et al., 2012) and ontogenetic changes of P. irwini (Aedo, Habit, 2009). There is concern about the present situation of the two species, which are currently catalogued as threatened in Chile.

So-called Australian percichthyids.
Under this subheading, we include all Australian genera that were interpreted as members of the family Percichthyidae by Johnson (1984; Tab. 1). However, new interpretations about them are addressed below and also in the sections Results and Discussion (see below).
The genus Macquaria currently includes Plectroplites (= Ctenolates), which was synonymized by MacDonald (1978) after examining morphological and electrophoretic characters, and Guyu Pusey, Kennard, 2001, which was informally synonymized based on the personal communication of Peter Unmack to Frickle et al. (2018). MacDonald (1978) found that the Australian percichthyid species sensu Gosline (1966) fall into two subgroups, the genus Maccullochella (see below) and the Macquaria group, the latter including Plectroplites. Cuvier (1830) described Macquaria australasica, the type species of the genus, based on specimens collected in the Macquarie River, although the species was historically distributed across both sides of the Great Dividing Range (GDR), Australia's most significant mountain range (about 3,500 km in length, up to ca. 300 km in width and 2,228 m in height) in South East Australia (Faulks, Gilligan, Beheregaray, 2010;Fig. 8). Macquaria australasica typically inhabits the cooler upper reaches of the Murray-Darling river system (MacDonald, 1978), is still present in the headwaters of this river system (Faulks, Gilligan, Beheregaray, 2010) and appears to have adapted to the environment of manmade lakes, thus is also found in some coastal streams of New South Wales and Victoria (MacDonald, 1978). Although the GDR, the most prominent geographical barrier within the distribution of M. australasica, is likely the major factor in shaping freshwater fish lineages for the last 90 million years, more recent and geographically small-scale processes must have been responsible for shaping the evolutionary history of freshwater taxa on either side of the GDR (Faulks, Gilligan, Beheregaray, 2010).
Macquaria (= Plectroplites) ambigua is a moderate to large fish known to reach 760 mm and 23 kg (Pusey, Keenan, Artington, 2004) that is commonly found in warm, turbid, slow-moving waters. It is restricted to the lower reaches of the Murray-Darling river system (Hammer et al., 2012) and inhabits some river systems of coastal south-eastern Queensland (MacDonald, 1978). In addition, it is reported to occur in coastal streams of northern New South Wales and to have been introduced into Western Australia and into streams and dams of northern coastal Australia (MacDonald, 1978;Fig. 8). Based on electrophoretic genetic differentiation, Musyl, Keenan (1992) recognized two distinct subspecies: M. ambigua oriens and Macquaria sp. B (subsp.). Macquaria sp. B also occurs in the Bulloo River and was suggested as a distinct subspecies based on the presence of a unique array of alleles (Pusey, Keenan, Artington, 2004 and citations therein).
Macquaria (= Guyu) wujalwujalensis is a small fish [maximum length of the type series being 101 mm SL (Pusey, Kennard, 2001)], limited to the Bloomfield River of north--eastern Queensland (145º20'E 15º55'S). It appears to be confined to the main channel between the Bloomfield River Falls and Roaring Meg Falls, being undetectable by snorkeling or electrofishing in other tributaries of the Bloomfield River either upstream or downstream (Pusey, Kennard, 2001). As recent phylogenetic molecular analysis placed M.
Macquaria (= Percalates) novemaculeata was originally described as Dules novemaculeatus by Steindachner (1866), and there has been considerable uncertainty about the taxonomy and systematics of the Australian bass P. novemaculeata and the morphologically similar estuary perch M. colonorum (Pusey, Keenan, Artington, 2004). Historically, these forms have been considered to represent a single species (often designated colonorum), two subspecies within a single species, or two distinct species (Pusey, Keenan, Artington, 2004 and references therein). According to Nelson et al. (2016) andFrickle et al. (2018), Percalates is considered a valid taxon, but MacDonald (1978), based on morpholo-gical and electrophoretic evidence, synonymized Percalates with Macquaria. Thus, some authors considered Percalates as a junior synonym of Macquaria (see Tabs. 1-2 and Figs. 1-2). Recently, Betancur-R et al. (2017) removed the genus Percalates from Percichthyidae (sensu Johnson, 1984) and included it in the suborder Percalatoidei within Centrarchiformes; however, both the Family and Suborder are awaiting a formal nomenclatural description in compliance with the International Code of Zoological Nomenclature (ICZN; Betancur-R et al., 2017).
Percalates colonorum is essentially estuarine, rarely being found above tidal influence, whilst Percalates novaemaculeata spends a large part of its life cycle in fresh water, periodically descending to waters under tidal influence to breed (MacDonald, 1978). These species occur in coastal southeastern Australian drainages between southern Queensland and eastern South Australia, commonly co-occurring from northern New South Wales to eastern Victoria (Trnski, Hay, Fielder, 2005; Fig. 8).  Australia's cod were described as Grystes macquariensis by Cuvier (1829), while other generic names were made available prior to 1929 (e.g., Gryptes, Gristes and Oligorus), when Whitley erected the genus Maccullochella (Berra, Weatherley, 1972). Using morphological and electrophoretic characters, Berra, Weatherley (1972) concluded that only two species (Ma. peeli and Ma. macquariensis) were found in southeastern Australia. Later, Jerry et al. (2001) suggested and Nock et al. (2010) confirmed that four extant taxa should be recognized within the genus: Ma. macquariensis, Ma. peelii, Ma. mariensis, and Ma. ikei. Maccullochella spp. occur on both sides of the GDR, which has had a key role in the freshwater fish speciation in eastern Australia (Fig. 8). The Murray River cod (Maccullochella peelii) and trout cod (Ma. macquariensis) are endemic to the Murray-Darling Basin on the inland side of the GDR and the eastern freshwater cod (Ma. ikei), whereas the Mary River cod (Ma. mariensis) persists in isolated coastal catchments to the east of the GDR (Harrisson et al., 2016 and references therein). Species of Maccullochella are highly regarded for their cultural, historical, angling and edible characteristics, and there have been significant declines in the distribution and/or abundance of each species, which are classified as vulnerable (Ma. peelii) or endangered (the other species; Nock et al., 2010).
The taxonomic placement of Gadopsis had proven problematic until Johnson (1984) placed it within Percichthyidae and interpreted it as more closely related to the Australian genera Bostockia, Nannatherina, and Nannoperca, than to other percichthyids. In the past, Gadopsis was first included within the Blennidae by Richardson (1848), an assignment that was followed by Steindachner (1866) and Goodrich (1909). Regan (1913) established a division within the Percoidei to contain Gadopsis based mainly on the absence of the entopterygoid, the jugular position of the pelvic fins (each reduced to a pelvic spine and a bifid soft ray), and 48 vertebrae, an assignment that was later followed by Jordan (1923). Berg (1947) placed Gadopsis in its own superfamily, the Gadopsoidae, whereas Scott (1962) placed it in its own order, the Gadopsiformes or it was included in its own family, Gadopsidae, within the Percoidei (Greenwood et al., 1966) or Ophioidei (Gosline, 1968).
Although Gadopsis is known by two species, new molecular evidence has suggested that Gadopsis marmoratus represents a complex of several cryptic species (Unmack et al., 2017). This is one of the largest temperate perch that may reach 60 cm in length (Bray, 2011). Species of Gadopsis are restricted to rivers and streams of southeastern Australia, including Tasmania (Hammer et al., 2014;Unmack et al., 2017;Fig. 9).
Bostockia porosa is a moderately small fish (16 cm maximum length, commonly reaching 9 cm SL) with nocturnal habits (hence the common name of "nightfish"), endemic to the streams of the southwestern drainage area in Western Australia (MacDonald, 1978;Fig. 9). Little is known regarding B. porosa, although its biology has been described based on monthly data obtained over three years during the 1980's by Pen, Potter (1990). Bostockia porosa sampled had most age classes belonging to 0+ and 1+ (81.8 and 12.5% of the 2,463 fish caught), and hardly reaches 6+ years old (only one individual of this age was found). Sexual maturity is attained commonly at the end of the first year for males and at the end of the second year for females. This species is carnivorous, feeding on a wide range of benthic organisms (Pen, Potter, 1990).
Originally Nannatherina balstoni was placed in the Atherinidae, but later was positioned in the Kuhliidae (Regan, 1940), until Johnson (1984) placed it in the Percichthyidae. Nannatherina balstoni is the rarest of all the endemic freshwater fishes of southwestern Australia (Morgan, Hill, Potter, 1998;Fig. 9). It is moderately abundant in a few shallow pools and creeks that dry up in summer, and in winter and spring this species typically is found amongst inundated riparian vegetation, where it presumably feeds and spawns. It is generally restricted to the small area of coastal peat flats in the southwestern corner of Australia and to two populations that have recently been found well to the north of this area (Morgan, Hill, Potter, 1998); however, the contemporary area of occurrence is now ca. 69% of its historical distribution (Morgan et al., 2014).
Edelia Castelnau, 1873-a synonym of Nannoperca according to Jerry et al. (2001)-was described with E. obscura and E. vittata, which are currently assigned to the genus Nannoperca. All these species are considered "pygmy perches" due to their small size that may reach a maximum length of 7 or 8 cm. Among these, Nannoperca pygmaea reaches a maximum length of 4 cm. Classification of pygmy perches has long posed problems for taxonomists. Over the years these species have been placed in Nannopercidae, the Kuhliidae, within the kuhliid subfamily Nannopercinae (see Jerry et al., 2001 andHoese et al., 2006). Johnson (1984) interpreted the genera Nannatherina, Nannoperca and Edelia as part of the Percichthyidae.
Recent molecular studies suggested that there are cryptic species within Nannoperca (Unmack et al., 2011;. A phylogenomic study published recently confirmed the paraphyletic nature of N. vittata, revealing two new cryptic lineages within the species and recognizing informally the existence of N. 'flindersi' (Buckley et al., 2018). However, Buckley et al. (2018) emphasized that a more comprehensive sampling may reveal the full geographic range of the delimited species, including morphological and ecological characters for establishing species boundaries.
These species are endemic to the southeast (e.g., Murray-Darling Basin and coastal drainages) and the southwest of Australia and have a puzzling biogeographic disjunction across southern Australia (Buckley et al., 2018;Fig. 9). Pygmy perches prefer lowland habitats, including low gradient rivers and creeks, as well as floodplain habitats with aquatic macrophyte cover in general, and occur in virtually all drainages inhabited by fishes across southern Australia (Unmack et al., 2011). 14 e180102[14]

Material
For institutional acronyms see Sabaj (2016). The material of South American and Australian temperate basses examined is listed below in alphabetic order. sp = specimen; cl&st = cleared and stained specimen. Considering the taxonomic discrepancies concerning certain species, the possible synonyms are identified below. High-level classification follows Betancur-R et al. (2017).   Methods. Ethanol specimens were included for studies on the external morphology of fishes and its squamation. Muscles were studied in dissected specimens; dissections were done by GA on the left side of specimens. Ligaments and tendons were observed with stereomicroscopes equipped with polarized light and phase contrast. Osteological structures were studied in cleared and double stained specimens for cartilage and bone following the techniques of Dingerkus, Uhler (1977) and Arratia, Schultze (1992). Scales from comparable body regions were studied on ethanol specimens; some scales were stained with a weak solution of alizarin to observe details under an Olympus compound microscope equipped with polarized light and phase contrast.
Phylogenetic methodology for morphological characters. Our major difficulty for this analysis is the controversy between the composition and taxonomic placement of percichthyids, perciliids and other percomorphacean taxa (see Tabs. 1, 2). For instance, between percichthyids sensu Johnson (1984) and other percomorphacean families and genera incertae sedis, the only morphological phylogeny published on the subject is that of Micklich (1987: fig. 16; see Fig.  1 herein), and percichthyids and perciliids were considered within the order Perciformes, as in Arratia (1982) and Johnson (1984). The most recent molecular hypotheses interpreted percichthyids as part of the Centrarchiformes (Betancur-R et al., 2017;Arcila et al., 2017;Hughes et al., 2018). Considering the current information concerning the placement of percichthyids, we use in morphological comparisons different percomorphaceans belonging to a variety of families including other centrarchiforms, such as species from Acropomatidae, Centrarchidae, and Percalatoidei and further representatives, for instance Moronidae (see list of comparative material). Species of Morone and Macquaria (= Percalates) were used as the outgroup (see Material studied).
The phylogenetic analysis was conducted using TNT version 1.5 (Goloboff, Catalano, 2016) and tested with PAUP (Phylogenetic Analysis Using Parsimony) software (version 4.0b10) for 32-bit Microsoft (Swofford, 2000). In TNT, the optimal score was hit 20 times independently, using xmult plus 10 cycles of tree-drifting, while a heuristic search using a random addition sequence with 1000 replicates, and the three bisection and reconnection (TBR) branch-swapping algorithm was performed in PAUP. The character matrix was constructed using MacClade for the analysis to run in PAUP. All characters are unweighted, unordered, and considered to be independent of one another. The phylogenetic analysis used a list of 54 characters for 12 taxa. The data were analyzed in PAUP, including both ACCTRAN and DELTRAN optimization. Bootstrap support was done with 1000 replicates by TBR-swapping the trees found in TNT. Mitogenomic analyses. We compiled published mitogenomes considered as part of Percichthyidae according to current molecular hypotheses Betancur-R et al., 2017)  Additionally, we sequenced the whole mitogenome of Percilia gillissi from the Maipo River (type locality), a second sample from south of Chile, and Percilia irwini from the Bureo River, Bío-Bío Basin, following Quezada-Romegialli et al. (2016a, 2016b. Briefly, total genomic DNA was extracted from dorsal muscle (Aljanabi, Martínez, 1997) and purified with the DNA Clean and Concentrator Kit (ZymoResearch, Irvine, CA, USA). Total DNA was sent to a next-generation sequencing facility (OMICS-Solutions, Santiago, Chile) and subjected to a shared run on an Ion Torrent PGM Sequencer (chip 318). Total reads per each sample were trimmed (at 5' and 3' low quality ends), filtered by quality (Q15/90% with FASTX--toolkit: http://hannonlab.cshl.edu/fastx_toolkit) and mapped to the closely related species Percilia irwini (Genbank accession no. AP014528; Lavoué et al., 2014) in Geneious v8.0.5 (Biomatters Ltd., Auckland, New Zealand). Annotations of 13 protein-coding genes, ribosomal units and tRNA genes were confirmed in DOGMA (Wyman et al., 2004) and MitoFish .
Considering the 34 taxa, separate alignments were done with MAFFT (Katoh et al., 2002; algorithm FFT-NS-i x 1000, scoring matrix 1PAM / k = 2; excluding the control region) for (i) the 12S and 16S rRNA sequences, (ii) all the concatenated 22 tRNA genes, and (iii) the protein-coding genes. The protein-coding genes alignment was checked by eye with respect to the translated amino acid sequences. We inferred phylogenetic trees for three different character matrices: first, a partitioned data matrix concatenating the 13 protein-coding genes except third codon positions (7712 positions), 22 tRNA genes (1545 positions) and the two ribosomal RNA genes (2921) as an initial matrix (Mt1, three partitions). Second, a partitioned data matrix with the first (3856 positions) and the second codon positions (3856 positions) of 12 protein-coding genes concatenated, 22 tRNA genes (1545 positions) and the two rRNA genes (2921 positions; following Lavoué et al., 2014) as Mt2 (four partitions). Finally, a third data matrix (Mt3) was partitioned according to the best fit partitioning scheme calculated with PartitionFinder2 (Lanfear et al., 2016), considering as input the three codon positions of the protein-coding genes per separate (29 data blocks), plus the 22 tRNA genes (1 data block) and the two rRNA genes (1 data block). We searched for maximum likelihood (ML) phylogenetic trees using RAxML (Stamatakis, 2014) at the CIPRES Science Gateway (Miller, Pfeiffer, Schwartz, 2010) under a GTR+ Γ model (general time reversible model, following a discrete gamma distribution) and the rapid bootstrapping and search for best-scoring ML tree as search algorithm, with 1000 bootstrap replicates for assessing robustness of the internal branches.
Anatomical Terminology. The terminology of the skull roof bones follows the traditional terminology that differs from that based on homologization (e.g., Schultze, 2008). The skull roof traditional terminology is fundamentally based on human anatomy. For instance, the bone called "frontal" in traditional terminology actually corresponds to the parietal bone in actinopterygians. Identification and nomenclature of cephalic sensory canals follows Northcutt (1989). Identification and terminology of muscles follow Datovo, Vari (2013, 2014, with comparisons of previous terminologies used by Winterbottom (1974), Stiassny (1981), and Johnson (1984). We indicate some of the differences below. Scale types follow Schultze (1966) and Roberts (1993), even though we are not dealing specifically with different types of ctenoid scales, but those of percichthyids (McCully, 1970 [-]. Presentation of characters follows body regions, except for a few miscellaneous ones presented at the end of this section. Most characters from Johnson (1984) are considered below; however, the construction of each character, including its character states, is based on our polarization depending on our outgroup (see Material and Methods). Comments. Fossil stem teleosts may have ganoid scales of lepisosteid-type and amioid-type (Schultze, 1966), whereas the presence of cycloid scales is a synapomorphy at the level of Leptolepis coryphaenoides plus more advanced teleosts or "true" teleosts (Arratia, 1997(Arratia, , 2015(Arratia, , 2017. The crown teleosts share the presence of cycloid scales and its further transformations (e.g., ctenoid and serrated scales) in some euteleost groups. The posterior exposed field of a scale may be smooth or with a serrated or pectinated posterior border or bear small ctenii (Schultze, 1966;Roberts, 1993;Fig. 10). In strict terms, scales bearing ctenii are ctenoid scales; however, pectinated and serrated scales have also been named ctenoid.

Scales
Both ctenoid and serrated scales develop from cycloid scales, and they represent evolutionary transformation of the cycloid pattern (Schultze, 1966). This is also supported by ontogenetic changes, as shown by small specimens bearing cycloid scales that develop ctenii in their posterior field. Ctenoid scales (Fig. 10) may lose their ctenii in the oldest individuals-as shown in large Percichthys chilensis and P. melanops-but always keep remnants of the ctenoid condition. Radial furrows or radii may develop in the anterior field of cycloid scales and are consistently present in percichthyids bearing ctenoid scales (Fig. 10). Ctenoid scales may adopt a variety of shapes and modifications of the ctenii within percomorphaceans (Robert, 1993), which are features in need of further investigation.
The presence of scales with simple, only slightly amputated needle-like ctenii (Fig. 10) was noticed first by McCully (1970) and later was proposed as a synapomorphy of percichthyids by Johnson (1984:p. 469). It is interesting to clarify that this feature is absent in some of the Australian temperate basses or perciliids with only cycloid scales (e.g., Bostockia, Gadopsis and Nannatherina; see Tab. 3), and their scales were interpreted as secondarily cycloid by Johnson (1984), but see section of Results. The ctenii develop as needle-like with sharp tips, and they lose their tips during growth. Thus, needle-like ctenii are observed at the posterior margin of the scale, whereas oldest rows near the focus of the scale present truncated ctenii. Comments. The lateral line may be continuous and positioned through a complete series of scales along the body (the common condition observed in percichthyids), or it may be interrupted or continuous, but runs throughout an interrupted series of scales (e.g., Percilia; Arratia, 1982). Ramifications of the lateral line at the level of its maximum curvature have been observed in some individuals of Percichthys colhuapiensis. A reduced, interrupted or absent lateral line canal is present in the Australian genera Nannoperca, Nannatherina, and Bostockia among the studied fishes (see also Johnson, 1984). In contrast, Gadopsis has a continuous lateral line (see Tab. 3).

Position of lower jaw versus anterior margin of snout.
Ch. 6. Anterior tip of the lower jaw: projecting in front of premaxilla/upper jaw [0]; premaxilla placed at the same level [1]; slightly posterior to premaxilla [2]. See Figs. 11 and 12. Comments. A lower jaw (Fig. 11) that projects anterior to the anterior margin of the snout is the generalized condition among the studied percomorphaceans. Among the studied specimens, three conditions are present: 1) The lower jaw extends anterior to the premaxilla (e.g., Bostockia diagramma, Macquaria (= Percalates) colonorum, Percichthys colhuapiensis, and Macquaria (= Plectroplites) ambigua; Figs. 11b-d). This pattern represents the primitive character state by comparison to the outgroup (Fig. 11a).
3) The premaxilla projects anterior to the lower jaw (e.g., Gadopsis marmoratus, Maccullochela macquarensis and Percilia gillissi; Figs. 12a, d); however, in certain individuals of Percilia the lower and upper jaws are at the same level.
Given the differences on the position of the premaxilla relative to the lower jaw, the osteology and ligaments of the anterior part of the head of all mentioned fishes were investigated without finding synapomorphies that could characterize the percichthyids; though variation among the fishes was used for comparisons. Tab. 3. Character-taxon data matrix for 54 characters used in the phylogenetic analysis. State [N] represents the character is not applicable. State P represents a polymorphic character coded as 0/1 in the matrix.  Fig. 13.

e180102[18]
Comments. In most examined taxa, the quadrate-mandibular articulation (Fig. 13b) is placed below the posterior half of the orbit or posterior to it (e.g., Gadopsis, Maccullochella, Nannatherina, and Percichthys colhuapiensis). In Percilia, the quadrate-mandibular articulation is placed below the anterior margin of the orbit (Fig. 13d), a condition also observed in Bostockia. In Percichthys chilensis and P. melanops, the articulation is placed at the mid-region of the orbit or posterior to it (Fig. 13c) so that the lower jaw is short. Among Percichthys, the quadrate--mandibular articulation is placed below the middle region of the orbit or posterior to it. Consequently, the longest lower jaw, is present in P. colhuapiensis (Fig. 13b) with the quadrate-mandibular articulation posterior to the orbit. The length of the lower jaw justifies its common name "perca bocona" (= perch with a big mouth), a feature used to characterize this percichthyid (MacDonagh, 1955;Ringuelet et al., 1967). Comments. Although the maxilla presents differences in the shape of the articular process for the palatine, or in the curvature between the articular anterior processes and the body of the palatine, we could detect no significant character that could be unique to percichthyids. However, the length of the maxilla may be of some value. A maxilla whose posterior end lies below the anterior half of the orbit or in front of the orbit is interpreted here to represent the plesiomorphic condition (e.g., in Bostockia, Gadopsis, Macquaria australasica, and Percichthys melanops; Figs. 11, 13c) by comparison with the outgroup (Fig. 11a). In contrast, a maxilla that extends below the middle or posterior half of the orbit is interpreted here as an advanced feature (e.g., Nannatherina, Percichthys colhuapiensis, Maccullochella macquarensis, and Macquaria (= Plectroplites) ambigua; Figs. 11, 12). The length of the maxilla varies in Percichthys colhuapiensis; it ranges from the posterior half of the orbit to posterior to it (Figs. 4d, 11c, 13b). The position of the posterior end of the maxilla, and therefore the maxillary length, may be of taxonomic value; for instance, in Percichthys melanops the maxilla is short, usually reaching the anterior rim of the orbit (Fig. 12); in contrast, the maxilla is longer, reaching the anterior half of the orbit in Percichthys chilensis (Figs. 4a-c Comments. The mandibular canal extending from the posterolateral side of the anguloarticular to the anterior part of the dentary exits to the skin through a series of pores (most examined species). The external exits of the mandibular canal perforate the skin in young specimens (e.g., Macquaria (= Plectroplites) ambigua, Percichthys colhuapiensis, P. chilensis, and P. melanops). However, in large specimens the count of pores and their size may be difficult, because the fish may still have all pores opened or a few of them closed (e.g., individual of 345 mm standard length of Percichthys colhuapiensis). Therefore, the study of the pores of the mandibular canal opening on the skin may be misleading if someone does not consider their ontogenetic variation.
Those regions of the skin of the lower jaw where pore(s) will open during ontogeny are associated with oval or round depressions (Figs. 14, 15) that are easy to observe in an external examination of the skin. Three or four depressions are commonly present in percichthyids. Percilia (Fig. 14d), as well as Nannatherina and Nannoperca, does not have depressions. Less than three depressions are found in Morone ( Fig. 14a; see Tab. 3).
As above stated, pores begin to perforate the skin of the depressions of the lower jaw during growth. The common condition found in the taxa used for comparisons is the presence of numerous small pores irregularly placed in such depressions (e.g., Acropoma, Synagrops, and Lateolabrax); this condition is also present in Macquaria (= Percalates) colonorum (Fig. 14b). The pores may be slit--like or oval-shaped and placed at the posterior corner of a depression (e.g., Maccullochella, Macquaria australasica, Macquaria (= Percalates) ambigua and Percichthys; Figs. 14, 15), or round (Percilia; Fig. 14d), or there may be a combination of slit-like pores and small numerous pores (e.g., Macquaria (= Percalates) colonorum; Fig. 14b).
The presence of "enlarged sensory pores of the dentary" is a character mentioned by Johnson (1984:p. 469) for Plectroplites, Macquaria, Maccullochella, Percichthys, and Percilia. However, this statement is not supported by the specimens studied and illustrated in Figs. 14-16 herein.

Ch. 13. Pore 1 or the most anterior of the mandibular sensory canal placed on the skin: at the most anterior portion of the dentary [0]; absent [1].
Comments. The first pore of the mandibular canal usually is round or oval, and it may open late in ontogeny (e.g., Percichthys chilensis, P. melanops, and P. colhuapiensis). The first pore is on the skin that covers the anterior part of the dentary. The first pore location varies. It may be positioned anteriorly or ventrally in the jaw (Figs. 14-16). When the first pore has an anterior placement, it is not observed in a ventral view of the head, with the exception of Maccullochella (Fig. 14c). The anterior position of the pore represents the generalized condition among the studied fishes. The first pore is absent in Percilia, Bostockia, Nannatherina, and Nannoperca (e.g., Figs. 16a, b). Comments. The mandibular canal (Figs. 12a, d, 16a) is placed in the dentary and the angular portion of the anguloarticular, enclosed in a bony tube that has openings to the skin (most common condition). The mandibular canal (Figs. 12b,c,16b,c) is placed in a groove on the lateral aspect of the dentary and anguloarticular in Bostockia, Macquaria australasica, Percichthys chilensis and P. melanops (see Tab. 3).

Ch. 14. Mandibular canal is: enclosed in a bony tube in
There are some differences in the size of the openings in the dentary that may be of taxonomic value. For instance, Percilia irwini has the most enclosed canal among all examined taxa, and in Percilia gillissi, the pores are comparatively larger.      Arratia (1982). ang-ar = anguloarticular; de = dentary; p1-6 = pores 1-6; rar = retroarticular.

Skin, infraorbital canal and infraorbital bones.
Comments. The pores of the infraorbital canal that open through the skin of infraorbital bones 1 and 2 have different shapes within the studied fishes. Usually they are round (the generalized condition), however, slit-like or oval pores are characteristic of percichthyids.
The infraorbital sensory canal is partially or totally enclosed by bone in infraorbital 1 in most examined taxa (Figs. 13a, b, 17); however, it is in a groove, only partially protected by a small ossification at its anterior portion in fossil and extant species of Percichthys chilensis, P. melanops and P. colhuapiensis (Figs. 13b, c). A similar condition is observed in †Santosius antiquus from the Tertiary of Brazil (Arratia, 1982:text- fig. 91).
The infraorbital canal associated with infraorbital bones 3 to 5 may exit to the skin by short, simple and few (two or three, usually) tubules, each opening to the skin through a round pore (e.g., Percilia gillissi and P. irwini), or ramified tubules (Figs. 11d, 12b) exit through many pores (e.g., Macquaria australasica, Macquaria [= Plectroplites] ambigua) and specimens used in comparisons.
Comments. Six infraorbitals (including the most dorsal bone that may correspond to the dermosphenotic at the junction of the supraorbital, infraorbital, and otic canals; otic and postotic canals were named temporal canals by Vega, Viozzi, 2016) represent the common condition found in most fishes examined herein (Figs. 13b, c). In contrast, Percilia commonly presents more infraorbitals (7 or 8), because of the presence of an additional bone-bearing canal between infraorbitals 1 and 2. Arratia (1982:p. 34) interpreted the presence of a high number of infraorbitals in Percilia as the primitive condition from which would be a reduction of elements due to fusion. Additional infraorbitals are also observed in Nannatherina and Nannoperca (Fig. 17). The high number of infraorbitals is re-interpreted herein as an apomorphic condition by comparison with the outgroup and other fishes examined (see Material studied). A reduction of infraorbitals is found in some individuals of Percichthys trucha and P. melanops (Arratia, 1982) that is due to fusion of infraorbitals 3 and 4 or 4 and 5, as illustrated by the size of the bones and the presence of fusion lines.

e180102[26]
Comments. The absence of a subocular shelf was mentioned first by Johnson (1984) for Bostockia, Gadopsis, Nannatherina, and Nannoperca. The study of this structure has shown that when the subocular shelf is present, it may be associated with infraorbital 2 or 3 in different species, indicating that the bone bearing the shelf should be identified to make comparisons possible, and consequently, homology statements. Comments. One medial pore associated with both epiphyseal tubules (named "supraorbital commissure" by Vega, Viozzi, 2016) of the supraorbital canal is present in most examined taxa. The median epiphyseal pore is placed between both orbital margins of the frontal bones, at the mid-region of the preorbital region in Percilia. It lies at the level of the autosphenotic or posterior to the preorbital region of the frontal bones in most examined taxa. In some of the species used for comparisons, the pore is confused among the ramifications and pores of the supratemporal canal (e.g., Morone and Percalates [= Macquaria] colonorum). Ch. 23. Basisphenoid: present [0]; absent [1]. Character from Johnson (1984).
Comments. Among the studied taxa, Bostockia and Gadopsis share the absence of the basisphenoid.
Ch. 24. Palatine (auto + dermopalatine) extending posterior to the orbital process of the lateral ethmoid by: a short posterior process or no process [0]; a long posterior process [1]. See Fig. 18. Comments. The auto plus dermopalatine (the so-called palatine) is a compound bone that results from the early ontogenetic fusion of the chondral autopalatine that originates at the anterior corner of the palatoquadrate, the pars autopalatina (Arratia, Schultze, 1991), and a dermal component bearing teeth, which is named dermopalatine. An elongate auto + dermopalatine or palatine (Fig. 18a) that extends posterior to the lateral ethmoid is present in all examined taxa, except Nannoperca and Percilia (Fig. 18b). In Percilia and Nanoperca (as well as in the outgroup), the palatine is only represented by its anterior section that articulates anteriorly with the maxilla and posteriorly with the lateral ethmoid. The articulation between palatine and lateral ethmoid is a hinge joint in all studied percomorphaceans.
Ch. 25. Palatine (auto + dermopalatine) bearing: few or no teeth [0]; numerous teeth [1]. See Fig. 18. Comments. All taxa examined herein-with the exception of Nannoperca, Percilia, and the outgroup, Morone-have numerous teeth along the complete or almost complete length of the palatine (Fig. 18a). In Percilia, a few (1 to 5, commonly 2) are present on the ventrolateral margin of the palatine (Arratia, 1982:text-figs. 23, 24); the teeth are so small and scarce in number that former authors described Percilia as lacking teeth on the palatine and vomer (Eigenmann, 1927;Fowler, 1951).

Comments.
A similar vertebral number (29-33) was interpreted as a synapomorphy of Bostockia, Nannatherina, and Nannoperca by Johnson (1984). However, and based on the variation present in the studied taxa, the character is re-interpreted here with three states. Among the Australian taxa, Gadopsis has 48 or 49 vertebrae. The number of vertebrae of the Chilean species of Percichthys and Percilia varies between 30 and 36 (Arratia, 1982:tab. 1).
Comments. The absence of the supracleithral sensory canal or lateral line was proposed as a synapomorphy of the Australian genera Bostockia, Nannoperca, and Nannatherina (Johnson, 1984) and also Gadopsis (see Tab. 3 for distribution of this character).

Ch. 36. Postpelvic process in basipterygium: well defined and elongate [0]; short or rudimentary [1].
Comments. Gosline (1966) considered a relatively long postpelvic process a diagnostic feature of his Percichthyidae. A well-defined elongate process (ca. 5 to 10 times in the basipterygium length) is common to the outgroup and most of the studied taxa, with the exception on the Australian Bostockia, Gadopsis, and Maccullochella, which have a short or rudimentary process. See Arratia (1982:tb. 5 Comments. Among the studied taxa, the greatest numbers of dorsal fin rays are found in Bostockia (with more than 16 rays) and Gadopsis (with about 30). A reduced number of dorsal spines was suggested as a synapomorphy of Bostockia, Nannatherina, and Nannoperca by Johnson (1984). Two or three dorsal spines are present in Gadopsis (among the studied taxa), a feature that has been used to distinguish G. marmoratus (commonly 3) and G. bispinosus (commonly 2).

Comments.
A procurrent spur at the ventral margin of the base of the posteriormost ventral procurrent ray is a character present in most perciforms (Johnson, 1975). However, a spur is missing in several taxa studied here, e.g., Bostockia, Nannatherina, and Nannoperca (see also Johnson, 1984). In others, its presence may be intraspecifically variable, as for instance in Percichthys chilensis and P. melanops, with a very small spur when it is present. A spur is absent in most specimens of Percilia, and when it is present, is rudimentary.
Comments. Reduced numbers of procurrent rays was mentioned as a synapomorphy shared by Bostockia, Nannatherina, and Nannoperca by Johnson (1984;p. 469), without establishing differences between the dorsal and ventral series of procurrent rays. Thus, the character has been changed, and both series are treated as separated for comparative and homology purposes.

Ch. 47. Promalaris: absent [0]; a fleshy connection unites it the main muscle mass of pars malaris [1]; separated from pars malaris and has a tendinous connection to hyomandibula and maxilla [2]. See
Comments. In Percichthys trucha and P. melanops the promalaris may be separated from pars malaris (Fig. 20e) and connects by independent tendons to the hyomandibula and maxilla. Promalaris may have a fleshy attachment to the main section of pars malaris (e.g., Maccullochella macquarensis and Percichthys colhuapiensis; Fig. 20c) and to the buccopalatal membrane plus imbedded ligaments (e.g., Macquaria australasica; Fig. 20g Fig. 20. Comments. Commonly, the levator arcus palatine (Fig. 20) is a small muscle in all taxa examined, except in Maccullochela macquarensis and Percichthys colhuapiensis (Fig.  20c). In these fishes, the muscle is particularly large, with most fibers lying almost 90 degrees to the cranial roof bones. The enlargement of the muscle seems to be correlated with the enlargement of the autosphenotic. In addition, the position of a separate section of the dilatator operculi (Fig. 20c Johnson (1984).  Ch. 51. Tubular anterior nostril placed near the margin of the lip: absent [0]; present [1]. Character from Johnson (1984).

Comments.
A distinctive nasal rosette and a tubular anterior nostril placed near the margin of the upper lip are morphological features shared by Bostockia and Gadopsis and were first mentioned by Johnson (1984). A similar anterior nostril is also present in Maccullochella peeli (Fig. 12a) Johnson (1984).

Comments. A premaxillary frenum has been observed in
Gadopsis and Maccullochella among the studied fishes.
3) Slightly demersal and non-adhesive (e.g., Macquaria [= Plectroplites] ambigua and probably other species of Macquaria [Lake, 1959]). 4) Eggs of Percilia gillissi are deposited in the gastral cavity of the freshwater sponge Spongila igloviformis where they develop (Kilian, Campos, 1969). To the best of our knowledge no percichthyid or centrarchiform is known for this behavior.
Phylogenetic analysis and results: morphological characters. Phylogenetic Analysis. To assess the phylogenetic relationships of Percilia and percichthyids, we have performed a cladistic analysis including 12 taxa and 54 morphological characters. The coding of 54 characters is presented in Tab. 3. Two parsimony analyses were performed using Morone as the outgroup. The parsimony analysis recovered one tree of 116 steps (retention index = 0.647; consistency index = 0.566). Figure 21 represents the topology of the tree found, which suggests two well-supported clades identified here, in a preliminary way, as Perciliidae (Node B) and Percichthyidae (Node C).
Node B, Perciliidae. The monophyly of the family Perciliidae is supported by eight synapomorphies: Absence of scales on the skin of maxilla (Ch. 4[1]); series of scales carrying the lateral line canal interrupted or reduced (Ch. 5[1]); without large depressions on the skin that covers the dentary (Ch. 9[1]); with a few independent round pores on the skin of the dentary (Ch. 12[1]); pore 1 or the most anterior of the mandibular sensory canal placed at the most anterior portion of the dentary absent (Ch. 13[1]); with more than five infraorbital bones (Ch. When the phylogenetic analysis included Macquaria (= Percalates) as the outgroup, there was no change in the topology of the tree (Fig. 21), with the exception that Percalates was included in a polytomy at node A.
Phylogenetic analysis and results: Mitogenomics. A total of 303,853 (mean 240 bp, sd 88 bp), 324,606 (mean 243 bp, sd 100 bp) and 308,889 (mean 228 bp, sd 97 bp) reads were sequenced for the samples of Percilia gillissi from the Maipo River and southern Chile, and Percilia irwini from the Bureo River, respectively. The complete mitochondrial genomes of Percilia gillissi were assembled with 372 (mean coverage 5x) and 292 (mean coverage 4x) reads, whereas the mitogenome of Percilia irwini had 717 (mean coverage 10x) reads mapped. Gene composition and structure of the newly assembled mitogenomes were similar to Percilia irwini (Genbank accession no. AP014528; Lavoué et al., 2014), and comprised of 13 protein-coding genes, 2 ribosomal RNA units, 22 tRNA genes and a control region, the latter being excluded for further analyses. The length of these mitogenomes were 16,622 bp (Maipo River sample), 16,624 bp (southern Chile sample) and 16,624 (P. irwini from Bureo River). Genbank accession numbers are MK239311, MK239310 and MK239312, respectively.

Discussion
Percichthyid and perciliid relationships. The results presented herein are the first published effort to analyze the possible phylogenetic relationships among fishes currently interpreted as percichthyids sensu Johnson (1984) that includes South American and Australian taxa. Prior research established the possible content of the families Percichthyidae and Perciliidae (Arratia, 1982;Micklich, 1987) or only Percichthyidae (Johnson, 1984) without a formal phylogenetic analysis and hypothesis of possible relationships. According to Johnson (1984:p. 469), "The monophyly of Percichthyidae is supported by a series of nested synapomorphies only a few of which are mentioned here. The scales of most of those genera are similar and unlike those of the excluded genera in having the posterior field filled with simple only slightly amputated (see McCully, 1970) needle-like ctenii (those of Bostockia, Gadopsis and Nannatherina are secondarily cycloid)." The parsimony analysis (TNT) does not interpret this feature as a synapomorphy of Perciliidae and/or Percichthyidae, an interpretation also supported by PAUP analysis. Nannoperca has ctenoid scales as in Percilia and percichthyids; thus, the absence of ctenoid scales is interpreted as a synapo-  Tab. 3 for distribution of characters). Johnson (1984:p. 469) provided 13 morphological characters supporting the taxa that he assigned to the family, and added five more specific characters (p. 470) supporting a probable relationship between the Australian Bostockia and Gadopsis and one shared by Bostockia and Maccullochella. From those, only two characters were not considered in the present study: (1) a reduced number of branchiostegal rays, because there is not much variation among the studied taxa, just one ray that can be variable (see Arratia, 1982:tb. 1); and (2) absence of a medial tabular (= medial extrascapula; only in Gadopsis and Bostockia).
The phylogenetic analyses performed herein differs from Johnson's (1984) hypothesis in recognizing two clades that we identify preliminarily as Perciliidae and Percichthyidae ( Fig. 21: nodes B, C). The two families are supported by numerous morphological characters. In addition, each branch has bootstrap support over 90 % (see Fig. 21).
According to this new hypothesis, the family Perciliidae includes the Australian genera Nannoperca, Nannatherina, Bostockia and Gadopsis with the South American genus Percilia at its most basal node (Fig. 21: node B). The enigmatic genera Bostockia and Gadopsis are interpreted as perciliids; they are sister to each other, as already suggested by Johnson (1984), and the most advanced taxa within the family. On the other hand, the family Percichthyidae has a new understanding, including the South American genus Percichthys and the Australian Macquaria australasica, Macquaria (= Plectroplites) ambigua, and Maccullochella.
The new phylogenetic hypothesis (Fig. 21) based on morphological characters is not congruent with previous molecular studies (e.g., Near et al., 2012Near et al., , 2013Chen et al., 2014;Lavoué et al., 2014;Betancur R et al., 2017)-where Percilia and Percichthys are part of the same clade, the family Percichthyidae sensu Johnson (1984). Those results are not confirmed either by the mitogenomic analysis performed herein (Figs. 22 and S2a-b). For example, Bostockia and Gadopsis (Fig. 21: node B3) are not recovered as sister taxa in any molecular study (i.e., Figs. 2, 22, and supplementary S2a-b). According to molecular hypotheses, percichthyids such as Maccullochella spp. are commonly found as sister taxa with perciliids (i.e., Gadopsis spp.; Near et al., 2013;Lavoué et al., 2014;Betancur-R et al., 2017;Figs. 2d, g, i-j). Nevertheless, it is important to note that a major problem is the differences in the taxonomic sampling used in the various studies. The morphological analysis performed here is the only one that includes representatives of all genera suggested as percichthyids by Johnson (1984). Consequently, the results between morphological and molecular analyses are not comparable, and even the results among the differently sampled molecular studies are not comparable.
Taxonomy of percichthyids and perciliids. The content and understanding of the so-called family Percichthyidae have been controversial from its beginning when Jordan (1923) created the family (see Tabs. 1 and 2). Over the decades, only a few papers revised the taxonomy and/or morphology of the South American percichthyids (e.g., Ringuelet et al., 1967;Arratia, 1982;López-Arbarello, 1999 and Australian ones (e.g., Berra, Weatherley, 1972;MacDonald, 1978;Johnson, 1984;Nock et al., 2010). In the most recent morphological and taxonomic reviews of the Argentinian species of Percichthys, only three species were recognized: Percichthys trucha, P. colhuapiensis and P. laevis (López-Arbarello, 1999, whereas only one species of Percichthys was recognized by the phylogeographic study of Ruzzante et al. (2006), an approach followed later by other researchers. Thus, under this paradigm, the genus Percichthys would be represented by two species in South America, P. trucha inhabiting both Andean slopes and P. melanops restricted to central-southern Chile (Fig. 3b). However, the results of this study-based on morphological characters-suggests that Percichthys colhuapiensis, as well as Percichthys chilensis, are valid species and should be re--investigated. Percichthys colhuapiensis is characterized by two autapomorphies: dilatator operculi placed almost horizontal to the main section of the muscle and a large subdivision of the levator arcus palatine muscle with fibers almost 90 degrees to the cranial roof bones (Fig. 20c). In addition, the position of a separate section of the dilatator operculi dorsal to the lateral surface of the autosphenotic and pterotic is a character shared by Percichthys colhuapiensis and Maccullochella macquariensis among the examined percichthyids. The presence of a large muscle mass formed by the ricto-retromalaris (rictalis + malaris) is another character shared by P. colhuapiensis and Maccullochella.
Currently, the taxonomy of the Australian percichthyids needs revision. MacDonald (1978), based on an electrophoretic analysis of protein variation at 19 genetic loci, proposed Percalates and Plectroplites as a synonym of Macquaria and recognized Maccullochella as a valid genus that differs extensively from Macquaria. However, as discussed in the previous sections, Macquaria and Percalates should be revised, because they have different positions and relationships in previous molecular studies (see for instance or Betancur R et al., 2017 and herein (Fig. 2), as well as the morphological phylogenetic analysis performed herein. The family Perciliidae was created by Jordan (1923) for the Chilean genus Percilia, an approach followed later by a few others. The results of the morphological phylogenetic analysis (Fig. 21) suggest that representatives of the genera Nannatherina and Nannoperca and the enigmatic Bostockia and Gadopsis are also members of Perciliidae.
Thus, under the new interpretation presented here, the family Percichthyidae includes three genera (Percichthys, Macquaria, and Maccullochella) and 12 species; five species are restricted to southern South America and seven to Australia. The family Perciliidae contains five genera (Percilia, Bostockia, Gadopsis, Nannatherina, and Nannoperca) and 12 species; two are restricted to central-southern Chile and 10 to Australia.   The results of the mitogenomic analysis (Fig. 22) reveal that under the names Percilia gillissi and P. irwini are other cryptic species. Similar results have been published concerning Macquaria (Musyl, Keenan, 1992;Beheregaray et al., 2017), Gadopsis (Hammer et al., 2014;Unmack et al., 2017), and Nannoperca (Unmack et al., 2011;Buckley et al., 2018). Consequently, there is much more work needed to clarify the taxonomy of the South American and Australian percichthyids and perciliids.
Geographical distribution and endemism. Percichthyids and perciliids are distributed in freshwaters of southern South America and Australia (7)(8)(9). Three freshwater fish families (Aplochitonidae, Percichthyidae, and Perciliidae), the diadromus fish family Galaxiidae, and the cyclostomata Geotria inhabit these two regions. Despite the fact that percichthyids and perciliids are distributed in freshwaters of the two continents, we are unable to establish sister group regions between South America and Australia with the available information.
The current distribution of percichthyids and perciliids is interesting, because it suggests an Antarctic connection to explain it. However, fossil and extant percichthyids and perciliids have not been reported from the Antarctica yet. The extant South American and Australian percichthyids and perciliids are not diadromus, and it is difficult to explain their present distribution as result of transoceanic dispersion as has been proposed for galaxioids (McDowall, 1970). In contrast, Campos (1984) suggested that the present distribution of galaxioids is merely the result of the disruption of the supercontinent Gondwana. The same suggestion is put forward here to explain the actual distribution of percichthyids and perciliids. We can hypothesize that the ancestral stock of percichthyids and perciliids inhabited southern South America, Antarctica, and Australia before the break-up of Gondwana. During the Late Cretaceous the three continents were continuous, although there were breaks between all crustal blocks; final separation took place in the Eocene-Oligocene (e.g., Craddock, 1982;Woodburne, Zinsmeister, 1984). According to the morphological analysis, the phylogenetic positions of the South American genus Percichthys as the sister group of Macquaria australasica and Macquaria (=Plectroplites) and of the South American genus Percilia as the sister group of all Australian perciliids (Fig. 21) suggest that the separation between the South American and Australian stocks occurred not so long ago, because of the few divergent morphological changes. However, the group seems to have a longer evolutionary history in Australia than in South America as shown by a major generic differentiation, a condition also observed in other fish families-Aplochitonidae and Galaxiidae-that occur in both continents. Regarding the molecular and mitogenomic hypotheses, Macquaria (including M. ambigua and M. australasica) were recovered as the sister group of Percichthys and Percilia (Chen et al., 2014;Betancur-R et al., 2017; mitogenomic analysis herein; Fig.  22 and suplementary S1a-b). Thus, considering the time--calibrated phylogeny of Betancur-R et al. (2017), the split between Percichthys chilensis and Percilia occurred about 20 Ma, whereas the split of these genera and Macquaria (= Plectroplites) ambigua + Nannoperca australis occurred ca. 45 Ma. In this view, Percilia would have appeared within the South American continent during the Cenozoic. A Paleoce-ne-Oligocene origin for crown Percichthyidae sensu Johnson (1984) was suggested previously by Near et al. (2013) and Chen et al. (2014), which agrees with the age of the fossil percichthyids recovered in Chile (further research on this subject is in progress by the authors).