Filling a Gap in Quercus Phylogeny: Molecular Phylogenetic Evidence, Morphometric and Biogeographic History of Quercus petraea subsp. pinnatiloba Matt. Liebl from Türkiye

Abstract

Quercus petraea subsp. pinnatiloba is a narrowly distributed oak taxon in southeastern Türkiye, and its taxonomic position has long remained uncertain. This study aims to clarify its distinctiveness by integrating morphological, molecular, and biogeographical evidence. Principal Component Analysis (PCA) and Stepwise Discriminant Analysis (SDA) of 14 leaf traits revealed that subsp. pinnatiloba constitutes a morphologically stable and distinctly differentiated group from other Q. petraea subspecies and closely related taxa, characterized by key diagnostic traits such as petiole length (PL), lamina length (LL), length of leaf blade at its broadest point (WP), and lobe width at the tip of the widest lobe (LW). Phylogenetic analyses based on nuclear ITS and plastid markers (rbcL, psbA-trnH) confirmed its placement within sect. Quercus, yet consistently distinguished it genetically from other subspecies for the first time. Molecular dating (BEAST) suggested divergence in the Miocene (11 Mya with 95% HPD 3.01, 20.95) while RASP biogeographical analysis indicated an origin in the Euro-Siberian region with later dispersal into the Mediterranean. These integrative results support its recognition at species rank as Quercus pinnatiloba, clarifying its phylogenetic placement and underscoring the conservation importance of this lineage.

1. Introduction

The genus Quercus L., encompassing 350–500 species, is widely distributed across the Northern Hemisphere [1]. These trees are prominent inhabitants of temperate deciduous forests in North America, Europe, and Asia. The genus is found across Türkiye from the Mediterranean maquis to the temperate rain forest in northern Anatolia, as well as dry forests in the steppes. Türkiye has a rich diversity of Quercus species, with 14 deciduous and 3 evergreen oak species documented [2]. Presently, based on morphological evidence, 23 taxa, including subspecies and variations, have been officially recognized in Türkiye [2,3,4,5]. Recent evolutionary analyses of nuclear DNA sequence data revealed two significant lineages within the genus Quercus: Subgenus Cerris Oerst. and Subgenus Quercus [6,7,8]. One important member of the Subgenus Quercus lineage is sessile oak (Quercus petraea Matt. Liebl.), a prevalent species of European white oak stretching from Great Britain to the Iberian Peninsula [9]. The Quercus petraea (Matt.) Liebl., one of the most economically important pure stands [3], is represented by three subspecies in Türkiye as Quercus petraea subsp. iberica (Steven ex M.Bieb.) D.I.Krassiln. (accepted name as Q. petraea subsp. polycarpa (Schur) Soó), Quercus petraea subsp. petraea, and Quercus petraea subsp. pinnatiloba (K.Koch) Menitsky [3]. The first, subsp. petraea, is characterized by cupule scales that are flat and adult leaves that are densely hairy on the underside. In contrast, subsp. iberica (accepted name as subsp. polycarpa) has tuberculate cupule scales, and its adult leaves are nearly glabrous underneath, shallowly lobed, typically with intercalary veins present. Lastly, subsp. pinnatiloba also has tuberculate cupule scales, but its leaves are either glabrous or pubescent beneath, often glaucous, deeply lobed, with intercalary veins absent [3]. Quercus petraea subsp. pinnatiloba have previously been considered endemic to Türkiye [10,11,12]. However, Mehrnia et al. (2013) reported that this subspecies occurs across a wide range in the Zagros forests of Iran, suggesting that its endemic status is no longer valid [13,14]. Common synonyms of Quercus petraea subsp. pinnatiloba are Q. pinnatiloba K.Koch and Q. cedrorum Kotschy, with common distributions in Lebanon, Syria, and Türkiye [13].

In Quercus classification, leaf characteristics are considered the most important traits for distinguishing species [15,16]. Leaf morphological assessment is a critical tool in taxonomic, infraspecific, and population-level studies of oaks, with standardized approaches involving a combination of direct measurements, observed traits, and synthetic variables, as demonstrated before [17,18,19]. Due to the blurred species boundaries, distinguishing between closely related oak species such as Quercus petraea and Quercus robur L. based on morphological traits alone is particularly challenging [18,19]. This difficulty becomes even more pronounced at the infraspecific level, where relying solely on morphological characters for taxonomic delimitation is often insufficient. For example, while Quercus petraea and Q. robur are typically distinguishable by certain morphological features, especially in the fruit, their identification can still be challenging in hybrid zones or in the absence of reproductive structures [18,20]. Moreover, Q. petraea shows closer morphological overlap with other white oaks such as Q. pubescens, particularly in sympatric regions [21,22]. This challenge extends within the Q. petraea collective group, which includes several morphologically similar taxa of uncertain placement, such as Q. polycarpa Schur, Q. banatus Kucera, and Q. dalechampii Ten., used in Eastern European and Balkan floras. While some studies suggest that Q. dalechampii sensu Auct. (Eastern Europe) aligns more closely to Q. petraea lineage [23,24,25], others have placed it within the Q. pubescens lineage [26,27,28]. Based on the lectotypification provided in Di Pietro et al. (2012), Q. dalechampii Ten. has been definitively assigned to the Q. pubescens complex [29]. Accordingly, all the previous records of Q. dalechampii (classified as belonging to the Q. petraea complex) from the Balkan Peninsula are to be assigned to a different taxon name [30,31]. Standardized approaches, typically involving a combination of direct measurements, observed traits, and synthetic variables, can overcome limitations [18].

Leaf morphometric investigations are increasingly integrated with molecular data in the Quercus genus to yield essential insights into species delimitation within taxonomically particularly intricate groupings, such as that of the pubescent white oaks [32,33,34]. Phylogenetic and evolutionary analyses based on nuclear and plastid DNA markers can help clarify relationships that are often obscured by introgression, hybridization, or phenotypic plasticity [8]. Thus, integrative taxonomy by combining morphological, molecular, and biogeographic evidence has emerged as a powerful and widely accepted approach for resolving species boundaries in oaks and other complex plant groups [35,36].

Phylogenetic relationships were established utilizing molecular markers, including allozymes [37], ribosomal gene restriction maps [38], cpDNA [2], and nrDNA polymorphism [39], which also facilitated haplotype analysis and spatial distribution delineation. Moreover, the rates of evolution and genetic diversity across different parts of both the nuclear and chloroplast genomes differ; hence, employing combinations of these highly variable genetic regions is crucial for elucidating phylogenetic relationships among species and subspecies. The integration of non-coding chloroplast DNA (cpDNA) regions psbA-trnH, the chloroplast coding gene ribulose-1,5-bisphosphate carboxylase/oxygenase large subunit (rbcL), and the nuclear ribosomal internal transcribed spacer (ITS) gene region of nuclear DNA (nrDNA) is frequently used in molecular phylogenetic research due to their capacity to furnish adequate data for addressing inquiries related to phylogenetic reconstruction and biogeographical history [40,41,42].

The distribution and division of plant taxa, including Quercus species, have been influenced by phylogeographically defined geographical barriers in Anatolia [10,43], primarily during the Pliocene period [44], such as the hypothetical break Anatolian diagonal (Southern Türkiye) and Colchic region (Northern Türkiye). These barriers also mark transition zones between the major phytogeographical regions of Anatolia, as the Euro-Siberian in the north, the Mediterranean in the south and west, and the Irano-Turanian in the interior and east, with the Anatolian diagonal acting as a southern land bridge between Europe and Asia and playing a crucial role in influencing species ranges [45]. Quercus petraea subsp. pinnatiloba occurs from Türkiye to Syria and South Transcaucasus, with its range likely influenced by the unique biogeographic position of Anatolia [14]. This taxon is particularly found in the mountain of Hatay in southeast Anatolia (Figure 1) [3]. According to most herbarium records in Türkiye, it mainly grows in both the Euro-Siberian and Irano-Turanian phytogeographic regions. Ecologically, Quercus petraea subsp. pinnatiloba represents a distinct oak lineage adapted to xeric environments, typically occurring on dry, rocky slopes and open forest habitats between 1200 and 2200 m above sea level. In contrast to other Q. petraea subspecies, which are commonly found in mesophytic deciduous forests [46], subsp. pinnatiloba is adapted to more arid conditions and is associated with open-canopy forest communities that permit greater light penetration to the understory [3]. It shows a preference for well-drained, acidic soils and occupies ecological niches on plateaus and mountainous terrain.

Despite their significance as a component of forest vegetation, sessile oaks in the Turkish Flora remain relatively poorly understood in terms of their taxonomic and phylogenetic positions. Compared to other countries, the Quercus petraea species in Türkiye has received less attention and research [2,19,47,48,49]. Only a few studies [2,12] focusing on the subspecies of Quercus petraea have been considered to date. Also, there is no exhaustive study of the evolutionary relationships between Quercus petraea subsp. pinnatiloba and other oak species. This study aims to clarify the taxonomic status, phylogenetic relationships, and biogeographic history of Quercus petraea subsp. pinnatiloba, a taxon historically regarded as endemic to Türkiye but recently reported from Iran [13]. We specifically address two main research questions: (1) Does Q. petraea subsp. pinnatiloba represent a morphologically and phylogenetically distinct lineage within the Q. petraea complex in Türkiye? (2) If the subspecies is not geographically restricted to Türkiye, what are its biogeographic origins and historical dispersal patterns? To investigate these questions, we employed an integrative approach combining morphometric analysis of leaf traits (Principal Component Analysis and Stepwise Discriminant Analysis), molecular phylogenetic reconstruction using nuclear and plastid markers (nrDNA ITS, cpDNA rbcL, and psbA–trnH), molecular date estimation, and biogeographic inference based on an ultrametric, time-calibrated Bayesian tree constructed with BEAST. Ancestral area reconstruction was conducted using RASP to explore historical distribution dynamics across biogeographic regions.

2. Materials and Methods

Samples belonging to endemic Quercus species from the southern region of Türkiye were collected mainly from natural stands and identified for morphological analysis. In addition, the sampling included herbarium specimens for molecular analysis using cpDNA (rbcl and psbA-trnH) and nrDNA (ITS) data. The field work for herbarium specimens was carried out from early spring to mid-summer. In the field, a large branch with leaves and fruit (if available) was taken and pressed into herbarium specimens for identification and conservation with voucher numbers TC555 and TC557. Information on the samples’ locations is presented in detail in Figure 1. Species were identified using Flora of Turkey Vol 7 and Aas (1993) [3,17]. Detailed information about the specimen used in the map (Table S1A), in morphological analysis assessment (Table S1B), in the phylogenetic tree (Table S1C), and in biogeographical history (Table S1D) is given in the supporting information file (Table S1). Some Quercus taxa (Q. kotschyana O. Schwarz, Q. x mannifera Lindl, and subspecies of Q. trojana Webb, Q. pubescens Brot) naturally found in Türkiye were excluded from analysis due to the absence of corresponding sequence accessions in the NCBI GenBank database. Even though the rubisco gene sequences for Q. vulcanica Boiss. & Heldr. ex Kotschy and Q. hartwissiana Steven could not be retrieved from the NCBI GenBank database, both taxa were included in the analyses. In addition, although Quercus petraea subsp. iberica has been synonymized as subsp. polycarpa [50], we retain the previous name in this study to ensure consistency with herbarium and NCBI GenBank accessions.

2.1. Morphometric Analysis

For morphometric assessment, three leaves were collected from each of the 61 specimens, including Q. petraea subsp. pinnatiloba (19), subsp. petraea (8), subsp. iberica (18), Q. robur (10), and Q. vulcanica (6) sampled herbarium specimens and analyzed across 14 distinct leaf traits within each population. These traits encompassed five directly measured characteristics, five dimensional, two observed, two counted, and five derived variables, following the methodologies outlined by Kremer et al. (2002) and Yücedağ et al. (2011) [18,19].

The measured traits included five dimensional morphological traits: (1) lamina length (LL), (2) petiole length (PL), (3) lobe width at the tip of the widest lobe (LW), (4) sinus width (SW), and (5) length of leaf blade at its broadest point (WP). Two counted traits: (6) number of lobes (NL) and (7) number of intercalary veins (NV). Two observed traits: (8) basal lamina shape (BS) and (9) lobe tip shape (LT), which were scored as 0 (not pointed), 1 (slightly pointed), or 2 (pointed) [18]. Five derived traits: (10) lamina shape or obversity (OB), (11) petiole ratio (PR), (12) lobe depth ratio (LDR), (13) venation percentage (PV), and (14) lobe width ratio (LWR). All leaf measurements were conducted in a standardized orientation, consistently from the lower right side. Leaf morphological variables used in the Principal Component Analysis (PCA), with their corresponding codes and the evaluated correlation between leaf variables and lamina length (LL), are shown in

Table 1. Leaf morphological variables used in the Principal Component Analysis (PCA), with their corresponding codes and their correlation coefficients with lamina length (LL).

No	Variable	Code	Coefficient of Correlation
1	Petiole length	PL	0.858128
2	Lobe Width at the tip of the widest lobe	LW	0.617328
3	Sinus Width	SW	0.659195
4	Length of Leaf blade at its Widest Point	WP	0.792767
5	Number of Lobes	NL	−0.091260
6	Number of Intercalary Veins	NV	−0.200110
7	Basal Lamina Shape	BS	−0.279870
8	Lobe Tip Shape	LT	−0.210570
9	Lamina Shape or Obversity	OB	−0.245720
10	Petiole Ratio	PR	−0.378410
11	Lobe Depth Ratio	LDR	0.309501
12	Venation Percentage	PV	−0.063140
13	Lobe Width Ratio	LWR	0.124491

. Image analysis was performed by using ImageJ software (Version 1.54g) [51]. Detailed information about correlation matrices and biplot variables can be found in Supplementary Figure S1, Supplementary Figure S2, and Table S1B.

To analyze morphological variation and support taxonomic classification, Principal Component Analysis (PCA) and Stepwise Discriminant Analysis (SDA) were performed using Python (v3) [52]. PCA was implemented as an ordination method using the scikit-learn library, with pandas and matplotlib used for data handling and visualization. The proportion of variance explained by each principal component was calculated as the ratio of its eigenvalue to the total sum of eigenvalues. To evaluate statistical separation among Quercus taxa based on 14 quantitative leaf morphological traits from 61 specimens, SDA was conducted using scikit-learn, pandas, and scipy. Group labels (species or subspecies assignments) were label-encoded, and both within-group and total scatter matrices were computed to derive Wilks’ Lambda. Significance testing was performed using Bartlett’s approximation to obtain Chi-square statistics and corresponding p-values. Classification accuracy was assessed by comparing predicted group assignments with original labels to evaluate the discriminant model’s effectiveness. Based on PCA-derived clusters, SDA identified the most influential traits for discriminant functions, which were then used to assign specimens probabilistically. Individuals showing consistent classification across both methods were assigned to a species, while those with conflicting results remained unclassified to reflect uncertainty (Table S1B). These morphometric analyses included the closest relatives of the taxa, namely Quercus petraea subsp. petraea, Q. petraea subsp. iberica, Q. robur [2], and Q. vulcanica [53].

2.2. Phylogenetic Analysis

The herbarium leaves were preserved in silica gel until DNA extraction. Total DNA was extracted using the DNeasy Plant mini Kit (Qiagen, Germany). The presence and quality of the DNA were assessed using a spectrophotometer. The non-coding chloroplast DNA (cpDNA) regions psbA-trnH, the chloroplast coding gene ribulose-1,5-bisphosphate carboxylase/oxygenase large subunit (rbcL) gene [41], and the nuclear ribosomal internal transcribed spacer (ITS) gene region of nuclear DNA (nrDNA) [54] were amplified with the universal primers. The nuclear ITS region was amplified using primers ITSL(F) and ITS4(R) [54], while the plastid regions rbcL and psbA-trnH were amplified using rbcL1(F)-rbcL2(R) [41] and psbA(F)-trnH(R) [55], respectively. Primer sequences and references are listed in Table S1E. The sequences were generated from TC herbarium specimens (n = 2) conserved at TC, with voucher numbers TC 555 and TC 557, and were deposited in the NCBI GenBank database with accession numbers PV945968, PV945969, PV963096, PV963097, PV963098, and PV963099 (Table S1C). PCR amplifications were conducted in a total reaction volume of 20 μL, comprising 4 μL of 5X HOT FIREPol Blend PCR Mix solution, 0.5 μL of each primer, 3 μL of template DNA, and 12 μL of distilled water, prepared in 0.2 mL sterile Eppendorf tubes. The thermal cycling technique comprised an initial denaturation at 95 °C for four minutes, then denaturation at 94 °C for two minutes, annealing at 57 °C for one minute, extension at 72 °C for two minutes, and a concluding extension at 72 °C for 10 min. PCR results were observed utilizing 1% and 2% agarose gels. BM Biotechnology Company (Ankara, Türkiye) conducted the purification and sequencing. Sequence chromatograms were analyzed utilizing FinchTV software (Version 1.4.0) created by the Geopiza Research Team [56], accompanied by manual verification of base peaks to guarantee the precision of base-calling. Multiple sequence alignments were performed via ClustalW (Version 1.7) [57]. In total, 21 sequences from world Quercus acccessions and two newly generated sequences for Q. petraea subsp. pinnatiloba were used. Molecular diversity statistics such as GC contents (%), nucleotide deletions and insertions, conserved and variable sites, parsimony informative sites, and nucleotide diversity were calculated via MEGA 11 software (Molecular Evolutionary Genetics Analysis) [58]. Maximum likelihood (ML) and Bayesian inference (BI) were estimated separately for the data with 26 sequences including outgroups (Fagus sylvatica, Castanea mollissima, and Trigonobalanus doichangensis).

Phylogenetic relationships and divergence times among Quercus petraea subspecies and closely related taxa were inferred using a combination of maximum likelihood (ML) and Bayesian inference (BI) approaches. Both ITS and cpDNA (rbcL, psbA–trnH) regions were tested independently for best-fitting nucleotide substitution models using MrModelTest v2 [59]. For both partitions, the GTR + G + I model was selected based on AICc scores. ML analysis was performed in MEGA X software (Figure S4). Bayesian molecular dating was conducted using the BEAST v 2.7.1 (Bayesian Evolutionary Analysis by Sampling Trees) package program [60] based on concatenated sequence data from the nuclear ITS and two chloroplast regions (rbcL and psbA–trnH). Both nuclear (ITS) and chloroplast (rbcL, psbA–trnH) sequence datasets were used and linked in the molecular clock analysis, with each sequence set treated as a separate partition and assigned GTR + I + G substitution with four discrete gamma categories.

Sequence partitioning and substitution model selection were separately carried out in BEAUti, with an uncorrelated lognormal relaxed clock applied to all partitions. The analysis was conducted under a Yule tree prior.

A single fossil-based calibration point was applied to the stem node of Quercus, defined as the most recent common ancestor (MRCA) of Quercus and its sister genus Trigonobalanus in priors. This node was calibrated using a lognormal prior distribution, with an offset of 60 million years, a mean of 1.0, and a standard deviation of 0.5 in log space. This calibration reflects the earliest fossil evidence of crown Fagaceae from the Campanian stage of the Late Cretaceous and is supported by fossil records [61].

In the BEAST analysis for each node, the mean divergence age, along with the 95% highest posterior density (HPD) interval, calculated from the maximum clade credibility (MCC) tree in TreeAnnotator, is reported. All divergence time were estimated

The final XML configuration file was manually validated to ensure the accurate specification of priors and model parameters. Markov Chain Monte Carlo (MCMC) analyses were run for 50 million generations, with sampling every 5000 generations. Convergence diagnostics and effective sample size (ESS) values were assessed in Tracer v1.6 [62]. The ESS values exceeded 200 for all calibration and substitution parameters, and posterior distributions showed well-formed, bell-shaped curves, indicating reliable convergence. TreeAnnotator v2.7.7 [63] was used to summarize the posterior distribution and generate the maximum clade credibility (MCC) tree based on a Bayesian posterior probabilities threshold of ≥0.95. In addition, the bootstrap values (limit of 100) constructed in the MEGA program to estimate bootstrap with 1000 replicates in ML analysis were added to the Bayesian phylogenetic tree (Figure 2). The final tree was visualized in Figtree 1.4.3 [64].

2.3. Biogeographical Analysis

Historical biogeographical reconstructions for Quercus species were performed using RASP (Reconstruct Ancestral State in Phylogenies) software [65,66] based on phylogenetic trees generated from linked ITS and cpDNA (rbcL, psbA-trnH) sequence data sampled from different phytogeographical regions across Türkiye. Ancestral area reconstruction was performed using Statistical Dispersal–Vicariance Analysis (S-DIVA), a model-based method that estimates ancestral geographic ranges by integrating both dispersal and vicariance events within a phylogenetic framework. Unlike traditional DIVA, S-DIVA incorporates phylogenetic uncertainty by summarizing ancestral range probabilities across a set of posterior trees, rather than relying on a single topology. This approach provides statistical support for alternative biogeographic scenarios by evaluating distributions across nodes of the phylogeny [67]. The analysis was conducted in RASP, using the maximum clade credibility (MCC) tree generated in BEAST, with a maximum of two areas per node, based on 50,000 trees and a 10% burn-in. The geographic distribution of each taxon was coded discretely based on its known occurrence within seven defined biogeographic regions of Türkiye: Marmara (A), Black Sea (B), Central Anatolia (C), Aegean (D), Mediterranean (E), Eastern Anatolia (F), and Southeastern Anatolia (G). Outgroup taxa (Fagus sylvatica, Castanea mollissima, and Trigonobalanus doichangensis) were assigned to non-overlapping geographic codes (H–J) to improve root inference while avoiding confounding signals within the focal region (Table S1D). This coding was based directly on the collection localities of herbarium specimens, allowing for precise and reproducible assignments, and follows approaches used in previous studies on endemic plant groups from Türkiye [68,69]. For biogeographic interpretation, these regions were further aligned with Türkiye’s recognized phytogeographical zones [70] to capture broader ecological and historical patterns: (i) Black Sea and Marmara = Euro-Siberian, (ii) Aegean and Mediterranean = Mediterranean, and (iii) Central and Eastern Anatolia = Irano-Turanian. The analysis was restricted to extant taxa, as fossil or extinct Quercus lineages with geographically assignable data were not included.

3. Results

This study integrates morphometric analysis, utilizing dimensional, observed, counted, and derived variables, with molecular data from rbcL and psbA-trnH genes of cpDNA and ITS of nrDNA, alongside molecular dating and historical biogeographical reconstructions, to thoroughly assess the status of Quercus petraea subspecies native to Türkiye.

3.1. Leaf Morphometric Analysis

Correlation between the variables used and the size of the leaf (leaf blade length) for 61 herbarium specimens was evaluated. The first three principal components (PC1 = 30.35%, PC2 = 16.08%, PC3 = 14.92%) together explained 61.35% of the total morphological variation, indicating substantial dimensional reduction and supporting the robustness of the observed clustering pattern. Considering variables of PC1, it was obvious that petiole length (Pl), leaf blade width at its broadest point (WP), lamina length (LL), and lobe width at the tip of the widest lobe (LW) had high loading scores. For PC2, the most influential variables were lamina shape or obversity (OB), petiole ratio (PR), lobe depth ratio (LDR), and lobe width ratio (LWR). In PC3, the dominant variables included the number of lobes (NL), lobe tip shape (LT), OB, and PR (

Table 2. The variables that contribute to the PC1, PC2, and PC3 (Loading Matrix).

Variables	PC1	PC2	PC3
LL	0.4224 *	−0.2619	−0.1118
PL	0.4423 *	−0.0508	0.0974
LW	0.4038 *	0.1296	−0.0971
SW	0.2914	−0.3618	0.1593
WP	0.4241 *	−0.0258	0.1863
NL	−0.1368	−0.3115	0.2142 *
NV	−0.2015	−0.1276	−0.1534
BS	−0.1224	0.2890	−0.037
LT	−0.1285	−0.1255	0.3578 *
OB	0.0258	0.3428 *	0.4935 *
PR	−0.0233	0.4045 *	0.4429 *
LDR	0.2147	0.3293 *	−0.2604
PV	−0.0393	0.2223	−0.4489
LWR	0.2416	0.3589 *	−0.0261

* The most significant value.

).

The 14 variables were selected a priori for their proven taxonomic relevance in Quercus petraea and Q. robur [18,19], making dimensionality reduction and collinearity removal unnecessary. PCA was applied primarily as an exploratory tool to visualize clustering and identify potentially discriminant variables, rather than to eliminate them [71]. SDA revealed statistically significant morphological differentiation among the examined Quercus taxa, yielding a Wilks’ Lambda value of 0.0433, indicating strong separation between groups, with a corresponding Chi-square statistic of 174.31 (df = 56, p < 0.0001). The model demonstrated robust classification performance, correctly assigning 81.82% of specimens to their respective taxonomic groups. When SDA was re-run using only the variables with the highest contributions to the first three principal components (SW, NV, BS, and PV removed), the classification accuracy decreased to 63.64% (Wilks’ Lambda = 0.1471; χ² = 110.20, df = 40, p < 0.0001). Due to this reduction in performance, the reduced-variable model was not used in the final analysis. Based on selected 14 morphometric characters, the two-dimensional configuration of the factor analysis of mixed data formed three major clusters as Quercus petraea subsp. pinnatiloba in blue, Q. vulcanica in purple, and mixed colored circles with Quercus robur, Q. petraea subsp. petraea, and Q. petraea subsp. iberica (Figure 2). Also, the mixed three taxa (Quercus robur in the red circle, Quercus petraea subsp. petraea in orange, and Quercus petraea subsp. iberica in the green circle) were not effectively discriminated in the analysis. A biplot of the Principal Component Analysis (PCA) was also created and is presented as Supplementary Figure S2 to illustrate the links between taxa and associated morphological characteristics. The members of the taxa Quercus petraea subsp. pinnatiloba and Q. vulcanica were grouped at the left and right corners, respectively, and differentiated by PC1, which explained 30.35% of the total variance (Figure 2). The central cluster formations in Figure 2 indicated that Q. petraea subsp. petraea and Q. robur were nearly differentiated by PC2, which accounted for 16.08% of the explained variance. Both taxa are in the same section (sect. Quercus (L.) Loudon) and show similarities in terms of obovate and deeply lobed leaves, except for pedunculate length in Q. robur, which can be up to 12 cm long.

Quercus petraea subsp. pinnatiloba from the Hatay provinces developed a completely distinct cluster in the left corner of Figure 2. Certain members of the taxa were clustered with other members of the subspecies (petraea and iberica) of the Q. petraea. The uniqueness of subsp. pinnatiloba is predominantly morphological, defined by deeply lobed leaves measuring up to 17 cm in length and the absence of intercalary veins.

3.2. The Level of Sequence Variation

The total lengths of the nrDNA ITS, rbcL, and psbA-trnH gene regions were 657, 1462, and 292 base pairs, respectively (

Table 3. Molecular diversity statistics calculated for cpDNA and nrDNA gene regions of Quercus species, with all sequence lengths reported in base pairs (bp).

	nrDNA	cpDNA		Total
	ITS	rbcL	psbA-trnH	total
Number of taxa	22
Number of newly generated sequences for the gene (nrDNA/cpDNA) concerned	2
Number of sequences used from the NCBI GenBank database	24
Number of outgroups used in the phylogenetic tree	3
Total number of sequences used in the tree	26
Total length (bp)	657	1462	292	2411
GC content (%)	60.06	43.3	24.7	46.9
Conserved sites	628	1442	291	2362
Variable sites	29	19	1	49
Parsimony informative sites	25	16	1	42
Transitional pairs	81.14	60.2	99.01	75.69
Transversional pairs	18.86	38.8	0.09	24.31
Transition/Transversion (tr/tv) (R) ratio	4.00	1.50	3.19	3.12
Nucleotide diversity	0.010	0.001	0.001	0.010

). The polymorphism sites were more variable in the nrDNA ITS (29 of 657) compared to the cpDNA (20 of 1754) gene regions. Among the studied gene regions, the most conserved sites were identified in psbA-trnH (291/292), whilst the most variable sites were in the ITS (29/657) gene region. The nucleotide diversity, indicative of overall polymorphism, was calculated at 0.01 for the nrDNA ITS region. The highest transition rates were observed in psbA-trnH (99.01%), the largest transversion substitution in rbcL (38.8%), and the transition–transversion bias (R) in ITS (4). The ratio of parsimony informative sites to variable sites was greater in the nrDNA ITS area (25 of 29) than in the total cpDNA regions (17 of 20) (Table 3).

3.3. Phylogenetic Analyses and Molecular Dating

The analysis of the nrDNA and cpDNA gene region (ITS + rbcL + psbA-trnH) identified two clade formations with a bootstrap support value of 95 (Figure 3). The major groups were sections Quercus (green, Clade 1), Ilex Loudon (blue, Clade 2), and Cerris Loudon (orange, Clade 3). Among Clade 1, the Quercus section, the subspecies of the Q. petraea, nested together with a high posterior probability value of 0.99. Other members of the sect. Quercus members, Q. macranthera, Q. vulcanica, Q. dalechampii sensu E-European Auct., Q. pontica, and Q. infectoria, are positioned very close to the subclade of Q. petraea subspecies subclade in Clade 1 with high supportive pp values of 0.96. The Quercus petraea subsp. pinnatiloba (TC555, TC557) samples formed a well-supported monophyletic group (pp = 1/bs = 84), distinct from Q. petraea subsp. petraea and subsp. iberica, suggesting its independent evolutionary lineage. Clade 2 was dominated by Ilex members (Q. coccifera and Q. aucheri) except for one Quercus section member, Q. ilex. Clade 3 consists predominantly of the members of the sect. Cerris.

The divergence between Quercus and its sister genus Trigonobalanus was dated to approximately 60 million years ago (95% HPD 59, 60.89 Mya), based on fossil-calibrated lognormal priors. Divergence events among major Quercus lineages were estimated to have occurred mainly during the Oligocene to Miocene, reflecting deep historical diversification within the genus. Molecular dating suggests that all Quercus species began diversifying around 30 Mya (95% HPD 13.15, 49.22 Mya), with the emergence of section Quercus estimated at 20.8 Mya (95% HPD 8.08, 36.01 Mya), and sections Ilex and Cerris dated to approximately 21.1 Mya (95% HPD 7.13, 37.83 Mya), respectively (

Table 4. Divergence time estimates from BEAST analysis with mean ages and 95% highest posterior density (HPD) intervals for the major nodes.

No	Node Name	Mean (Mya)	95%HPD Lower Mya	95%HPD Upper Mya	Fossil Calibration Point (**) [Reference]
1	The crown age of Quercus sp.	30.13	13.15	49.22
2	The crown age of sect. Quercus	20.83	8.08	36.01
3	The crown age of sections Ilex and Cerris	21.1	7.13	37.83
4	The crown age of Quercus petraea subsp. pinnatiloba from Q. petraea	11.04	3.01	20.95
5	The crown age of Quercus petraea subsp. pinnatiloba from sect. Quercus	16.02	5.75	28.41
6	The crown age of Trigonobalanus from Quercus sp.	60	59	60.89	60 Mya [61,72]

). The crown age of Quercus petraea subsp. pinnatiloba, indicating its divergence within the Q. petraea subspecies, was estimated at 11 Mya (95% HPD 3.01, 20.95 Mya), while its stem age, reflecting divergence from other members of section Quercus, was assessed at 16 Mya (95% HPD 5.75, 28.41 Mya) (Figure 3). These findings indicate a relatively recent emergence of Q. petraea subsp. pinnatiloba, succeeded by an extended phase of ongoing evolution, aligning with its unique morphological and genetic characteristics.

3.4. Historical Biogeography Reconstructions (RASP)

The RASP analysis using ITS + rbcL + psbA-trnH sequences, with selection of the S-DIVA model and MCMC runs, was carried out to estimate a biogeographical history in which divergence and vicariance have been important in the speciation of Quercus subspecies (Figure 4). The analysis inferred a total of 48 dispersal, 10 vicariance, and 5 extinction events have been postulated with ITS + rbcL + psbA-trnH data. The Black Sea region (B) appears to be a major source of dispersal, with 26 events originating. The RASP analysis indicated that the ancestors of all Q. petraea subspecies in Türkiye most likely evolved within two major phylogeographic regions: the Euro-Siberian Region of the Black Sea zone (B), situated along the northern slopes of the Anatolian Diagonal, and the Mediterranean region (E) in southern Türkiye (Figure 4). At present, members of sect. Quercus are predominantly distributed within the Euro-Siberian Region (B), whereas members of sect. Cerris show greater adaptation to the Irano-Turanian (F) and Mediterranean (E) regions. The Mediterranean region (E) exhibited a balanced pattern of dispersal events, with frequencies ranging from 10.0 to 6.0, coupled with a comparatively high number of intra-regional dispersals (6).

Ancestral area reconstruction suggests that the most recent common ancestor of Q. petraea subsp. pinnatiloba likely originated in the Euro-Siberian region (Black Sea(B)), with subsequent dispersal into the Mediterranean region (E), as indicated by Node 32. Nodes 27, 28, 45, and 46 exhibit low probability values (<0.1) for restricted ancestral areas, which may reflect a historically widespread distribution or past gene flow and range expansion events across regions. Nodes 41, 42, 43, and 44 show a dispersal event with full support (probability = 1), indicating movement from the Euro-Siberian region (Black Sea(B)) to a broader area encompassing both the Black Sea and Mediterranean regions (BE). In the ancestral area reconstruction of the Euro-Siberian region, it may have served as a Colchic origin and a historical center of expansion for this lineage.

4. Discussion

4.1. Morphometric Analysis

The Principal Component Analysis (PCA) in this study provided valuable insights into morphological differentiation among Quercus taxa, particularly highlighting the distinctiveness of Q. petraea subsp. pinnatiloba. In particular, the observed reduction in classification accuracy when utilizing only high-loading PCA variables suggests that characteristics with lower PCA loadings can still make a significant contribution to effective group discrimination. Because PCA maximizes overall variance without explicitly considering group structure, the removal of such variables can reduce the discriminatory power of SDA [73,74]. Retaining all 14 morphometric traits, therefore, ensured the most robust classification results and provided the strongest support for the taxonomic distinctions observed. Among the principal components, PC1 explained 30.35% of the total variance and was predominantly influenced by leaf traits, including petiole length, lamina length and width, and lobe width. These traits have also been identified in previous studies as major contributors to both inter- and intraspecific variation in oaks, further supporting their diagnostic value. Early work by Aas (1993) demonstrated that leaf morphological variation in Quercus petraea and Q. robur is significantly influenced by environmental conditions and hybridization, showing the complexity of juvenile phenotypes [17]. Jensen et al. (1995) and Bruschi et al. (2003) also highlighted the importance of quantitative leaf measurements in distinguishing white oak taxa across broad geographic scales [15,21]. Adding to this, Fortini et al. (2024) applied landmark-based geometric morphometrics across 18 European populations of Q. petraea, revealing clear patterns of leaf shape and size variation correlated with latitudinal gradients and isotopic signatures [75]. Moreover, Gugerli et al. (2007) found spatial discontinuities in leaf morphology that coincided with genetic lineage between Q. petraea and Q. robur in mixed forests, illustrating how even subtle leaf traits can reflect introgression and phylogeographic structure [76]. The significance of utilizing morphometric techniques to elucidate fine-scale variation within Q. petraea is confirmed by all of these findings, which also support the use of such features in complex taxonomic and evolutionary investigations.

The PCA-based clustering revealed three main morphological groups: Q. petraea subsp. pinnatiloba, Q. vulcanica, and a mixed group comprising Q. robur, Q. petraea subsp. petraea, and Q. petraea subsp. iberica. The significant overlap among the last three taxa aligns with prior research highlighting the morphological continuum and possible hybridization between Q. petraea and Q. robur [18]. This pattern further exemplifies the increased complexity within the Q. petraea group, which remains taxonomically unclear in numerous areas of Eastern Europe and the Balkans [77]. In particular, Q. dalechampii sensu E-European Auct. has been shown to align more closely with the Q. pubescens group by herbarium specimens, further putting forward the challenges of species delimitation within morphologically similar white oaks [29]. These closely related taxa (Q. robur, Q. petraea subsp. petraea, and Q. petraea subsp. iberica) frequently co-occur and exhibit overlapping leaf characteristics, rendering morphological differentiation challenging. However, our morphometric analyses reveal a notable pattern: Quercus petraea subsp. pinnatiloba consistently separates from both Q. petraea subsp. petraea and subsp. iberica, while the latter two cluster more closely with Q. robur. In addition, Q. petraea subsp. pinnatiloba, especially individuals from Hatay province (southern Türkiye), formed a clearly separated cluster in the ordination space. This separation is supported by distinctive morphological features in the Flora of Turkey [3], such as deeply lobed leaves reaching up to 17 cm long and the absence of intercalary veins, which set it apart from other Q. petraea subspecies. These characteristics, while observed in earlier regional floristic works [19], have rarely been evaluated through quantitative analysis. This unexpected result suggests that subsp. pinnatiloba is more morphologically distinct from other members of the Q. petraea complex than previously assumed. The distinctiveness of subsp. pinnatiloba is further supported by both qualitative and quantitative traits, such as deeper lobes, longer petioles, and distinct lamina proportions, as well as by its genetic divergence revealed in our phylogenetic analyses.

Q. vulcanica constituted a different group in the analysis, corroborating prior reports that it preserves a stable leaf morphology and does not easily hybridize with closely related taxa. Its location substantiates its identification as a distinct species; however, additional DNA analysis may be required for confirmation. This discovery corresponds with overarching trends in Quercus, wherein leaf morphology is frequently inadequate for species delimitation because of significant intraspecific variation and the potential for hybridization [19,43]. Our PCA and clustering analyses indicate that Q. petraea subsp. pinnatiloba can be morphologically differentiated from other subspecies and species, particularly by traits including vein absence, petiole length, maximum lamina width, lamina length, and lobe width at the apex of the widest lobe. Thus, a morphometric basis to support its taxonomic validity suggests that subsp. pinnatiloba represents a morphologically stable and geographically localized lineage. It is proposed in this study that the taxa can be elevated to the species level as Quercus pinnatiloba, in accordance with the original classification by C. Koch (Linnaea 22: 326, 1849), where it was first described under this name [78].

4.2. Phylogenetic Position

The analysis of sequence polymorphism among the nrDNA and cpDNA regions (ITS + rbcL + psbA-trnH) revealed clear differences in their informativeness for phylogenetic reconstruction. The nrDNA ITS region displayed the highest level of variation, with 29 polymorphic sites out of 657 bp and a nucleotide diversity of 0.01. In contrast, the cpDNA regions were more conserved, particularly psbA-trnH, which exhibited only one variable site among 292 bp. The limited variation in cpDNA regions compared to nuclear ones, as previously reported in studies on Fagaceae [79,80], was similarly observed in the present study. The high transition rate in psbA-trnH and the higher transversion rate observed in rbcL reflect different evolutionary dynamics between coding and non-coding plastid regions. Consistent with previous studies, the nrDNA ITS region exhibited greater polymorphism than the chloroplast regions (rbcL, psbA-trnH), showing the typically faster evolutionary rate and higher variability of nuclear markers [81]. The elevated variability in the ITS region provided enhanced phylogenetic resolution within the Quercus petraea group, enabling clearer distinction of subsp. pinnatiloba from closely related taxa, consistent with previous studies [79,82].

The phylogenetic analyses based on the combined dataset (ITS + rbcL + psbA-trnH) supported the formation of three major clades corresponding to the recognized sections Quercus, Ilex, and Cerris. Within Clade 1, which represented the sect. Quercus, Q. petraea subspecies formed a well-supported subclade (pp = 0.99) showing a close genetic relationship among these taxa. The placement of Q. petraea subsp. pinnatiloba adjacent to other morphologically similar species such as Q. macranthera, Q. vulcanica, Q. dalechampii sensu E-European Auct. Q. pontica, and Q. infectoria may reveal shared ancestral traits or historical gene flow within a specific geographic or ecological context. The coexistence of these taxa within the same subclade illustrates the intricate evolutionary relationships in the Quercus section, characterized by reticulate evolution and significant hybridization potential [8,18]. Tekpinar et al. (2021) also put forward high allelic richness and diverse relations among members of sect. Quercus [2]. Importantly, Q. petraea subsp. pinnatiloba was nested within this group but maintained a consistent and distinguishable position, showing its distinctiveness from other subspecies such as petraea and iberica. This phylogenetic placement complements the morphological evidence that Q. petraea subsp. pinnatiloba represents a genetically coherent lineage within Q. petraea. This conclusion is further confirmed by Kansu et al., unpublished data 2025 [83], which showed that ten chloroplast microsatellite markers successfully identified all cpDNA lineages in other Q. petraea populations across Türkiye, but failed to amplify in subsp. pinnatiloba, pointing to a potential divergence in its chloroplast genome.

The results of this study support the placement of Q. petraea subsp. pinnatiloba within Quercus sect. Quercus, in line with both its morphological characteristics and current taxonomic treatment. The integrated morphological and molecular evidence suggests a degree of differentiation that may justify its classification as a distinct species. This interpretation illustrates a wider trend noted in recent biosystematic research from Eastern Europe, Türkiye, and the Caucasus, where comprehensive morphological and molecular investigations increasingly endorse taxonomic differentiation within once extensive oak complexes [84,85]. Conversely, research from Central and Western Europe, especially on the Italian Peninsula, predominantly highlights morphological continuity and genetic mixing, advocating for taxonomic consolidation within the Q. petraea and Q. pubescens group [21]. This disparity may arise not just from methodological issues but also from biogeographic trends, such as environmental heterogeneity, regional lineage diversification, and the influence of eastern glacial refugia as reservoirs of ancestral diversity [75].

4.3. Molecular Dating and Biogeographic History

The molecular dating results presented here align with prior studies that place the origin of the crown Fagaceae in the Late Cretaceous, 60 million years ago (95% HPD 59, 60.89 Mya), consistent with fossil-calibrated phylogenies [61,72]. The diversification of Quercus around 30 Mya (95% 13.15, 49.22 Mya) corresponds with major climatic shifts during the Oligocene–Miocene boundary, a period widely recognized as a driver of oak radiation across Eurasia [86,87]. The concurrent divergence of sections Quercus, Cerris, and Ilex further supports a scenario of rapid lineage diversification under increasing ecological heterogeneity [7,75]. The estimated Miocene origin of Q. petraea subsp. pinnatiloba supports a scenario of post-Oligocene diversification, likely driven by regional climatic fluctuations and geographic isolation, as proposed for other Anatolian and Caucasian oak taxa [7,82]. This evolutionary timescale aligns well with the ancestral area reconstruction, which identified a primary diversification center in the Euro-Siberian. These findings suggest that Q. subsp. pinnatiloba represents a long-isolated lineage shaped by the complex biogeographic history of Anatolia, a region known for its high oak endemism and role as a Pleistocene refugium [87,88].

S-DIVA analysis of ITS, rbcL, and psbA-trnH inferred 48 dispersal, 10 vicariance, and 5 extinction events, indicating dispersal as the dominant process [66,89]. The Euro-Siberian region appears to be a major source of dispersal, with numerous events originating there. The phytogeographical region may have functioned as a historical center of expansion or a glacial Colchic region [90,91]. In contrast, the Mediterranean showed balanced incoming and outgoing dispersals with many within-region events, suggesting both in-situ diversification and long-term persistence [92,93]. Although extinct lineages were not incorporated, the time-calibrated phylogeny provides a temporal framework for the interpretation of biogeographic patterns and emphasizes the importance of incorporating fossil data in future studies.

The ancestral area reconstruction revealed that the ancestors of all Q. petraea subspecies in Türkiye likely evolved in the northern climates of the Anatolian Diagonal, specifically within the Black Sea region of the Euro-Siberian phytogeographical zone, and in the southern parts of Türkiye corresponding to the Mediterranean portion of the Irano-Turanian region. This dual-origin signal supports the idea that both northern mesic forests and southern xeric habitats contributed to the diversification of white oaks in the region [6,94]. The current distribution patterns also support these reconstructions: members of section Quercus are now largely restricted to the Euro-Siberian, while section Cerris taxa are more common in the Mediterranean. This spatial segregation aligns with ecological differentiation between regions due to the Anatolian diagonal [10] and may reflect long-term climatic adaptations [18,82]. Recent phylogeographic research also indicates that Quercus section Quercus possesses ancient plastid lineages concentrated in the Middle East and Caucasus, representing longstanding refugial and diversification hotspots [84,85,95]. The existence of ancestral haplotypes in the eastern Mediterranean, as documented by Fortini et al. (2024), reinforces a common evolutionary origin of section Quercus [75]. These results are consistent with prior research and serve as additional evidence of the unique evolutionary paths that the Quercus and Cerris sections have pursued and the diverse regions in which they have been found throughout Türkiye.

Evidence of additional refugial areas in central and southern Italy (including Sicily and Sardinia islands) [96], the Balkans, and Anatolia supports a broader biogeographic model in which both eastern and southern Europe played integral roles in shaping the genetic structure of modern European white oaks [97]. These historical patterns have profound implications for understanding postglacial recolonization processes and show the importance of the Eastern Mediterranean, particularly Anatolia, as a critical center of oak diversification and long-term persistence [2,98,99]. Many Quercus populations in Türkiye are near potential glacial refugia, especially along the western Black Sea coast [19], consistent with the role of these areas as long-term reservoirs for temperate forest species [100].

These marginal and Mediterranean populations of Quercus petraea subsp. pinnatiloba faces increasing threats from habitat loss driven by urban expansion, recurrent wildfires, and intensive silvicultural practices. In addition, in Eastern Anatolia, prolonged pressure from cutting and overgrazing has resulted in its occurrence primarily as low-growing, shrubby forms rather than mature trees [3]. These disturbances not only reduce the taxon’s ecological functionality but may also limit its reproductive capacity and long-term survival potential.

5. Conclusions

This study provides integrative evidence from morphometric, molecular, and biogeographical analyses supporting the distinctiveness of Quercus petraea subsp. pinnatiloba. Morphometric analyses clearly separated the taxon from closely related species and subspecies based on leaf traits such as lamina length, lobe depth, and the absence of intercalary veins. Molecular phylogenetic analyses using nuclear and plastid DNA gene regions confirmed that while Q. petraea subsp. pinnatiloba belongs within the Quercus section; it reveals a phylogenetically different lineage from other Q. petraea subspecies. Biogeographical reconstruction further indicates the Euro-Siberian region served as the most likely ancestral area, with the recent origin of the taxon estimated to date back to the Miocene. Subsequent dispersal into the Mediterranean reflects a dynamic history of refugial persistence and expansion. In accordance with its original description, the recognition of Q. petraea subsp. pinnatiloba as a distinct evolutionary lineage is warranted by all these findings, which collectively support its elevation to species rank as Quercus pinnatiloba (C. Koch, 1849). Given its restricted distribution and increasing anthropogenic pressures in southeastern Türkiye, this taxon should also be prioritized in regional conservation planning. These findings not only support its morphological and ecological distinctiveness but also emphasize its relevance for conservation and future evolutionary studies in the genus Quercus.