Niche evolution in a northern temperate tree lineage: biogeographical legacies in cork oaks (Quercus section Cerris)

Abstract Background and Aims Cork oaks (Quercus section Cerris) comprise 15 extant species in Eurasia. Despite being a small clade, they display a range of leaf morphologies comparable to the largest sections (>100 spp.) in Quercus. Their fossil record extends back to the Eocene. Here, we explore how cork oaks achieved their modern ranges and how legacy effects might explain niche evolution in modern species of section Cerris and its sister section Ilex, the holly oaks. Methods We inferred a dated phylogeny for cork and holly oaks using a reduced-representation next-generation sequencing method, restriction site-associated DNA sequencing (RAD-seq), and used D-statistics to investigate gene flow hypotheses. We estimated divergence times using a fossilized birth–death model calibrated with 47 fossils. We used Köppen profiles, selected bioclimatic parameters and forest biomes occupied by modern species to infer ancestral climatic and biotic niches. Key Results East Asian and Western Eurasian cork oaks diverged initially in the Eocene. Subsequently, four Western Eurasian lineages (subsections) differentiated during the Oligocene and Miocene. Evolution of leaf size, form and texture was correlated, in part, with multiple transitions from ancestral humid temperate climates to mediterranean, arid and continental climates. Distantly related but ecologically similar species converged on similar leaf traits in the process. Conclusions Originating in temperate (frost-free) biomes, Eocene to Oligocene ranges of the primarily deciduous cork oaks were restricted to higher latitudes (Siberia to north of Paratethys). Members of the evergreen holly oaks (section Ilex) also originated in temperate biomes but migrated southwards and south-westwards into then-(sub)tropical southern China and south-eastern Tibet during the Eocene, then westwards along existing pre-Himalayan mountain ranges. Divergent biogeographical histories and deep-time phylogenetic legacies (in cold and drought tolerance, nutrient storage and fire resistance) thus account for the modern species mosaic of Western Eurasian oak communities, which are composed of oaks belonging to four sections.


INTRODUCTION
The genus Quercus L. (oak trees) is one of the most economically and ecologically important woody angiosperm genera in the Northern Hemisphere. Oaks comprise ~425 species and occur in a wide range of habitats, from dry woodlands to swamp forests and from lowlands to elevations ≤4500 m a.s.l. (Camus, 1936(Camus, -1954Denk et al., 2017a). They are a dominant component of the northern temperate forests (Martinetto et al., 2020). Traditionally, the taxonomy of this genus has been based on key morphological characters, and different classification schemes have been proposed over the centuries (e.g. Ørsted, 1871;Trelease, 1924;Schwarz, 1936;Camus, 1936Camus, -1954Menitsky, 1984;Nixon, 1993; for differences in these schemes, see Denk and Grimm, 2010;Denk et al., 2017a). In recent years, a number of morphological (Solomon, 1983a(Solomon, , 1983bDenk and Grimm, 2009;Denk and Tekleva, 2014) and molecular (Manos et al., 2001Oh and Manos, 2008;Denk and Grimm, 2010;Hipp et al., 2014Hipp et al., , 2018Hipp et al., , 2020Hubert et al., 2014;Cavender-Bares et al., 2015;Simeone et al., 2016Simeone et al., , 2018McVay et al., 2017aMcVay et al., , 2017bPham et al., 2017;Vitelli et al., 2017;Deng et al., 2018;Ortego et al., 2018;Cavender-Bares, 2019;Jiang et al., 2019;Crowl et al., 2020) studies have provided a robust phylogenetic framework along with a revised subgeneric and sectional classification. Together, these studies provide a framework to place the extensive fossil record of the genus in a phylogenetic context (e.g. Bouchal et al., 2014;Grímsson et al., 2015Grímsson et al., , 2016. Leaf phenology and climatic niche have evolved in concert across woody angiosperms (Woodward et al., 2004;Hawkins et al., 2014;Zanne et al., 2014;Edwards et al., 2017;. A broad phylogenetic comparative study has shown that leaf phenology can evolve as a response to a change in environment ('climate first') or arise first and predispose lineages to freezing tolerance ('trait first'), with the 'climate first' pathway being more frequent, particularly in deciduous woody plant lineages (Zanne et al., 2014). Understanding the detailed history of these patterns of evolution requires dissection of individual clades. In Viburnum, for example, deciduousness evolved in situ as populations were subjected to gradual cooling (Edwards et al., 2017). In the American ('New World') oak clade (Quercus subgenus Quercus), one of two major clades within Quercus, >20 independent shifts from deciduous to evergreen leaf phenology in Mexican white and red oaks (Quercus sections Quercus and Lobatae) are associated with climatic and edaphic shifts, and the evergreen habit is inferred to have evolved in response to decreases in temperature seasonality and decreases in both winter and summer temperature extremes . In contrast, in the Eurasian ('Old World') oak clade (subgenus Cerris), leaf phenology is stable within sections. Both the deciduous cork oaks (section Cerris) and the evergreen holly oaks (section Ilex) show considerable distributional overlap with each other and subgenus Quercus. Moreover, section Cerris is the only oak section to reach its highest species richness and absolute phylogenetic diversity in Western Eurasia.
Here, we investigate these two closely related sections of the Eurasian ('Old World') oak clade. We test the trait first and climate first pathways and provide detailed insights into two sister lineages of oaks. We use a reduced-representation nextgeneration sequencing method, restriction-site associated DNA sequencing (RAD-seq; Ree and Hipp, 2015), to infer a fully resolved species phylogeny of the Eurasian cork oaks, including 25 ingroup specimens representing 14 of the 15 species (all except for the narrow-endemic Quercus euboica). We use 47 fossil taxa to date the phylogeny using a fossilized birth-death (FBD) approach and to reconstruct the biogeographical history of the cork oaks. Based on habitats, climatic preferences and leaf morphologies of modern and fossil cork oaks, we investigate the history of climatic niche evolution and the correlation of niche shifts with leaf evolution. Furthermore, we compare the biogeographical histories of sections Cerris and Ilex to explain modern ranges and niche occupancy of these sections across Eurasia. Finally, we discuss our results against the background of palaeogeographical and tectonic changes in Eurasia during the past 40 million years and diversification patterns established from previous studies of the nuclear and plastid genomes.

Sampling
Samples from 62 individuals included in the study by  were re-analysed for this study. Sampling details, vouchers, National Center for Biotechnology Information (NCBI) short-read archive (SRA) project and accession numbers are provided in the Supplementary data (Table  S1). Twenty-five samples covered all species of section Cerris except for Q. euboica (Papaioannou) K.I.Chr., for which we were unable to obtain fresh material with sufficient DNA yield. To represent the sister section Ilex, we included all its Western Eurasian species: four Mediterranean species (five if Quercus calliprinos Webb is considered a separate species) plus the western Himalayan-Hindukush Quercus baloot Griff. (Clade VI in the study by Jiang et al., 2019; see also Simeone et al., 2016Simeone et al., , 2018. Additional East Asian Ilex species were selected to represent the major lineages within this section: five species, including Quercus floribunda Lindl. ex A.Camus, for the Himalayan clade carrying Ilex-specific plastomes (cf. Yan et al., 2019;Hipp et al., 2020; Clade IV+V in the study by Jiang et al., 2019); and three species representing the East Asian clade: the (central) Chinese Quercus baronii Skan, morphologically similar to Cerris oaks but with a unique plastome; Quercus dolicholepis A.Camus, a montane central Chinese species; and Quercus phillyreoides A.Gray, a widespread north-eastern Asian subtropical to temperate species. The latter two have Cerris-similar plastid signatures (Simeone et al., 2016; Clade II in the study by Jiang et al., 2019). Five species of section Cyclobalanopsis (Oerst.) Benth. & Hook.f. represented the third lineage within subgenus Cerris, resolved as early diverged sister lineage of sections Cerris + Ilex . As a further outgroup, we included 15 samples covering Western Eurasian members of subgenus Quercus (sections Quercus and Ponticae Stef.), one eastern North American red oak (Quercus coccinea Münchh., section Lobatae Loudon), and the western North American relict genus Notholithocarpus Manos, Cannon & S.H.Oh, the most probable closest living relative of oaks Zhou et al., 2022).

RAD-seq data generation and clustering
Next-generation sequencing libraries were prepared at Floragenex (Portland, OR, USA) following the methods of Baird et al. (2008) with PstI, barcoded by individual, and sequenced in 150-bp single-end reactions on an Illumina HiSeq 2000, 2500 or 4000 at the University of Oregon Genomic Facility; past analyses (Hipp et al., 2014 demonstrate that phylogenetic results in this sample set are not obviously influenced by variation in the sequencing platform. FASTQ files were demultiplexed and filtered to remove sequences with more than five bases of quality score <20 and assembled into loci for phylogenetic analysis using ipyrad v.0.7.24 (Eaton, 2014) at 85 % sequence similarity. Consensus sequences for each individual for each locus were then clustered across individuals, initially retaining loci present in at least four individuals and possessing ≤20 single-nucleotide polymorphisms and eight indels across individuals. Data were imported into R using the RADami package (Hipp et al., 2014) to filter loci for analysis into three datasets, containing a minimum of 15, 20 or 25 individuals per locus (m15, m20 and m25, respectively). Loci were concatenated into a single data partition for maximum likelihood (ML) and Bayesian phylogenetic analyses, and locus identities were preserved for D-statistic analyses of possible introgression (see below).

Maximum likelihood tree inference and bootstrapping
Initial phylogenetic tree inference and bootstrap analyses were performed under ML with RAxML v.8 (Stamatakis, 2014). Analysis was conducted using the general timereversible model with rate variation (GTR+Γ; Rodriguez et al., 1990), and 200 fast non-parametric bootstraps to estimate branch support. To assess the possible role of introgressive hybridization in the clade, we used Patterson's D-statistic test (Durand et al., 2011) as implemented in ipyrad (Eaton and Overcast, 2020). Set-up details and full results are provided in Supplementary Data S1. Two primary hypotheses were tested: the hybrid origins of Quercus afares (Mir et al., 2009) and of Quercus crenata (Schwarz, 1936;Pignatti, 1982). In addition, we performed follow-up tests for admixture between Quercus cerris and Q. afares; among subsections Aegilops, Suber and Libani; and between Quercus ilex and Q. suber, representing increasingly large phylogenetic distances. Supplementary files, data matrices and analysis scripts are archived at https://github. com/andrew-hipp/cerris-fbd (v.1.0-1; https://doi.org/10.5281/ zenodo.7547523).

Fossils
We compiled a set of 47 fossils as age distribution priors for the fossilized birth-death model (see next subsection) and mapped the spatiotemporal distribution of oaks, with a focus on section Cerris. Of these fossil occurrences, 24 localities are dated by radiometric dating and/or by vertebrate fossils; for two localities, ages are constrained using palaeomagnetic data; and the remaining localities are dated by lithostratigraphic correlation or dinocyst stratigraphy (Supplementary data Tables  S2, S3). Four fossil taxa are represented by fruit/cup remains, seven by pollen, and 36 by leaves. Pollen taxa were assigned to sections based on synapomorphies shared with particular lineages (Denk and Grimm, 2009;Denk et al., 2017a), whereas fruit and leaf fossils were chosen based on (sub)sectiondiagnostic traits. The full list of selected fossil taxa, with their taxonomic assignments, information on the plant organ, the branch to which they are assigned, geographical origin, ages, constraints and relevant references (if not occurring in the main text) are provided in the Supplementary data (Table S2). See Supplementary Data S1 and S2 for notes on stratigraphic units and mapping of fossils mapped onto palaeoglobes (Scotese, 2013a(Scotese, , 2013b(Scotese, , 2013c(Scotese, , 2013d; early Eocene to Last Glacial Maximum).

Fossilized birth-death dating analyses
For FBD analyses, the RAD-seq matrix was reduced to 29 tips, with a single tip per species within both section Cerris and section Ilex, with the exception of Quercus cerris, for which two individuals were kept that did not group together in any analyses and might represent pseudo-cryptic or cryptic species. Loci were retained if they were present in at least ten individuals. A NEXUS file was exported using the RADami package, including 47 additional lines of undetermined positions (coded as '?'), one per fossil included in the FBD analyses. The FBD analyses were conducted in BEAST2 (Bouckaert et al., 2014).
Markov chain Monte Carlo (MCMC) runs of 50 million generations each were run from ten independent random starting points on each of three random draws from the uniform distribution of the fossil age ranges. Analyses were conducted using a nucleotide substitution model that allows for rate variation and invariant sites (Γ+I), with the shape parameter (α) and proportion of invariant positions estimated, and four gamma categories. The relaxed log-normal clock was used, with the clock rate estimated. Details of the analysis are in the Supplementary Methods (Supplementary Data S1). Scripts and RAD-seq data matrices are archived in the code repository for this paper (release v.1.0-1: https://github.com/andrew-hipp/ cerris-fbd; https://doi.org/10.5281/zenodo.7547523).

Köppen-Geiger climate types, WorldClim climate data and major forest biomes
We used grid-weighted 'Köppen signatures' (Denk et al., 2013;Grímsson et al., 2018;Bouchal et al., 2020), henceforth 'Köppen profiles', to summarize the climatic niches occupied by species of Cerris and to investigate climatic niche evolution within and among subsections of Cerris (Table 1; Supplementary Data S3). A Köppen profile reflects the proportional coverage of the various Köppen-Geiger climate zones (cf. Kottek et al., 2006;Peel et al., 2007) by a modern species based on gridded distribution data. To simplify interpretation, Köppen profiles are summarized into five climatic niches (see the subsection Maximum likelihood reconstructions of major climatic niches and main biomes); additional details on interpretation and coding are in the Supplementary Methods (Supplementary Data S1).
Modern species distributions were connected to fossil distributions by using georeferenced occurrence data for each species, downloaded from the GBIF database (www.gbif.org; Supplementary Data S3). Each dataset was checked for natural distribution outliers (e.g. specimens from botanical gardens). Published chorological data were used to detect these outliers (e.g. Browicz and Zieliński, 1982;Fang et al., 2009;San-Miguel-Ayanz et al., 2016;Caudullo et al., 2017). The cleaned georeferenced occurrence data were then plotted onto a 5 arc min Köppen-Geiger grid (1986Rubel et al., 2017) to establish Köppen profiles for all species of section Cerris; and on major terrestrial biome maps (Olson et al., 2001;Supplementary Data S3) to assess the forest biome preferences of species.

Characterization of modern leaf types
Leaf morphologies of modern section Cerris species were characterized using leaf texture, lamina size, tooth type and other traits (Table 1; Supplementary Data S1). The overall morphological differentiation patterns were visualized using a neighbour net (Bryant and Moulton, 2004), a planar (meta-) phylogenetic network (cf. Denk and Grimm, 2009). Twelve traits were scored as a categorical character matrix (11 binary and 1 ternary character; matrix LeafMorphs in Supplementary Data S1: Data File S1-1) and used to infer simple (Hamming) pairwise distances and to reconstruct character evolution on  (Maddison and Maddison, 2011; Supplementary Data S1).

Maximum likelihood reconstructions of major climatic niches and main biomes
Based on the quantitative assessment of biome and climate zone preferences of the modern-day species, we binned extant and fossil species into five basic categories, accounting either for biome or climate zone preferences (Table 1; Supplementary Data S3). Our generalization and categorization make use of the terminology and concepts introduced by Schroeder (1998;cf. Denk et al., 2013) and allow: (1) direct comparison of biome and climate zones preferences, which are commonly correlated but not synonymous; and (2) relation of quantitative modern-day categorization qualitatively to our fossil taxon set. Towards that end, we first defined the putative covered biomes for each fossil taxon of section Cerris (columns Biome/major Köppen climate type in Supplementary data Table S2). Assignment of fossils to climates is described in the Supplementary Methods (see also Supplementary data Table S2; Supplementary Data S1: Data File S1-1). Explicit connection of biomes to Köppen profiles is explained in the Supplemental Methods (Supplementary Data S1). In brief, the biomes and general climate preferences reconstructed are as follows:  (Maddison and Maddison, 2011). We used two different input trees: (1) the original dated tree for standard top-down reconstruction of ancestral states (i.e. using only the information scored for the modern-day species); and (2) the dated tree with nodes and tips added to account for states of fossil taxa. Fossil taxa that could be associated with a distinct branch (lineage) were treated as sister lineages and used to break down the according branch. We used the oldest possible age of the fossil taxon as the age of the putative most recent common ancestor (MRCA) and the youngest possible age to define the MRCAadded tip distance. The Mesquite-NEXUS file is included in the Github repository/Zenodo submission (Supplementary Data S1: Data File S1-1).

Phylogenetic inference
The RAD-seq locus dataset maintaining loci with a minimum of 15 individuals (m15 dataset; Fig. 1 . The m15 and m20 datasets recovered the same topologies, and the m15 provided the strongest mean bootstrap support. Consequently, we report on the m15 dataset topology here ( Fig. 1) and used that topology for our FBD constraints. The FBD dataset included 5075 loci and 444 591 aligned nucleotide positions, with 52.2 % missing data for the extant species (those with RAD-seq data). The RAD-seq ML phylogeny using the complete tip set ( Fig.  1) indicates an initial divergence between the East Asian (subsection Campylolepides A.Camus) and the Western Eurasian species of section Cerris (Fig. 2). Within the Western Eurasian clade, the earlier nuclear ribosomal 5S Intergenic Spacer (5S-IGS)-identified species groups (Cluster 1-4 in the study by Simeone et al., 2018) comprise four unambiguously supported clades, with two corresponding to subsections (Suber and Aegilops) and two unnamed (within what we are calling the Cerris core clade) ( Table 2). Subsection Suber (Spach) Maleev comprises the Western Mediterranean cork oak, Q. suber, and the southern French-Italian-Croatian Q. crenata, commonly considered a hybrid between Q. cerris × Q. suber. Diverging next is subsection Aegilops (Reichenb.) Menitsky, including Quercus macrolepis, which ranges from south-eastern Italy to southern Turkey, and Quercus ithaburensis occurring further east and south to Israel (Browicz and Zieliński, 1982;Jalas and Suominen, 1988). The third species, Quercus brantii, ranges from south-eastern Turkey and north-western Syria to the Persian Gulf. The most recently diverged section Cerris core clade collects members of subsections Libani (new) and Cerris (Dumort.) Guerke. The former includes the North African and East Mediterranean-Near East species Q. afares, Q. trojana and Q. libani; subsection Cerris includes the narrow endemics Q. castaneifolia (Hyrcanian forest region of Azerbaijan and northern Iran) and Quercus look (northern Israel to Syria), and the widespread, genetically and morphologically heterogeneous Western Eurasian Q. cerris. The D-statistic tests demonstrate inter-locus phylogenetic discordance in the relative placement of subsections Aegilops, Suber and Libani, but none affects the conclusions presented here. There is no evidence of specieslevel introgression involving Q. afares, Q. canariensis, Q. suber or Q. ilex (Supplemental Data S1). According to Simeone et al. (2018), Q. euboica, not included in the present study, is a distinct species with subsection Libani morphology but genetically closer to subsection Cerris as defined here. It would thus be part of the section Cerris core clade.

Dating and historical biogeography
For each of three random draws from the uniform age distribution for all fossils, six to nine of the ten independent MCMC runs converged (Supplementary data Table S4). Burn-in was assessed by visual inspection of model likelihoods and estimated age distributions of the constrained nodes, and independent post-burn-in MCMC runs were pooled for each random draw of ages. Given that confidence intervals overlap strongly for all node ages across all three random draws from the fossil age distributions (Supplementary data Table S5), we report only the first of the three age draws in the main text of this paper, but provide results of all three in the Supplementary data (Tables   S4, S5). The FBD dating with 47 fossils ( Fig. 3  Quercus section Cerris is the only oak section to reach its highest species richness and absolute phylogenetic diversity in Western Eurasia (12 spp., vs. 3 spp. in East Asia; Table 1). Each clade shows a broadly cohesive geographical distribution of parapatric to allopatric species (Fig. 2) that replace each other along a climatic cline (details in Supplementary Data S3).
The East Asian subsection Campylolepides covers a region from eastern Nepal to Japan and Laos. Its species can be categorized as Moist-Subtropical, Meridio-Nemoral or Nemoral. They thrive predominantly in warm, fully humid or winter-dry climates, occasionally extending into arid and cool climates. Their habitats are characterized by synchronous temperature and precipitation yearly minima (December-January). Only one species, the Nemoral Q. variabilis tolerates substantial frost during winter.
The Meridional to Full-Mediterranean subsection Suber of the Western Eurasian clade covers a region from Croatia and northern Italy south-and westwards to Tunisia, Algeria and Morocco, and to France, Spain and Portugal, specializing in Mediterranean climates but ranging into temperate climates. Here, Q. suber, the most west-extending species of the section, occurs predominantly in distinctly Mediterranean climates (hot summers with pronounced drought), whereas its eastern sister species, Q. crenata, thrives in sub-Mediterranean to fully humid climates (rare droughts  Fig. 3. First of three chronograms for Quercus section Cerris and its sister clade section Ilex inferred with the FBD approach (pruned to modern-day tip set). The phylogenetic, stratigraphic (time/time slice) and geographical position of the used fossil dataset is indicated (23 and 14 fossil species for section Cerris and section Ilex, respectively; see Supplementary data Table S2). See Supplementary data (Table S5) for details across all three runs. Median rates and 95 % highest posterior density intervals are depicted in the Supplementary data (Fig. S4). The widespread Q. cerris is ecologically variable (summerdry Mediterranean into cool, frost-and snow-prone climates). Its climatic niche covers most of the total niche of section Cerris. It is the only species in its section with a main distribution in areas with a fully humid temperate climate with warm summers, and the only (extant) species in the section producing lobed leaves. The Full-Mediterranean Q. look is a narrow endemic in the Levant (Middle East: Mount Hermon, Anti-Lebanon and Lebanon Mountains), in a distinct Mediterranean climate setting, most similar to that of Q. ithaburensis in subsection Aegilops. Quercus castaneifolia is another endemic species south of the Caspian Sea separated from the nearest Q. cerris population by >600 km. Like other species categorized as Full-Mediterranean, it prefers (sub-)Mediterranean climates, which commonly are transitional to fully humid or summer-dry warm and cool climates. Its general niche resembles that of Q. afares, with mild but drier winters and no extensive summer drought; the precipitation maxima are concentrated in autumn. In contrast to Full-Mediterranean species, Q. castaneifolia typically occurs in mesic Temperate Broadleaf and Mixed Forests (i.e. it can be categorized as Meridional). Quercus cerris (Meridional or Full-Mediterranean) forms part of Mediterranean Forests, Woodlands and Scrub and of deciduous Temperate Broadleaf and Mixed Forests and, at higher elevations, Temperate Conifer Forests.

Maximum likelihood reconstructions of ancestral climate zones and biomes
Using only extant species, the MRCA of Western Eurasian Cerris oaks is reconstructed as Full-Mediterranean/Meridional, given that most modern species thrive in Mediterranean Forests, Woodlands and Scrub biomes under a Mediterranean climate with hot and dry summers, including the oro-Mediterranean belt (perhumid). For the East Asian subsection Campylolepides (biome-wise nemoral, climate-wise variable), the result is similarly biased (MRCA biome: nemoral; climate: ambiguous; small signatures in Fig. 4; Supplementary data Fig. S4).
In contrast, ML reconstructions incorporating fossil taxa suggest late (Early to Middle Miocene) biome and climatic niche shifts in East Asian members of section Cerris from Moist-Subtropical to (Meridio-)Nemoral. A shift from warm fully humid (Subtropical) to winter-dry monsoon climates (Meridio-Nemoral) is reconstructed for the Late Miocene (11.6-5.33 Ma; Supplementary data Fig. S4). In Western Eurasia, probable climate shifts from fully humid climates to climates with temperature and/or precipitation seasonality are inferred for the Middle Miocene (16.0-11.6 Ma) for the Cerris core group (sections Cerris and Libani). This is in accordance with high-resolution palynological data from East Mediterranean strata, which suggest a transition from equable warm-humid temperate climates to more seasonal (precipitation) and cooler climates (Bouchal et al., 2020). Shifts to fully Mediterranean climates are reconstructed with high confidence for the Pleistocene (2.58-0.012 Ma). No shifts are reconstructed with high confidence for the Mediterranean Forests, Woodlands and Scrub biome before the Pliocene when fossil assemblages containing the fossil taxa are considered. The strong effect of including fossils in ancestral state reconstructions is illustrated in subsection Suber, where the fossil(s) cause a shift in inferred biome and climate from summer-wet to summer-dry conditions during the latest Miocene. Without information from the fossil record, no biome shift is reconstructed, while already by the Oligocene a preference for summer-dry climates (Meridional; Full-Mediterranean) is reconstructed within the Western Eurasian clade of section Cerris.

Leaf evolution and climatic niches in section Cerris
Section Cerris exhibits high leaf variability in response to temperature (mean temperature of the coolest month; Fig. 5). A potentially ancestral leaf type, with narrow elliptic lamina, triangular teeth or reduced teeth with long bristle-like extensions, is present in all members of the East Asian subsection Campylolepides (Fig. 5A-C) thriving in summer-wet climates across a wide temperature range (Supplementary Data S3). This leaf type is also found amongst the earliest known leaf fossils of the section (Quercus gracilis [Pavlyutkin] Pavlyutkin; Supplementary data Table S2) and has been retained in the Western Eurasian subsection Libani, part of the Cerris core clade (Fig. 5L). Within Campylolepides, a correlation between leaf size and petiole length and cold tolerance is seen in the increase in both from Q. chenii to Q. variabilis and Q. acutissima (Fig. 5A-C; Supplementary Data S3).
A second leaf type is represented by the fossil species Quercus kraskinensis Pavlyutkin, which co-occurs with the fossil species Q. gracilis in the early Oligocene site of Kraskino (Supplementary data Table S2). This leaf morphotype is potentially symplesiomorphic in section Cerris, because it shares features with Western Eurasian species that are absent from the modern East Asian oaks (Supplementary Data S1: Fig. S1-6): teeth are usually strongly developed, with mucronate to cuspidate apexes and convex to sigmoid basal and apical sides. Early-diverging species in all Western Eurasian clades, irrespective of their diverse niche preferences, possess such leaves (subsection Suber: Q. crenata, Fig. 5D; subsection Aegilops: Q. brantii, Fig. 5G; subsection Libani: Q. afares and Q. trojana, Fig. 5J, K; subsection Cerris: Q. castaneifolia, Fig. 5N; and in Q. euboica, Cerris core clade, Fig. 5M).
Subsection Suber exhibits a leaf morphological gradient from a crenata type to a suber type, expressed by a reduction in leaf size and tooth area and a change from (semi)deciduous to (semi)evergreen, more leathery leaves. This gradient is associated with a climatic gradient from mesic (Nemoral; Q. crenata) to (Full-)Mediterranean conditions (Q. suber; Supplementary Data S1 and S3). Small ovate leaves with reduced cuspidate teeth lacking soft bristle-like extensions are found only in the semi-evergreen western Mediterranean Q. suber (Fig.  5E) and superficially resemble evergreen leaves of Q. ilex in section Ilex. Subsection Libani (Fig. 5J-L) exhibits a similar decreasing gradient in leaf size and tooth area associated with a decreasing gradient in cold tolerance and aridity from Q. afares to Q. trojana to Q. libani. In contrast to the diversity found in subsections Suber and Libani, all species in subsection Aegilops (the only subsection in which all members are adapted to pronounced summer-drought as well as winter-cold climates; Supplementary Data S3) have medium-sized leaves with teeth that range from weakly developed to complex, coarse teeth with subsidiary teeth (Fig. 5H, I).
Finally, in subsection Cerris, the greatest tooth area and largest leaf size are seen in Q. cerris, the most Nemoral species in Western Eurasia, and Q. look, a Full-Mediterranean species occupying a climatic niche that overlaps with both subsections Aegilops and Libani. Lobed leaves displaying an enormous variability are found exclusively in the widespread Q. cerris (Fig. 5O), which also exhibits a broad climatic niche, surpassed in the section only by the East Asian Q. variabilis. The remarkable leaf polymorphism of Q. cerris includes leaf types seen

W. Eurasian Cerris
Oldest evidence for sect. Ilex in WEA  Fig. 4. Maximum likelihood mapping of preferred biomes on the chronogram of the first run (Fig. 3), scored as five categories: Moist-Subtropical, Meridio-Nemoral, Nemoral, Meridional and Full-Mediterranean. Large pie charts give proportional likelihoods for most recent common ancestors (MRCAs) of modernday species (at nodes) and (along branches) additional (shadow) MRCAs inferred by using fossil taxa (stars connected to tree) to break down subsequent branches. Coloured outlines of stars indicate the provenance of included fossil taxa and fossil taxa that could not be assigned clearly to a branch in the dated tree (unconnected stars). Small pie charts above big charts give the results when only modern-day states are considered. Numbers to the left of representative leaves refer to leaf (tooth) types as defined in Supplementary Data S1 (Fig. S1-7).
in Q. castaneifolia and Q. look (Fig. 5N, P; cf. fig. 7 in the study by Denk et al., 2021a). The ecological-climatic and leafmorphological variation of Q. cerris parallels the lack of genetic coherence in our RAD-seq dataset, with some Q. cerris sharing the genotype of Q. castaneifolia, whereas others exhibit a high degree of genetic similarity to Q. look (hence, represented by two tips in the subset used for dating). This might indicate ongoing speciation in subsection Cerris.

Evolutionary and biogeographical history of section Cerris
Section Cerris appears to have originated and diversified morphologically in northern East Asia by the early Oligocene. The oldest fossils of section Cerris are dispersed pollen grains from early Eocene (Ypresian, 56.0-47.8 Ma) strata of the Russian Far East (Shkotovskii Basin; Naryshkina and Evstigneeva, 2020;Pavlyutkin et al., 2020). By the early Oligocene, section Cerris was present with at least two distinct fossil species based on leaves in the Russian Far East Kraskino Flora (34-30 Ma; Pavlyutkin et al., 2014;Pavlyutkin, 2015). Importantly, these oldest East Asian fossil records pre-date the earliest known fossils of Cerris in Western Eurasia (dispersed pollen from Germany, Altmittweida; earliest Miocene, 23-20.5 Ma; Standke et al., 2010;Kmenta, 2011) by >10 million years, and unambiguous leaf records of section Cerris are not known in Western Eurasia before the Miocene (e.g. Knobloch and Kvaček, 1976;Mai, 1995). Foliage described as 'Q. gracilis' (Pavlyutkin) Pavlyutkin (nom. illegit.) is very similar to modern leaves of East Asian members of section Cerris (Fig. 5). Another species described from the Kraskino Flora, Q. kraskinensis Pavlyutkin (Pavlyutkin, 2015), is strikingly similar to a number of modern Western Eurasian Cerris oaks, in particular to Q. crenata, the root-proximal species in the second-diverging subsection Suber, and to a lesser degree to Q. trojana and Q. cerris, members of Cerris core clade (present study; Supplementary Data S1: Plate S1-1). Thus, this leaf morphology is characterized by shared features, ancestral (possibly symplesiomorphic) within section Cerris. Quercus kraskinensis might represent a precursor or early member of what would become the Western Eurasian clade of Cerris (Fig. 3). The East Asian Palaeogene record thus demonstrates that most of the range of Cerris leaf morphological diversity evolved at the dawn of the section, with the 'kraskinensiscrenata' leaf type originating in East Asia and surviving in Western Eurasia (Supplementary Data S1). The northern East Asian origin of section Cerris (cf. Fig. 3) is also supported by evidence from previous molecular studies. First, all modern Cerris share the same, section-unique plastid lineage, indicative of a single point of origin and quick dispersal, with the East Asian subsection Campylolepides showing the overall highest plastid divergence (Simeone et al., 2018;Zhang et al., 2020;Li et al., 2022). Second, the Cerris plastomes are part of a haplotype lineage shared with a group of section Ilex species thriving in modern-day Japan and the mountains of northern and central China (East Asian clade in Figs 1 and 3; Simeone et al., 2016: Quercus engleriana Seemen, Q. phillyreoides and Q. spinosa David; Zhou et al., 2022: Quercus dolicholepis, Q. engleriana, Q. spinosa and Q. pseudosetulosa Q.S.Li & T.Y.Tu), i.e. geo-historically close to the oldest Cerris fossils of the Russian Far East (Fig. 6). Most of these species belong to Jiang et al.'s (2019) early-diverging 'clade II' (before the mid-Eocene, ≥ ~40 Ma, in Jiang et al., 2019; median divergence stem age of 49 Ma, present study, Fig. 3). In combination, the fossil and molecular data work together to provide a robust picture of the East Asian origins for the iconic European section Cerris.
After the earliest appearance of Cerris in Western Eurasia in Early Miocene deposits of Germany, dispersed pollen grains of section Cerris are also known from slightly younger Burdigalian strata of Turkey and Greece (Denk et al., 2017b). Given the complex tectonic situation and the availability of different more or less temperate niches during the Early and Middle Miocene, it is conceivable that Cerris evolved several lineages within its first 10 million years (subsects Suber, Aegilops and Cerris core clade; Fig. 3). One outcome of a rapid diversification might be partial reproductive compatibility between these lineages, as evidenced by admixture among subsections (Supplementary Data S1). A rapid origin and spread with gene flow between lineages is also supported by the low plastid differentiation, decoupled from main intrasectional lineages (subsections) and species. Cerris oaks then became relatively widespread (Quercus kubinyii) in Western Eurasia during the Middle Miocene, ranging from Denmark to Anatolia (Kováts, 1856;Christensen, 1976;Knobloch, 1986;Güner et al., 2017;Denk and Bouchal, 2021). These leaf remains resemble modern species of section Cerris that possess ancestral leaf types (Fig. 5;leaf types '0' and '1'). Section Cerris thus appears to have colonized and diversified in Europe and the Mediterranean from the Early to mid-Miocene, providing the ancestry for the modern-day species and subsections. Our combination of phylogenomic data with a rich set of fossils suggests that main lineages in Western Eurasia might have been established before their modern morphologies (Figs 3 and 5; Supplementary Data S1).
The Cerris history becomes all the more interesting in light of its parallels with the section Ilex history. Fossil pollen of both Quercus sections Cerris and Ilex in early Eocene strata of the Russian Far East (Naryshkina and Evstigneeva, 2020) indicate that both lineages might have originated in high-latitude warm temperate biomes (cf. Scotese et al., 2014). But while section Cerris had a first radiation in Northeast Asia and subsequently migrated to Western Eurasia north of the progressively enlarging Qinghai-Tibet Plateau, its sister clade, the evergreen section Ilex, initially migrated southwards and south-westwards into tropical China and south-eastern Tibet (Linnemann et al., 2017;Su et al., 2019;Hofmann, 2010), thence westwards into Europe and the Mediterranean along the proto-Himalayas south of the Qinghai-Tibet Plateau (Jiang et al., 2019). Section Cerris makes it into southern China only in Late Miocene strata of western Yunnan (Xia et al., 2009;Xu et al., 2012). The co-occurrence of the sections in East Asia is further supported by shared plastids of northern East Asian species of section Ilex (mostly Jiang et al., 2019, clade II) with section Cerris (Simeone et al., 2016;Yan et al., 2019;Zhou et al., 2022). Moreover, the Cerris-Ilex shared plastome lineages differ substantially from plastome lineages shared between section Ilex and the East Asian section Cyclobalanopsis (Yang et al., 2018), further pointing to the divergent history of these sections in Southeast Asia. Thus, molecular data corroborate that Western Eurasian Cerris came into contact with Ilex only after the East Asian Eocene and Oligocene members of Cerris had begun to move westwards. Plastid phylogeography thus fits with a scenario of largely isolated early evolutionary histories of Cerris (northern Asia) and Ilex (Himalayan Corridor).
Ecological and climatic niche evolution Our reconstructions of biome and climatic niche suggest persistence of the Tropical and Subtropical Moist Broadleaf Forests biome during the Oligocene, consistent with fossil records (e.g. Kvaček, 2010;Pavlyutkin et al., 2014). In contrast, reconstructions that ignore the fossils (Fig. 4) suggest shifts into the Mediterranean Forests, Woodlands and Scrub biome that are at odds with the Western Eurasian fossil record, where this biome is not recorded before the Plio-Pleistocene (Suc, 1984;Velitzelos et al., 2014). The early Eocene (Ypresian, Ma) split between sections Cerris and Ilex must have coincided with the origins of deciduousness in Cerris, whereas Ilex retained the original evergreen leaf habit of subgenus Cerris (which is also characteristic of section Cyclobalanopsis, sister to the Cerris + Ilex clade). The subsequent biogeographical histories of the two lineages reflect this change. Members of section Cerris moved westwards as a northern lineage, part of a more or less temperate forest biome dominated by deciduous  (Fig. 3). The map was produced with the software QGIS (QGIS, 2021). tree species during the Oligocene (Kryshtofovich et al., 1956;Popova et al., 2013;Pavlyutkin et al., 2014;Scotese et al., 2014;Willis and McElwain, 2014;Averyanova et al., 2021;Denk et al., 2021b). Extensive lowlands in large parts of Siberia and northern Kazakhstan provided nutrient-rich substrates dominated by deciduous woody plants ('warm temperate biome' in the paper by Willis and McElwain, 2014;'warm [temperate]' climate zone in the paper by Scotese et al., 2014). To the south of this more or less temperate forest belt, a drier region was occupied by the 'subtropical summerwet biome' (coined wooded savannah to semi-desert; Willis and McElwain, 2014) corresponding to an arid climate zone with frosts according to Scotese et al. (2014). Here, section Cerris oaks would have had an advantage owing to the earlier origins of deciduous leaf phenology.
This trait diversity is likely also to have shaped the ecological diversification of section Cerris and the high community-level diversity of Mediterranean oaks. In Western Eurasia, Cerris oaks colonized a wide range of habitats, reflected in their Miocene distribution from southern Scandinavia to southern Turkey, a range of almost 20° latitude (Fig. 6). Three key morphological features of Cerris oaks might have provided advantages during subsequent shifts into their modern Mediterranean habitats: deciduous or semi-deciduous leaves, large acorns protected by sturdy cups, and corky stems. In contrast, the evergreen oaks of section Ilex that spread southwards and south-westwards during the Eocene from the temperate to the subtropical summer-wet biome differentiated geographically (Jiang et al., 2019) but remained evergreen. The resulting climatic niche partitioning in the Western Eurasian subsections of Cerris oaks (Supplementary Data S3) and between sections Ilex and Cerris might have enabled the coexistence of several summer-and wintergreen oak lineages: ≤13 species of both deciduous and evergreen oaks from three sections (Quercus, Cerris and Ilex) presently co-occur in southern Turkey, for example (Hedge and Yaltirik, 1982;Blumler, 2015).
Ecological lability of section Cerris also manifests in both niche convergence and trait convergence. Quercus suber, for example, is the only European tree species that is able to resprout after fire damage (Pausas, 1997;Houston Durrant et al., 2016). It is also highly flexible in the timing and duration of leaf abscission and can rapidly replace old leaves with new shoots (Escudero and del Arco, 1987), characteristics typical of all members of section Cerris. Yet a number of Q. suber adaptations are more reminiscent of evergreen species of section Ilex: its ability to switch between annual and biennial fruit maturation in response to climate, semi ringporous wood anatomy that reduces the risk of embolism in droughty springs, and sandy lowland habitat (Elena-Rosello et al., 1993;Sousa et al., 2009). In contrast, Q. crenata, the sister species of Q. suber, retains many putatively ancestral traits of section Cerris (corky bark, leaf texture, leaf abscission, leaf shape, ring-porosity and partly humid temperate climatic niche) and is morphologically similar to some of the oldest known fossil species of Cerris (e.g. Q. kraskinensis), perhaps owing to niche conservatism and repeated phases of introgression (Table 3). Quercus cerris, the most temperate of the Cerris oaks, provides an even more pronounced example of convergence between sections. It is the only oak in all of subgenus Cerris with complex lobed leaves, reminiscent of the lobed white oaks of Quercus section Quercus (subgenus Quercus). Lower hydraulic resistance in deeply lobed leaves might provide a mechanism for improving water balance in dry atmospheric conditions (Siso et al., 2001) and also enable leaves to pack more efficiently into buds (Edwards et al., 2016;Givnish and Kriebel, 2017;Zohner et al., 2019). Lobedness is thus a convergent trait that might contribute to the ability of Q. cerris to grow sympatrically with (co-)dominant lobed white oaks, such as Quercus robur L., Q. petraea, Q. frainetto Ten. and Q. vulcanica Boiss. ex Kotschy, and account for its long history of cultivation in the British Isles (Loudon, 1838). Section Cerris thus contributes to our growing understanding of the importance of both divergence within sections and convergence between them in shaping oak diversity and patterns of coexistence (cf. Cavender-Bares et al., 2004.

Conclusions
Two different migration routes resulted in distinct diversity patterns in the sister sections Cerris and Ilex. The deciduous section Cerris is most ecologically and taxonomically diverse in Western Eurasia (12 of 15 species), whereas the evergreen section Ilex is most diverse in East and Southeast Asia into the Himalayas (21 of ~25 species). Globally, evergreen broadleaf species occur in relatively humid (~1500-3000 mm mean annual precipitation) and warm (mean temperature of the coldest month > 0 °C) climates, whereas winter deciduous broadleaf species typically occur in relatively less humid (mean annual precipitation ~700-1500 mm) and cooler climates (mean temperature of the coldest month < 0 °C; Woodward et al., 2004). Nevertheless, both evergreen sclerophyllous and deciduous oak species are, at present, highly diverse in summerdry Mediterranean areas. Extensive research in modern Mediterranean ecosystems suggests that neither evergreen nor deciduous species grow optimally there (Escudero et al., 2017).
Our time-calibrated phylogenetic reconstruction using 47 fossils suggests that suboptimal adaptation to current Mediterranean climate might be a deep-time evolutionary legacy in both the evergreen section Ilex and the deciduous section Cerris, resulting from their differential early range expansions from Northeast Asia. Section Cerris shifted to deciduous leaves in frost-free environments, which would have preadapted the lineage to the dry and cold climates it encountered in its westward expansion (trait first pattern). Section Ilex retained its evergreen leaf phenology and did not shift to deciduous leaves when colonizing winter-cold and -dry habitats in the Himalayas and warm, summer-dry environments in the Mediterranean region. Western Eurasia thus became a meeting ground for old relatives of divergent sources: the Northeast Asian Cerris, the Southeast Asian Ilex and, ultimately, the eastern North American white oaks of section Quercus, which joined them between 10 and 20 Ma (Denk and Grimm, 2010;. These legacies explain why species of all three sections co-occur in contemporary Mediterranean climates of Western Eurasia and how their distributions follow environmental and climatic gradients within the wider Mediterranean region.

SUPPLEMENTARY DATA
Supplementary data are available online at https://academic. oup.com/aob and consist of the following.
Supplementary Figure S1. Raw maximum Likelihood tree inferred from the m20 RAD-Seq data (3145 loci, 277,006 aligned nucleotide positions, and 58.2% missing data), rooted with Notholithocarpus. Numbers at branches indicate non-parametric bootstrap support.
Supplementary Figure S2. Raw maximum Likelihood tree inferred from the m25 RAD-Seq data (1132 loci, 100,841 aligned nucleotide positions, and 46.4% missing data), rooted with Notholithocarpus. Numbers at branches indicate non-parametric bootstrap support.
Supplementary Figure S3. Same dated phylogenetic tree as shown in main-text Figure 3 (first of three runs). Top: chronogram with fossil-species included (unlabelled tips) and branches coloured according to the estimated median substitution rates. Bottom: chronogram showing the 95% HPD intervals for inferred species divergences of extant species.
Supplementary Figure S4. Maximum likelihood mapping of preferred Köppen climates on the chronogram of the first run (main-text Fig. 3); scored as five categories: Moist-Subtropical, Meridio-Nemoral, Nemoral, Meridional, Full-Mediterranean. Large pie charts give proportional likelihoods for most-recent common ancestors (MRCAs) of modern-day species (at nodes) and (along branches) additional (shadow) MRCAs inferred by using fossil-taxa (stars connected to tree) to break down subsequent branches. Coloured outlines of stars indicate the provenance of included fossil-taxa and fossil-taxa that could not be clearly assigned to a branch in the dated tree (unconnected stars). Small pie charts above big charts give the results when only modern-day states are considered.
Supplementary Data S1. Extended methods and results. Supplementary Data S2. GoogleEarth locality (kmz) files indicating the probable position of all here-used fossils within their palaeogeographic context (palaeogeographic globes from Scotese, 2013aScotese, , 2013bScotese, , 2013cScotese, , 2013d Supplementary Data S3. Köppen distribution and biome maps for all species of section Cerris, as well as x-y and boxplots for selected bioclimatic parametres.