Astroniumxylon, Schinopsixylon, and Parametopioxylon n. gen. fossil woods from upper Cenozoic of Argentina: taxonomic revision, new taxon and new records

Abstract. This paper presents new descriptions of Anacardiaceae fossil woods from the Ituzaingó Formation (late Cenozoic) at the Toma Vieja, Curtiembre, and Arroyo El Espinillo localities, Argentina. We describe eight silicified woods assigned to four different species in three genera, one of which, Parametopioxylon crystalliferum n. gen. n. sp., is new. Similarities between these three genera and the six Anacardiaceae species previously recorded from the late Cenozoic in northeastern Argentina are investigated using multivariate analysis techniques (correspondence and cluster analysis). Our study is based on 33 characters scored for 17 fossil specimens (10 Astroniumxylon Brea, Aceñolaza, and Zucol 2001; five Schinopsixylon Lutz, 1979; and two Parametopioxylon n. gen.) and four extant species (Astronium balansae Engl., Astronium urundeuva Engl., Schinopsis balansae Engl., and Metopium sp.). Our main goal is to determine the wood anatomical features useful for distinguishing among these species. Results of the multivariate analyses support the previous classification where Schinopsixylon is distinguished from Astroniumxylon by having exclusively paratracheal axial parenchyma, ≥30% multiseriate rays, and multiseriate rays that are ≥5 cells wide and commonly 301–400 µm in height. Additionally, we propose that Schinopsixylon heckii Lutz, 1979 is synonymous with S. herbstii Lutz, 1979. A diagnostic key for the fossil species studied is given. Wood anatomy of Anacardiaceae fossil woods from Argentina (late Cenozoic) suggests a warm, dry to semi-humid climate for this region, supporting previous studies.

Some salient secondary xylem features of the Anacardiaceae are: simple perforation plates (except in some species with both simple and scalariform plates); alternate intervessel pits; parenchyma paratracheal, scanty, vasicentric, or aliform, *Corresponding author with marginal bands in some genera; rays mostly 2-3 cells wide, but larger in some species, usually heterocellular; fibers with simple pits, sometimes septate; and intercellular canals in lots of genera (Metcalfe and Chalk, 1950;Terrazas, 1999;Giménez et al., 2015). The anatomical structure of the Anacardiaceae xylem is fairly homogeneous (Heimsch, 1942;Agarwal and Gupta, 2008;Gupta and Agarwal, 2008;Schweingruber et al., 2011), but some features have been proposed to differentiate genera from each other (Heimsch, 1942;Terrazas, 1994). Because multivariate analysis of comparative wood anatomy has been shown to be helpful in distinguishing species or groups of species in extant and fossil genera, such as in the Anacardiaceae, Nothofagaceae, Leguminosae, and Podocarpaceae (Terrazas and Wendt, 1995;Poole, 2002;Pujana et al., 2014a, b), we conducted a similar analysis to clarify the systematics of Anacardiaceae fossil woods recovered in southern South America.
The aims of this study are: (1) to make a taxonomic revision of the late Cenozoic Anacardiaceae fossil woods from northeastern Argentina; (2) to describe eight new silicified woods assigned to four different species of Anacardiaceae; (3) to determine anatomical variability and similarities among fossil Anacardiaceae woods found in the Ituzaingó, Paraná, El Palmar, and "Salto Chico" formations, and related extant genera using correspondence and cluster analyses to establish the taxonomic limits and the diagnostic characters in fossil taxa; and (4) to propose a key to distinguish species of Parametopioxylon new genus, Schinopsixylon, and Astroniumxylon from each other.

Materials and methods
Eight new wood specimens are described, all of which were collected from the Ituzaingó Formation at the Toma Vieja, Curtiembre and Arroyo El Espinillo localities (Fig. 1). These fossil woods are silicified and were thin-sectioned in transverse, tangential, and radial planes following standard techniques for petrified woods. The specimens are registered in the "Colección Paleobotánica del Laboratorio de Paleobotánica, CICYTTP (CONICET-Prov. ER-UADER)" at Diamante city under the acronyms CIDPALBO-MEG for the fossil woods and CIDPALBO-MIC for the slides, and in the "Museo de Ciencias Naturales y Antropológicas Profesor Antonio Serrano" at Paraná city under the acronym MAS-PALEOBOT, Entre Ríos Province, Argentina. These sections were studied with a Nikon Eclipse E200 light microscope and photomicrographs were taken with a Nikon Coolpix S4 digital camera. For quantitative data, at least 25 measurements were taken. The mean values are shown, followed by minimum and maximum, in the description of each specimen. In addition, small fragments of wood were studied with a scanning electron microscope (SEM). The material was prepared for SEM by cutting a 1 cm 3 block of wood, mounting it on SEM stubs without coating, and then observing under low vacuum using a Phenom Pro Desktop SEM scanning electron microscope at the Electron Microscope Laboratory (Laboratorio de Microscopía Electrónica-EMLAB "Dr. Domingo Liotta"), CICYTTP (CONICET-Prov. ER-UADER), Diamante, Entre Ríos, Argentina.
The systematic assignment follows the APG IV (2016). Relationships to extant and fossil groups were investigated by using the references found in bibliographic lists made by Gregory (1994), Gregory et al. (2009), and Kew Micromorphology Bibliographic web site (https://www.kew.org/epic/datasources.  htm). With respect to the nomenclature used in the new fossil woods, the specimens were assigned to fossil rather than modern genera because: (1) the fossils are represented only by an isolated organ for which the remainder of the plant is unknown, and (2) there is a convention to match taxonomic rank with geological time. In this case, the fossils were found in the Ituzaingó Formation (late Miocene?) and it has been proposed to use modern generic names from the Pliocene (Collinson, 1986;Avise and Johns, 1999). This protocol allowed us to unify criteria with previous work and facilitated the comparison of the material with other fossils. In addition, a taxonomic revision based on 17 fossil specimens (10 Astroniumxylon, five Schinopsixylon, and two Parametopioxylon n. gen., including type species of each) and four extant species (Astronium balansae Engler, 1881; Astronium urundeuva (Allemão) Engler, 1881; Schinopsis balansae Engler, 1885; and Metopium sp.) was performed. The specimens examined are shown in Table 1.
The anatomical characters used in this study were considered in previous taxonomic papers to analyze phylogenetic and ecological aspects of the Anacardiaceae (Terrazas, 1994) and historical biogeography of Anacardiaceae based on fossil wood anatomy (Martínez-Millán, 2000). We used the characters that Terrazas (1994) and Martínez-Millán (2000) found useful for distinguishing Anacardiaceae species (Supplementary Material 1), to which we added the percentage of uniseriate, biseriate, and multiseriate rays.
Our complete Data Matrix (DM) consist of 21 binary and 12 multi-state characters, using the criteria of Crisci and López Armengol (1983). The characters used in the multivariate analysis with their corresponding states and codes and the DM are shown in Supplementary Material 1 and 2. If the fossil material was too badly preserved to identify the value of a character, the character was coded as "not comparable," and a question mark was used in the DM (Supplementary Material 2). Therefore, these character states were not included in the calculation process. The multivariate analyses were performed on a final DM that includes 22 wood anatomical characters because 11 were excluded either because they do not show variability, they are present in only one OTU, or the character can be highly influenced by environmental conditions (Baas and Wheeler, 2011, and references herein).
A Correspondence Analysis (CA) was performed to summarize similarities of anatomical variability among fossil specimens, and corroborate the accuracy of the systematic placement. Also, groups were classified by cluster analysis constructed using the UPGMA algorithm and based on Gower's distance matrix among OTUs. This coefficient can be used with a mixture of character types (binary, qualitative, and quantitative) as well as allowing missing values. All multivariate analyses were performed with the PAST software package version 2.17c (Hammer et al., 2001).

Systematic paleobotany
All the specimens studied in this paper show diagnostic features of the Anacardiaceae: diffuse to semi-ring-porous, tyloses present, solitary, radial multiples, and in cluster vessels, simple perforation plates, alternate to opposite intervessel pits, parenchyma and ray-vessel pits with much reduced borders to apparently simple; pits horizontal or similar to intervessel pits in size and shape throughout the ray cells, predominantly paratracheal vasicentric parenchyma, heterocellular rays, presence of radial canals in multiseriate rays and prismatic crystals common (Metcalfe and Chalk, 1950;Terrazas, 1994Terrazas, , 1999Gupta and Agarwal, 2008;Giménez et al., 2015).
Diagnosis.-As for type species.
Etymology.-The generic name denotes the wood's resemblance to extant Metopium Browne, 1756.
Description.-The fossil specimen is in a good state of preservation and is based on two fragments: the first one has a diameter of 69 cm and is 141 cm long, and the second one has a diameter of 76 cm and is 222 cm long. Growth rings are distinct to indistinct. Vessels are diffuse porous. Vessels are circular in cross-section with tyloses and other deposits. Vessels are mainly solitary (67%), frequently in radial multiples of two (17%), sporadically in radial multiples of three and more elements (10% and 4%, respectively), and in clusters (2%). The mean vessel tangential diameter is 80 (46-111) μm and the mean vessel radial diameter is 96 (26-171) μm. The mean vessel element length is 176 (85-275) μm. There are, on average, 12 (8-15) pores per mm 2 . The mean vessel wall thickness is 11 (7-20) μm. Vessel elements have oblique end walls. Perforation plates are simple. Intervessel pits are circular to polygonal in shape, bordered, and alternately arranged, with an average diameter of 7 (5-7) μm. Vessel-ray pits have much-reduced borders to apparently simple; pits horizontal. Fibers are non-septate, polygonal in outline, and form the ground in radially oriented files. The mean fiber diameter is 12 (8-18) μm and the wall thickness is 5 (3-7) μm. Axial parenchyma is paratracheal, scanty, and vasicentric.
Based on the presence of diffuse porous, vessels solitary and in radial multiples of 2-5, abundant tylosis, perforation plate exclusively simple, bordered and alternate intervessel pitting, paratracheal axial parenchyma vasicentric and scanty, rays uniseriate to multiseriate, heterocellular, and rays multiseriate with canals, our material most closely resembles the extant genus Schinopsis and conforms to the circumscription of the fossil species Schinopsixylon herbstii . We therefore emended the specific diagnosis to include the presence of solitary vessels and the phrase "rays heterocellular composed of procumbent and upright cells."

Multivariate analyses
The cluster resulting from the analysis of 17 fossil specimens and four extant species shows three well-delimited groups: the first one is comprised by Metopium sp., PMET1 and PMET2; the second, by SCHK1, SCHT1, SCHK2, SCHK4 and Schinopsis balansae; and the third, Astronium balansae, Astronium urundeuva, all the specimens of Astroniumxylon and SCHK3 (Fig. 6.2). In this third group, it is possible to further distinguish three subgroups: the first one includes all the specimens of Astroniumxylon bonplandianum, the second has all the specimens of Astroniumxylon parabalansae and SCHK3, and the third includes Astronium balansae, Astronium urundeuva and Astroniumxylon portmannii.
The CA performed for the specimens indicates that the first four dimensions accounted for 63.781% of the total variation ( Table 2). The first dimension accounts for 21.708% of the total variance (Table 2, Fig. 6). The characters that contributed most to the first dimension are: vessels in diagonal or dendritic pattern (3), porosity (2), shape (12) and arrangement (13) of intervessel pits, mean vessel element length (10), height of  multiseriate rays and rays mostly 4-10 seriate (32) (Supplementary Material 3). Four characters-height of multiseriate rays (28), and rays mostly 4-10 seriate (32), shape of intervessel pits (12), and vessels in diagonal or dendritic pattern (3)-contributed most to the second dimension, which accounts for 17.672% of the total variance (Table 2 and Supplementary Material 3). Dimension three accounts for 12.801% of the total variance (Table 2); characters 13 (arrangement of intervessel pits), 10 (mean vessel element length), 29 (prismatic crystals in rays), and 30 (percentage of uniseriate rays) have the highest coefficients on this dimension (Supplementary Material 3). Dimension four accounted for 11.527% of the total variance; Figure 6. (1) Scatterplot of scores on Axis I versus Axis II from Correspondence Analysis (CA) of specimens analyzed (for OTU numbers see Table 1). (2) Cluster analysis constructed by means of UPGMA algorithm and Gower's Similarity Distance Coefficient (for OTU numbers see Table 1).

Discussion
Multivariate analyses.-According to the cluster analysis, the group formed by PMET1, PMET2 (both specimens of Parametopioxylon crystalliferum n. gen. n. sp.), and Metopium sp. shows the lowest dissimilarity with values lower than 0.30. This is consistent with the fact that Parametopioxylon crystalliferum n. gen. n. sp. has more affinity with the descriptions made for Metopium sp. (see Terrazas, 1994;InsideWood, 2004-onwards).
Another group is made up of different specimens of Schinopsixylon and Schinopsis balansae. In 1979, Lutz created the genus Schinopsixylon and erected two species, S. herbstii and S. heckii. According to the diagnosis, S. heckii differs from S. herbstii by the presence of mostly solitary pores (80% solitary pores and 20% in radial multiples) and only one radial canal per ray . Based on our multivariate analysis, we determined that the characters 5 and 33 (vessels mostly solitary and number of radial canals per ray, respectively; Supplementary Material 1 and 3) lack enough taxonomic value to justify the separation of the two species in Schinopsixylon (Fig. 6) and correspond to intraspecific variations. Thus we proposed that S. heckii is synonymous with S. herbstii (Table 3; Fig. 6). These results are consistent with wood anatomy and multivariate analyses carried out by Gimenez et al. (2015) in three species of extant Schinopsis (S. lorentzii, S. balansae, and S. marginata Engler, 1883). These authors did not observe differences in the 29 anatomical characters studied and proposed that Schinopsis wood anatomy is substantially homogeneous with only some differences found in quantitative variables as well as density, diameter, and length of the vessels, all resulting from ecological adaptations caused by altitudinal gradient (Gimenez et al., 2015).
The characters enabling the distinction between Schinopsixylon and Astroniumxylon are the presence of rays mostly 4-10 seriate, exclusively paratracheal axial parenchyma, rays ≥6 cells wide common, and multiseriate rays commonly 301-400 μm in height (Supplementary Material 3). Therefore, as a result of our multivariate analysis, we transfer CIDPALBO-MEG 12 specimen (under the acronym SCHK3), previously assigned to Schinopsixylon heckii by Brea et al. (2010, see Table 1) to Astroniumxylon parabalansae on the basis of possessing apotracheal axial parenchyma and multiseriate rays ≤5 cells wide, and of not possessing prismatic crystals and mostly 4-10 seriate rays (Table 3, Fig. 6, Supplementary Material 3).
The CA delimits a group composed of Astronium balansae, A. urundeuva, and all the fossil woods with affinity to these. Two subgroups can be differentiated inside of that one. One subgroup consists of A. urundeuva and Astroniumxylon bonplandianum fossils; the other subgroup consist of A. balansae, Astroniumxylon parabalansae, and A. portmannii fossils ( Fig. 6.1). In the cluster analysis, the fossil groups mentioned above are delimited, the extant species and A. portmannii are not related to the subgroups as in the CA (Fig. 6.2). Astroniumxylon parabalansae and Astroniumxylon portmannii can be differentiate because the first one has vessels in radial multiples of 4 or more elements and height of multiseriate rays are <300 μm.
Based on the multivariate analysis, ten variables are the most efficient in discriminating among species: porosity type (2), vessels arrangement (3), mean of vessel element length (10), shape (12) and arrangement (13) of intervessel pits, apotracheal axial parenchyma present (19), height of multiseriate rays (28), prismatic crystals in rays (29), percentage of uniseriate rays (30), and rays mostly 4-10 seriate (32). In the present paper, these characters are considered the most useful for delineating Some peculiar features of the Anacardiaceae fossil woods, such as distinct growth rings, semi-ring-porous, radial canals, and scanty and paratracheal axial parenchyma, suggest a dry or seasonally dry warm climate. Others features present in some fossil woods from the Paraná Formation, such as wider vessels and low vessel frequency, suggest that some taxa were exceptionally adapted to humid conditions associated with riparian forests (see Brea and Franco, 2013 and references herein).
Comparisons with the nearest living relatives of the Paraná fossil woods can also provide insight into paleoclimate and paleoecology. The genus Metopium includes evergreen to semideciduous, medium-size trees, ranging from West Indies to Florida, Mexico, and Central America (Terrazas, 1994). This area corresponds to the Caribbean subregion sensu Morrone (2001) and is known as Halffter's Mexican Transition Zone (MTZ). The MTZ is a complex area where the Neotropical and Nearctic biotas overlap (Halffter and Morrone, 2017). The genus Astronium has endemic species with disjunct distribution throughout different areas in the Neotropical Dry Forests (Pennington et al., 2000), many of which are characterized by a strongly seasonal climate. Astronium urundeuva is an important member of the Caatingaas, a seasonal dry forest in north-eastern Brazil (Prado and Gibbs, 1993). Today, these genera do not occur where the fossils described here were collected. The finding of fossils with affinities to Metopium and Astronium affinity could be explained by a greater southward and eastward extent of Neotropical Dry Forests during the late Miocene than occurs today (Prado and Gibbs, 1993;Pennington et al., 2000;2004;Prado, 2000, DRYFLOR, 2016Flanklin et al., 2018). These results are consistent with the fact that during the late Miocene, large areas of the continents experienced drying, enhanced seasonality, and a restructuring of terrestrial plant and animal communities. These global climate changes are seen throughout the subtropics, but have been attributed to regional tectonic forcing (Herbert et al., 2016;Stevens Goddard and Carrapa, 2018).

Conclusions
A taxonomic revision and multivariate analyses were conducted on seventeen specimens of late Cenozoic Anacardiaceae fossil woods from northeastern Argentina and four extant species. Eight new silicified wood specimens assigned to four different species of the Anacardiaceae family were described: two specimens of Parametopioxylon crystalliferum n. gen. n. sp, three of Astroniumxylon bonplandianum, two of Astroniumxylon parabalansae, and one of Schinopsixylon heckii. Parametopioxylon crystalliferum n. gen. n. sp. has been erected based on exceptionally wellpreserved fossil woods recovered from the Ituzaingó Formation.
A correspondence analysis of 22 anatomical characters of anacardiaceous woods was conducted. Taxonomically relevant characters, anatomical variability, and relationships among Anacardiaceae species were identified using correspondence and cluster analyses and allowed us to differentiate and revise the Anacardiaceae genera and species included in this paper. Schinopsixylon heckii  and Schinopsixylon herbstii  are not clearly segregated and can be explained by intraspecific or individual variability; we propose that S. heckii is synonymous of S. herbstii. Schinopsixylon could be distinguished from Astroniumxylon by the presence of mostly 4-10 seriate rays, commonly ≥6 cells wide, 301-400 μm in height, and axial parenchyma exclusively paratracheal. A taxonomic key is proposed in this paper to differentiate anacardiaceous fossil woods found in the late Cenozoic of northeastern Argentina.
Lastly, in accordance to previous studies, wood anatomy suggests that in the late Cenozoic, Anacardiaceae were linked to a dry or seasonally dry warm climate, with some trees and shrubs adapted to xeric conditions and seasonality and others exceptionally adapted to humid conditions associated with riparian forests.