Paradiospyroxylon kvacekii gen. et sp. nov. from the Paleogene of the Czech Republic: a case study of individual variability and its significance for fossil wood systematics

ABSTRACT The wood anatomy of Paradiospyroxylon kvacekii gen. et sp. nov. is described and illustrated based on material originating from the Ústí Formation’s volcanic deposits of České středohoří Mts. (Paleogene, Czech Republic). The sample, identified earlier as Manilkaroxylon sp., was critically examined and is interpreted as root wood and proposed as the paratype of Paradiospyroxylon kvacekii. This paper discusses how wood anatomical variation needs to be considered when making systematic and palaeoecological interpretations.


Introduction
České středohoří volcanic complex in north-western Bohemia is rich in fossil plant remains and a valuable source of information for the flora of the European Paleogene. There are fossil leaves and reproductive structures, which have been described from numerous localities (e.g., Kvaček and Walther 1998, 2001, 2004. Fossil woods also occur there and have been studied since the nineteenth century, for a summary see Koutecký et al. (2019). A revision of some published samples done by Sakala et al. (2010) indicates that five fossil-wood species of four fossil-wood genera, defined there by Prakash et al. (1971), correspond probably only to two botanical species. This underlines the importance of the intraspecific and individual variability of non-ecological significance (e.g., Sakala 2000b), which will also be discussed hereafter.
The present paper deals with a new find of an angiosperm wood Paradiospyroxylon kvacekii from the locality Divoká rokle near the regional capital of Ústí nad Labem. Manilkaroxylon sp. was described from the same locality by Koutecký and Sakala (2015). We demonstrate here how these two wood types are related and discuss how individual variability in wood anatomy needs to be considered delimiting species of fossil wood. The aim of our research is therefore a presentation of the new fossil taxon with specific vessel arrangement to scientific community and especially a general discussion about the importance of wood anatomical variability in palaeobotany.

Geological settings
The Paleogene and Neogene of north-western Bohemia (Czech Republic) are represented by a volcanosedimentary area linked to the so-called Ohře Rift, which is parallel to the Czech/German boundary. It is formed of two volcanic complexes (České středohoří and Doupovské hory Mts.) developed in late Eocene-Miocene by an alternating deposition of lavas, tuffs, debris flows and other volcanic rocks (e.g, Cajz et al. 2009) and predominately lower Miocene sedimentary basins (Cheb, Sokolov, Most and Žitava Basins). There is a sparse record of fossil angiosperm woods in the sedimentary basins (e.g, Sakala 2002), but a rich record in the Doupovské hory Mts. (e.g, Sakala et al. 2010;Koutecký et al. 2019). The locality Divoká Rokle belongs to the České středohoří volcanic complex (Figure 1), which was developed in four main phases with volcanic products assigned to four respective formations of distinct geochemical composition. There are only a few published descriptions of fossil angiosperm woods from this area and so any new contribution has importance.
The sedimentary area of Divoká Rokle locality represents the parastratotype of the lower Oligocene Ústí Formation (Cajz 2000). There were two main sedimentary parts: the older, which was deposited in subaqueous area after an outburst of magma, and the younger deposited in subaerial environment. Within these parts, three lithofacies associations are described: breccias and conglomerates, lacustrine sandstones and alluvial fine-grained conglomerates (Kratochvíl 2007). The presence of fossil wood is confined to the subaerial breccias and conglomerates.

Material and methods
The two studied samples were collected by the first author in the Divoká rokle locality in 2013 (DR 02; thin slides: XYL12 a-c) and 2016 (DR 23; thin slides: XYL13 a-i). The samples were cleaned and after preliminary examination with a Leica EZ5 Stereo Microscope, thin sections were prepared using standard techniques. Thin sections were studied with Olympus BX-51 optical microscope in normal transmitted light, paired with an Olympus DP 73 camera. The photos and pen drawings were adjusted in GIMP 2.10. The scales were added in Quick PHOTO MICRO 3.0. The anatomical descriptions are in accordance with IAWA Hardwood list (IAWA Committee 1989;InsideWood 2004-onwards;Wheeler 2011;Wheeler et al. 2020) and the range of features is presented as: minimum-mean-maximum or minimum-maximum. A value in brackets is a rarely occurring one. Data analysis was performed in accordance with Moya et al. (2018) using the Principal component analysis. In order to visualise the data distribution, a dendrogram supplemented by a cophenetic correlation was used. This correlation is a measure of how faithfully the dendrogram preserves pair distances between the original unmodeled data points. The thin sections and the samples of holotype and paratype are housed in the Czech Geological Survey, Prague.

Generic diagnosis
Wood semi-ring-porous to diffuse porous; vessels narrow, predominantly solitary, oval to angular in outline and arranged in radial pattern, often closely spaced with single pores separated by axial parenchyma cells. Vessel elements short, perforation plates simple, pits alternate. Rays typically narrow, distinctly heterocellular with procumbent body ray cells and more than four rows of upright/ square marginal cells, multiseriate portions often as wide as the uniseriate portions. Axial parenchyma scanty paratracheal and apotracheal diffuse-in-aggregates with a slight tendency to interrupted, one cell wide, bands. Thin-to thick-walled non-septate fibres.

Etymology
Paradiospyroxylon is derived from the extant genus Diospyros.

Specific diagnosis
Wood semi-ring to diffuse-porous; earlywood vessel tangential diameters small to medium (50-140 μm), predominantly solitary, oval to angular outlines and arranged in radial pattern with single pores often separated by axial parenchyma cell or fibres. Vessel element lengths up to 550 μm, perforation plates simple, intervessel pits alternate, minute, vessel-ray parenchyma pits similar to intervessel pits. Rays up to four cells wide and up to 1 mm high, distinctly heterocellular with procumbent body ray cells and up to 11 rows of upright/square cells and often with multiseriate portions as wide as the uniseriate portions; non-storied. Axial parenchyma scanty paratracheal and apotracheal diffuse-in-aggregates with a slight tendency to make interrupted, one cell wide, wavy bands. Thin-to thick-walled nonseptate fibres. Synonymy 2015 Manilkaroxylon sp.; pl. 5,15.

Etymology
The species is named after Zlatko Kvaček as an honour to a great man and palaeobotanist who dedicated his life to the study of Tertiary floras and contributed greatly to our current knowledge.

Holotype
Thin slides XYL13 a-i from the sample DR 23 (Czech Geological Survey, Prague)

Paratype
Thin slides XYL12 a-c from the sample DR 02 (Czech Geological Survey, Prague)

Root form (paratype DR 2).
A detailed description was given by Koutecký and Sakala (2015). Revised anatomical features are presented in Table 2.

Discussion
The combination of these features: semi-ring porosity, narrow vessels arranged in radial rows, simple perforation plates, scanty paratracheal and diffuse apotracheal axial parenchyma with 4-11 cells per strand, narrow rays with long uniseriate extremities and some vertically fused rays occurs in the families Apocynaceae, Rubiaceae and Ebenaceae (Metcalfe and Chalk 1950). While the family Ebenaceae consists only of four genera, i.e., Diospyros L., Royenna L., Euclea L. and Lissocarpa Benth., the Apocynaceae and Rubiaceae, both in the order Gentiales, have hundreds of genera, many with similar wood anatomy, making it difficult to identify an isolated wood from these two families to genus.

Comparison with recent wood
The literature contains only a few descriptions of woods with the combination of features observed in these Oligocene woods, namely: predominantly solitary, closely spaced, narrow vessels arranged in radial rows. Metcalfe and Chalk (1950, p. 768) mention in their summary of Rubiaceae secondary xylem anatomy: '. . . commonly with a tendency for single vessels to be grouped together radially but separated by a single row of fibres . . . ' showing as an example Mitragyna stipulosa (DC.) Kuntze, which belongs to the 'wood type I' sensu Jansen et al. (2002). Other descriptions of narrow vessels in radial arrangement separated by fibres or axial parenchyma cells were given by Fahn et al. (1986, p. 71, Nerium oleander L., Apocynaceae) and in Kräusel's (1939, p. 104) description of Ebenoxylon aegyptiacum Kräusel: '. . . einzeln oder in radialen Reihen zu 2-8, mitunter 2 solcher hintereinander liegend, durch kleinere oder Gefäßtracheiden verbunden . . .'. Vessel elements in this wood are very short. Short vessel elements occur in the families Ebenaceae and Rubiaceae (Jansen et al. 2002;Jahanbanifard et al. 2020) and in several clades of Rauvolfideae subfamily of Apocynaceae, e.g., Carisseae, Hunterieae, Melodineae, Plumerieae and Willughbeieae and also the whole APSA clade (Lens et al. 2008(Lens et al. , 2009).
In the Rubiaceae, there are two main wood types (Jansen et al. 2002), based on fibre type, vessel groupings, ray width and structure, and axial parenchyma distribution. However, any of the combinations of the characteristics for 'wood type I' or 'wood type II' sensu Jansen et al. (2002) do not fully correspond to our wood. Although the predominantly solitary vessels and narrow rays with long uniseriate margins in our wood agree with the 'wood type I', this group is characterised also by fibre-tracheids, a feature not observed in our fossil. Another distinctive feature of this fossil wood is semi-ring-porosity. The semi-ring-porous Rubiaceae woods differ from our wood: Cephalanthus L. has only uniseriate rays; Chiococca P. Browne, Pausinystalia Pierre ex Beille, Sinoadina Ridsdale and Vangueria Comm. ex Juss have distinctly bordered pits on fibre walls; Coprosma J. R. Forster and G. Forster has distinctly bordered pits on fibre walls and frequent scalariform perforation plates and/or helical thickenings in vessel elements; Posoqueria Aubl. has vessels typically in radial multiples; Emmenopterys Oliv. has helical thickenings. Overall none of the types mentioned above have as pronounced a radial vessel arrangement as seen in our fossil wood.
In accordance with results by Lens et al. (2008Lens et al. ( , 2009), short vessel elements are present in the later derived members of Apocynaceae as mentioned above. However, another trend they observed in vessel grouping, with solitary vessels in the earlydiverging clades and long radial multiples and vessel clusters in the more-derived Apocynaceae. This is unlike our wood which has both short vessel elements and mostly solitary vessels. Also, there are trends (Lens et al. 2008(Lens et al. , 2009) that do not correlate with our wood: paratracheal parenchyma occurs in the more-derived members and axial parenchyma strands up to 11 cells occur in the socalled 'primitive' tribes of Aspidospermeae and Alstonieae. With regard to the porosity, it is overall diffuse-porous except for some species of Allamanda L., Alyxia Banks ex R. Br., Ichnocarpus R. Br., Malouetia A. DC. and Tabernaemontana L. which are semi-ring to ring-porous (Metcalfe and Chalk 1950).
Most Ebenaceae have diffuse-porous wood, except for Diospyros virginiana L., D. kaki L. and D. lotus L., which are (or can partly be) semi-ring porous (Jahanbanifard et al. 2020). All the aforementioned semi-ring porous species also have storied structure (Metcalfe and Chalk 1950;InsideWood 2004onwards). Diospyros species typically have apotracheal axial parenchyma in narrow, usually uniseriate, tangential bands varying from irregular and short to longer continuous and regularly spaced ones; paratracheal axial parenchyma occurs always as combination of scanty paratracheal or scanty paratracheal to vasicentric (see Table A1 in Jahanbanifard et al. 2020, pp. 21-37). However, some species have also a combination of diffuse-in-aggregates axial parenchyma and scanty paratracheal or vasicentric (e.g.,  (Metcalfe and Chalk 1950;Jahanbanifard et al. 2020). According to Jahanbanifard et al. (2020), there are only 1-4 marginal rows of square/upright cells, but according to Metcalfe and Chalk (1950) and InsideWood (2004-onwards) there can be over 4 rows of square/upright marginal cells (e.g., D. borbonica, Détienne and Jacquet 1993).
Overall the anatomy of our fossil wood is equally similar to/ different from Apocynaceae, Rubiaceae and Ebenaceae. Therefore, we performed a PCA analysis to assess the similarity of the fossil to these three families, following the examples of Oakley and Falcon-Lang (2009) for Upper Cretaceous woods of the Czech Republic or Moya et al. (2018) for Pliocene woods of Argentina.
Principal component analysis. As a support for our conclusions from comparison with recent woods, principal component analysis was done. It is one of two ways to measure the similarity between entities (Lessig 1972;Sneath and Sokal 1973). Cophenetic correlation was found to be R = 0.890. According to Clarke et al. (2016), this value is sufficient. The result of the clustering is, therefore, a suitable representation of the original distances. Also, cluster analysis was used to make a classification of the groups. The analysis was performed using InfoStat version 2015 (Di Rienzo  Due to the large amount of data, it is not possible to include the whole dendrogram for all 124 monitored items (genera or tribes). To simplify it, first we only worked those elements that the variable (sample DR 23) most closely resembles or where the smallest distance to Diospyros-Royena, Diospyros-Royena-Euclea, Diospyros-Royena-Euclea-Lissocarpa (i.e., the whole Ebenaceae) occurs. Subsequently, elements from the Apocynaceae (first the subfamily Rauvolfioideae, then Apocynoideae) were added to this cluster, and elements of the Rubiaceae. The individual distances corresponding with the first connection between DR 23 and an element of a given family or families are in more detail within the dendrogram in Table 1.
The following dendrogram ( Figure 5) shows how the interaction between DR 23 -Ebenaceae and DR 23 -Apocynaceae works. For clarity, elements from the Rubiaceae family are not included in the dendrogram as the given variables are associated with them last. The whole dendrogram is attached as Appendix 1 (see Supplemental Data with this article).
It is obvious from the graph that the variables form three clusters. The first is saturated with elements 1-9, which correspond to the subfamily Apocynoideae, the second with elements 121-124, which correspond with family Ebenaceae. This second cluster is then gradually connected with elements 10-19 corresponding to the Rauvolfiodeae subfamily. The layout of all the elements is shown in the following biplot ( Figure 6). The analysis of the main components was performed with the assignment of all analysed groups of Ebenaceae, Apocynaceae, and Rubiaceae families. Principal components 1, 2 and 3 explained 29.2% (see Figure 6), 15.7% and 8.8%. That is in total 53.7%. Due to capacity reasons, it was not possible to insert all biplots and, thus, the option PC1 -PC2 was used primarily due to the highest saturation.
It is clear from the graph that DR 23 is closer to Ebenaceae family elements and subsequently to those of Apocynaceae family. The length of the projection of individual original variables in the space of factor axes describes their contribution to the definition of a given factor space.
As part of the data analysis, a similar analysis (cluster analysis, PCA analysis) was performed, where the possibility of occasional occurrence of a given feature (with its own value) was included within the data matrix, and its result is only statistical evidence of the relationship between Rubiaceae and Apocynaceae or the distance of Ebenaceae. Our sample DR 23 is remote from Ebenaceae in this analysis because in this family there is a large representation of the value 3 (occasional occurrence) which does not prove the similarity of one particular sample to a larger group, but only of the groups among themselves. This is mainly due to the large number of missing values in the data matrix which is caused by the acquisition of information from comparative studies where the authors (Jansen et al. 2001(Jansen et al. , 2002Lens et al. 2008Lens et al. , 2009 Jahanbanifard et al. 2020) observe especially typical features for a given group/family, not fully correlatable with each other, and the effort to intersect as much as possible across the three large units. In our case, the DR 23 sample was closer to Apocynaceae family. A dendrogram and PCA graph is attached in Appendix 2 (see Supplemental Data with this article). As a result of comparison with recent wood and our statistical analyses we concluded that our wood belongs to the Ebenaceae and is especially close to Diospyros.

Comparison with fossil wood
Although we attribute our wood to the Ebenaceae, we also compared it with fossil woods of Rubiaceae and Apocynaceae as listed in Gregory et al. (2009).
The oldest description affiliated of Rubiaceae was presented by Crié (1888). The wood was however later recognised as being dipterocarpaceous (Kräusel 1926;den Berger 1927). Later on, there are several wood descriptions related to this family, e.g.,
In summary, we did not find similarity between this Oligocene wood and any previously described fossil wood of the Apocynaceae, Ebenaceae, or Rubiaceae. We believe this wood is most similar to Diospyros and Ebenaceae, even if there is no a similar combination of features within a present-day species of Diospyros. The name Diospyroxylon was first used without a proper diagnosis: 1) in a summary (Selmeier 1976) and 2) for a fossil wood sample thought similar to extant Diospyros ebenaster Ret and called Diospyroxylon sp. (Greguss 1967). Although a generic diagnosis is missing, Petrescu (1978) published an 'emended diagnosis' of Diospyroxylon Greguss 1967 and made new combinations, e.g., Diospyroxylon knollii (Hofmann) Petrescu. As discussed above, our fossil wood does not correspond to any living species of Diospyros, so we propose a new fossil genus Paradiospyroxylon, for Diospyros-like woods that are diffuse-porous and do not have regular tangential bands of axial parenchyma.

Individual variability vs. palaeoecological interpretation
There are two samples in the Divoká rokle locality (DR 02 and DR 23) which agree in most anatomical features except for porosity. The more poorly preserved sample DR 02 was originally described by Koutecký and Sakala (2015) as Manilkaroxylon sp. (Sapotaceae). However, when we compared the distribution of vessels in radial rows of both samples, we found that each whole growth ring in DR 02 corresponds to the earlywood in DR 23(see Figure 7). Apparently, there is no latewood in DR 02, and we consider that both samples belong to the same species.
The virtual absence of latewood can be explained in two ways by: 1. ecological conditions, when there are many stress factors, which could prevent the formation of latewood, such as flooding, drought, volcanic activity or burial (e.g., Corcuera et al. 2004;Schweingruber 2007;Stoffel et al. 2010); and/or 2. individual variability, where the sample DR 02 can be interpreted as a root, given the roots of ring-porous woods can be diffuseporous (e.g., Patel 1965;Cutler et al. 1987).
All 16 growth rings of sample DR 02 are nearly uniform. Moreover, in its central part, it seems there is no pith preserved, indicating the sample is probably a root wood. Also, slight differences in some other features (see Table 2) might indicate DR 02 is a root wood: larger intervessel pits, larger tangential diameter of the earliest vessels and larger ray parenchyma cells (e.g., Patel 1965;Machado et al. 1997;Psaras and Sofroniou 2004). Consequently, we think that we are dealing here with individual variability, rather than with palaeoecologically based differences, and so are treating DR 23 (stem wood) and DR 02 (root wood) as the same species. We have previously used a similar approach, both for angiosperm (Sakala et al. 1999) and conifer (Koutecký and Sakala 2015) fossil wood.

Fossil record of Ebenaceae and its biogeographic implications
Recently, overviews of the fossil record of Ebenaceae were presented (  Kvaček et al. 2014). This fossil species is thought to be similar to modern D. virginiana, D. lotus or D. kaki (Kvaček et al., 2019), all of which have semi-ring-porous wood as does our new fossil wood. However, an attribution to any single living species remains problematic as well as is an unambiguous association of the wood of Paradiospyroxylon kvacekii with leaves and reproductive structures of Diospyros brachysepala.
Concerning eventual biogeographic scenario and mode of dispersal of Diospyros based on this newly defined early Oligocene wood, there is little to be added to Denk and Bouchal (2021), and it is generally difficult to reconcile it with an African origin of the genus, but we believe, similarly to Denk and Bouchal (2021) that this either 'results from a sampling bias with Europe being much better sampled than Africa', or Diospyros present in the African Paleogene 'could also be the result of immigration from Europe' (see in Denk and Bouchal 2021;compare with Grímsson et al. 2019).

Conclusions
Based on an extensive comparison of the literature and a statistical analysis, a new Palaeogene wood of the family Ebenaceae was recognised from volcanic rocks of the Czech Republic and named Paradiospyroxylon kvacekii gen. et sp. nov. The new fossil wood species is represented by two samples: the holotype DR 23, which represents a stem wood and the paratype DR 02, described originally by Koutecký and Sakala (2015) as Manilkaroxylon sp., which, we believe is root wood. Both wood pieces could come from the plant that produced the leaves and calyx of Diospyros brachysepala, which is the only known fossil member of Ebenaceae in the Paleogene-Neogene of northwestern Bohemia, however, is not yet known from the Divoká Rokle locality. Our study of the two samples, collected at different times, demonstrates some of the pitfalls of systematic palaeontology, where the authors often deal with isolated fossil wood samples. Intraspecific and individual variability and variability associated with ecological conditions have long been documented for modern wood (e.g., Bailey and Faull 1934;Patel 1965;Cutler et al. 1987;Corcuera et al. 2004;Falcon-Lang 2005;Schweingruber 2007;Stoffel et al. 2010), however there are fewer examples in the fossil record with many species based on but one samples. Samples DR 02 and DR 23 have a distinctive vessel arrangement; DR 02 has features consistent with root wood and sample DR 23 has features consistent with stem wood. We propose a new fossil wood taxon Paradiospyroxylon kvacekii gen. et sp. nov. with two informal forms (stem and root), which documents another record of the Ebenaceae in the Paleogene/Neogene of north-western Bohemia. Generally, the systematical interpretation is based here on the difference between stem and root wood and demonstrates how important is the individual variability for the definition of new fossil wood taxa.