Ancient Antarctica: the early evolutionary history of Nothofagus

ABSTRACT The genus Nothofagus (southern beech) has an extensive fossil record and extant species exclusively distributed in the Southern Hemisphere. It is divided into four subgenera widespread across eastern Australasia and southern South America. The origin and evolution among closely related species remain an important question in palaeontology. The goal of this work is to reconstruct the biogeography of Nothofagus incorporating a complete leaf fossil dataset to better understand its origin, diversification, and colonisation history. The most ancient fossil leaves were discovered in Antarctica and are herein included for the first time into phylogenetic and biogeographic analyses. We employed statistical biogeographic methods implemented in BioGeoBEARS to estimate ancestral areas. The results support a high probability that the ancient ancestor of Nothofagus may have originated in Antarctica during the Late Cretaceous which is also supported by the fossil pollen record found in the Antarctic Peninsula. Subgenera Fuscospora and Lophozonia resulted in the most ancient clades, while the subgenera Nothofagus and Brassospora evolved later (Palaeogene). Our model supports that subgenera divergences were characterised by both dispersal and vicariance events from the Late Cretaceous to the early–middle Eocene. Graphical abstract


Introduction
The past reconstruction of life on Earth is one of the most enigmatic issues to be resolved.New findings on the distribution of fossils play an important role in the reconstruction of historical biogeographic events as well as provide evidence for the evolution of life over millions of years.Nothofagus Blume (southern beech), the sole extant genus of family Nothofagaceae, is represented by four extant subgenera: Lophozonia, Fuscospora, Brassospora, and Nothofagus.It is a key taxon with an extensive fossil record and numerous extant species widely distributed in southern South America and eastern Australasia (Hill 2001;Hill and Read 1991;Swenson et al. 2000Swenson et al. , 2001a, b;, b;Vento and Agraín 2018).Nothofagus is a key genus in order to understand southern forest biogeography due its past and present diversity of species restricted to the Southern Hemisphere (Romero 1986;Hill 1992).The origin, nature, and timing of the evolution among closely related species of this genus remains an important question in palaeontology and evolution subjects (Hill and Jordan 1993;Dutra and Batten 2000;Swenson et al. 2000).The South Pacific disjunction has intrigued biologists for over a century and the Nothofagus biogeographic history was an enigma for several decades becoming perhaps the best-known example for understanding evolution (Heads 2006).
Every discovered fossil represents a snapshot giving us information about the evolution processes.Therefore, reliably identified fossils in accurately dated sediments play a crucial role in providing minimum age constraints on the divergence of lineages and evidence for their palaeogeographic distribution (Cantrill and Poole 2012;Sauquet et al. 2012).Many attempts have been made to reconstruct the inter-relationships of Nothofagus species and the relationships among subgenera, its biogeography and taxonomy, dispersion capacity, and the origin by a variety of methods using both morphological and molecular data of extant species (Tanai 1986;Hill 1991;Hill and Jordan 1993;Manos 1997;Jordan and Hill 1999;Swenson et al. 2000;Knapp et al. 2005;Premoli et al. 2011;Sauquet et al. 2012;Fernández et al. 2016).However, the inclusion of fossils of this genus in both phylogenetic and biogeographic analyses is limited.The reconstruction of ancestral areas of Nothofagus is an important factor for understanding the biogeographic diversification history of its lineages (Hill and Jordan 1993;Swenson et al. 2000).The fossil record should be considered complementary to the extant species and help overcome the problems that limited extant distribution and diversity may cause (Hill 2001).It has been shown that throughout the Cretaceous and even much in the Tertiary, Nothofagus has an extremely richer fossil record than today especially considering the palynological record (Couper 1960;Dettmann et al. 1990;Swenson et al. 2000).Fossil leaves of Nothofagus have been found in the Southern Hemisphere (Dutra and Batten 2000;Hill 1983aHill , 1983bHill , 1991;;HaoMin and ZheKun 2007;Pujana et al. 2021;Vento et al. 2017;Vento and Prámparo 2018, and cities therein).The first findings of Nothofagus leaves from Antarctica were recorded by Dusén (1899Dusén ( , 1908) ) and new macro and microfossils were discovered during the last decades (Askin 1990; Barreda et al. 2019;Dettmann et al. 1990;Dutra and Batten 2000;Dutra 2001Dutra , 2004;;HaoMin and ZheKun 2007;Kellner et al. 2007;Scasso et al. 2020;Tosolini et al. 2021;Zastawniak 1981Zastawniak , 1994;;Zastawniak et al. 1985 among others).
Nowadays, the genus Nothofagus, mainly characterised by the presence of simple or compound teeth (with an exception of a few taxa with smooth margin) and is a typical element of the relictual Southern Hemisphere temperate floras that has radiated out from its putative 'center of origin'.The centre of origin and diversification of Nothofagus has been suggested to be a region covering southern South America, western Antarctica, southeastern Australia and New Zealand, when all these land masses were connected (Cantrill and Poole 2012).Pioneer works about the biogeography of Nothofagus postulate the almost irrefutable theory of a southern origin (Couper 1960;Swenson et al. 2001aSwenson et al. , 2001b;;van Steenis 1971).The ancestral area of this genus is thought to be a region between southern South America and the Antarctic Peninsula.This area became the most likely centre of origin in the initial stages of differentiation and diversification (Cantrill and Poole 2012;Leppe et al. 2021;Swenson et al. 2000).The most ancient Nothofagus fossils (leaves and pollen) date back to the Late Cretaceous of western Antarctica (Dettmann et al. 1990;Dutra and Batten 2000;Leppe et al. 2016;Barreda et al. 2019;Romero et al. 2019) and provide an excellent evidence for understanding the extant disjunction in the Southern Hemisphere floras.Therefore, the available fossil evidence reflects that the origin can be traced to the Late Cretaceous in the Antarctic Peninsula (Leppe et al.;2016;Barreda et al. 2019;Romero et al. 2019;Pujana et al. 2021).Nevertheless, the biogeographic history of Nothofagus is still controversial and some points were not resolved yet due to the lack of the addition of fossils into the analysis.Until now there is not any biogeographic analysis including fossil leaves of Nothofagus from Antarctica.Only a few works have attempted to incorporate fossil taxa of this genus from South America and Australasia into a phylogenetic framework (Christophel 1985;Romero 1986;Hill and Read 1991;Carpenter et al. 2014;Vento and Agraín 2018).The fossil plants of Antarctica are fundamental to better understand the evolution of vegetation in the Southern Hemisphere (Zastawniak 1981(Zastawniak , 1994;;Cantrill and Poole 2012).
The goal of this work is to analyse the biogeography of Nothofagus in the Southern Hemisphere including morphological features of fossil and extant representatives from Antarctica, southern South America, and Australasia.We attempt i) to place fossil leaves within a phylogenetic framework, ii) to calibrate the phylogeny using a minimum age dating protocol, and iii) run a modelbased analysis that reconstructs ancestral areas, vicariance, and patterns of movement through time.We focus our research on the origin and the dispersion routes of Nothofagus across the Gondwana along the time and expect that the inclusion of fossils from some crucial locations like mainland Antarctica into the biogeographic analysis will provide relevant information on its history and improve dating resolution.
We revised the literature that included published fossil leaf taxonomic descriptions from Antarctica and used the most complete and better-preserved specimens for our analysis (Zastawniak 1981(Zastawniak , 1994;;Zastawniak et al. 1985;Dutra and Batten 2000;Dutra 2001Dutra , 2004;;HaoMin and ZheKun 2007;Romero et al. 2019).We also included fossils from Australasia in order to improve the phylogenetic resolution among different members of Nothofagus.There are many fossil taxa from Australasia in the literature; however, some of them are not clear enough to observe the diagnostic characters necessary for our analysis.Therefore, we focused on those ones that presented clear anatomical and morphological descriptions in order to minimise the amount of missing data.We coded these taxa following the taxonomic descriptions and the score made by Jordan and Hill (1999) and Carpenter et al. (2014).We included fossils of subgenera Brassospora, Nothofagus, Fuscospora, and Lophozonia from Australasia previously described (Hill 1984(Hill , 1988(Hill , 1991;;Jordan 1999;Paull and Hill 2003;Carpenter et al. 2014).
The selection of extant species from South America and Australasia was based on the previous work made by Heenan and Smiseen (2013) and Vento and Agraín (2018).We also examined herbarium sheets held at L. H. Bailey Hortorium (BH), Ithaca, New York, Herbario Ruíz Leal (MERL), Mendoza, Argentina, Museo Argentino de Ciencias Naturales (BA), Buenos Aires, Argentina, and the herbarium catalogue of the Royal Botanical Gardens available online at www.kew.org.Extant hybrids of Nothofagus were not considered in our analysis.

Morphological evidence
Morphological characters 0-35, including fruit, flower, anatomical leaf features, and pollen features of the extant species were taken directly from Heenan and Smissen (2013); characters based on leaf morphology (36-45) were coded for both fossil and extant species.The complete character list with their respective states (Table S1) is modified from Vento and Agraín (2018).The fossil characters scored in this contribution were mainly based on the leaf margin and the venation pattern based on the descriptions made by Tanai (1986) and Gandolfo and Romero (1992).

Phylogenetic analysis
The phylogenetic analysis included both fossil and extant taxa of the genus Nothofagus as follows: from Antarctic Peninsula (only fossils, because there are not extant species), Southern South America (Argentina and Chile), Australasia (southern Australia, New Caledonia, New Guinea, New Zealand, and Tasmania).As Nothofagus species were previously included as members of the Fagaceae (Hill and Jordan 1993;Manos 1997;Jordan and Hill 1999) but were considered later to be more closely related to Betulaceae (Nixon 1982(Nixon , 1989;;Jones 1986;Dutra 1997;Premoli et al. 2011), we selected two outgroups: Fagus Linnaeus (Fagaceae) represented by Fagus grandifolia Ehrhart y Fagus sylvatica Linnaeus and Betulaceae represented by Betula pendula Roth.The data matrix comprises a total of 46 morphological characters (Table S2).Morphological and anatomical characters were informative and treated as non-additive.The phylogenetic analysis was performed using TNT v.1.5parsimony software (Goloboff et al. 2008).We used implied weighting analyses exploring the topologies of the strict consensus trees resulting from a range of concavity constants (k) (Goloboff 1993).The topologies from k = 1 to k = 30 were explored both using implied weighting (Piwe) and the extended implied weighting analysis approach (XPiwe).For xpiwe we used the command 'xpiwe (*' to avoid that the missing entries generate too much homoplasy of the observed characters during the optimisation of the most parsimonious trees MPTs (i.e. to receive a high fit).This command is comparable to use different values of k for each character according to its percentage of missing entries.Missing entries are assumed to have 50% of the homoplasy of observed entries as part of the extended implied weighting functions of TNT, that allow to assign lower values of k to those characters with more missing data, (i.e.down-weighting homoplasy more strongly), see Goloboff (2014) for details.For both analyses (piwe and xpiwe) we used a traditional heuristic search that was performed on the base of Wagner trees with 3000 random addition sequences, followed by the tree bisection reconnection (TBR) swapping algorithm, saving 10 trees per replicate, and collapsing trees after the search.It was followed by a branch and bound search based on these trees from RAM.The evaluation of branch support was performed using symmetric resampling (Goloboff et al. 2003) with a change probability set up at 0.33 and 500 replicates (values are indicated as a frequency difference).
According to Goloboff (1993) there is no optimal criterion to choose any particular value of k.Recently, Goloboff et al. (2018) stated that better results can be obtained when weighting more gently against homoplasy (i.e. using larger values of k).This notion seems to be supported by our data when considering the congruence of the results obtained with k values equal and higher than nine and the current knowledge of Nothofagus phylogeny.Thus, we discuss the results on the base of the most parsimonious tree (fit = 6.16437) obtained with xpiwe k = 9 (Figure S1a) following the same criterion as in our previous contribution (Vento and Agraín 2018) that combines a satisfactory balance between robustness and resolution which is also coherent with the current morphological and molecular evidence for the genus Nothofagus.Symmetric resampling support measures are provided in Figure S1a.

Time-calibrated estimation
We estimated divergence times using both fossil and extant taxa in R (R CoreTeam 2019) package Strap (Stratigraphic Tree Analysis for Palaeontology) designed by Bell and Lloyd (2015).The tree calibration uses the first (FAD) and the last appearance datum (LAD) trusted record (Table S3).The formatting matches that used in the R package paleotree (Bapst 2012) and allows for easy swapping of time-scaled trees between packages (Bell and Lloyd 2015).The minimum and maximum possible age should be considered in a time-calibrated tree using geochronological information of fossil taxa (Pol and Norell 2006).Therefore, to extract the temporal information from the extinct taxa, the age of each fossil was entered as non-contemporaneous date representing millions of years (Ma) before the present.This age is determined by biostratigraphy or radiometric isotopic methods in the sequence in which the fossils were derived.The geological age of each fossil (FAD and LAD) was obtained directly from the literature (Table S3).In order to plot the time-calibrated tree against geologic time, the function 'geoscalePhylo' was used.The geologic time follows Gradstein et al. (2012) or the published time scales by the International Commission on Stratigraphy (Bell and Lloyd 2015).

Historical biogeography analysis
We performed a biogeographical analysis to estimate ancestral ranges and examine dispersal patterns employing the ready-to-use version of Reconstruction Ancestral States in Phylogenies (RASP 4.2) (Yu et al. 2020).We reconstructed ancestral areas for the main nodes within our tree topology to examine the geography of speciation and evolutionary history of Nothofagus.The geographic distributions were delimited into nine areas: Antarctic Peninsula, New Caledonia, New Guinea, New Zealand, Southern Australia, Tasmania, Southern South America, North Hemisphere, and Eurasia.The inference analysis of the best-fit test model for biogeographic history of Nothofagus was performed running the BioGeoBEARS model testing R package (Matzke 2013) (Landis et al. 2013).To account for the possibility of jump dispersal, an additional J parameter was added for all models.The latter parameter enables descendant lineages to occupy a different area than its direct ancestor (Matzke 2013).Therefore, we tested the three biogeographic models with and without the parameter J totalising six models.The ancestral area analyses were conducted using the posterior distributions of the time-calibrated phylogenetic tree that was previously estimated from the Strap package.As BioGeoBEARS requires fully binary trees, in this case, we used the tree obtained with piwe k9 but without collapsing the zero length branches (Figure S1a-b).We performed two biogeographic analyses using the known distribution area of B. pendula.These are: (A) using 3 as the maximum range of occupancy (MRO) and (B) fixing 9 as MRO.For both cases, we selected the model with the highest Akaike information Criterion (AICc_wt) to select the fittest model for our dataset, allowing all area combinations with equal probability.For the assessment of the effect of parameter J, we used a likelihood ratio test for nested models considering a p value <0.01 as statistically significant.
We preferred to place no constraints on range or movement ability, and considered that parameter J is worth to be included in our statistical model comparison.Recently, Matzke (2022) has shown that the Dispersal-Extinction-Cladogenesis model + J parameter (DEC+J) fits better on most datasets, demonstrating its statistical validity.All results from A-B analyses are shown in Table S3-S4.

Historical biogeographic analysis
Results for the two analyses (i.e.A-B using different MRO and distribution of the ancestor) indicate that the best statistical fit to the data is DEC+J.As the results were similar, here we describe the 'A' scenario, using the known area distribution of B. pendula and fixing 3 as MRO.Complete results of analyses can be found on tables S4-S5.Therefore, for the 'A' scenario, and according to the selection criterion, the best-fit model to our data is DEC+J with the highest AICc_wt = 0.93.The latter model identifies transitions between ranges in evolutionary time on a phylogenetic tree, then it uses likelihood methods to estimate ancestral ranges at nodes (Ree and Smith 2008).

Early evolution of Nothofagus
According to our results, the ancestral area (Figures 1 and 2) of the genus Nothofagus is suggested to be the Antarctic Peninsula with a probability of 56.9% (optimal area reconstruction at node 85).Therefore, the ancestral reconstruction suggests that the common ancestor of Nothofagus probably originated approximately between 80 and 90 Ma in the Late Cretaceous in Antarctic Peninsula (Figures 1 and 3a).Dispersal and vicariance events occurred in the Antarctic Peninsula and may have influenced the distribution of Nothofagus which could have expanded to other areas such as New Zealand, southern Australia, Tasmania, and southern South America (Figure 3b).The distribution continued later towards tropical regions such as New Caledonia and New Guinea with geographical diversification events of dispersal and vicariance in the Palaeocene (~60-65 Ma) and late Eocene-Oligocene (~45-30 Ma).This distribution pattern is similar to nowadays species distribution (Figure 3c).According to our biogeographic model, cladogenetic events in Nothofagaceae were characterised by jump dispersal that may occur in Antarctic Peninsula within the same geographic region (narrow sympatry) during Late Cretaceous and migration occurred later to other closer areas such as South America, New Zealand, and Australia.

Major cladogenetic events
Our phylogenetic tree supports the monophyly of family Nothofagaceae and shows clear relationships between fossil and extant representatives (Figure 1).The results show that most of the major cladogenetic events for Nothofagus, subgenera divergences, are characterised by a combination of dispersal and vicariance events  area from a common ancestor of these taxa with a probability of 91%.The ancestral area of subgenera Fuscospora, Brassospora, and Nothofagus (node 71) is inferred to be the Antarctic Peninsula (51.5%) and southern South America (13.5%).Moreover, the ancestral area of Fuscospora (node 70) is shown as Antarctic Peninsula with a probability of 62.2%.The sister clades corresponding to subgenera Nothofagus and Brassospora (node 63) might have shared their ancestral area in southern South America with a high probability of 40.8% and New Guinea with a probability of 32.7%.For the subgenus Brassospora (node 62), New Guinea is the most probable ancestral area with 78.4%, and for the subgenus Nothofagus (node 53), southern South America is the most probable ancestral area with 91.4%.At this node (53) both dispersion and vicariance events are hypothesised to have separated the species Nothofagus nitida (Philippi) Krasser (southern South America) from Nothofagus lobata Hill (Tasmania).Thus, the common ancestor (node 52) of both taxa probably evolved in southern South America (79.5%).

New biogeographic insights from Antarctica
A region covering South America, Antarctic Peninsula, Australia, and New Zealand has been suggested as a possible centre of origin and dispersion of many plant species during millions of years ago, when the mentioned land masses were connected (Cantrill and Poole 2012).The unique connection between South America and western Australasia in the supercontinent Gondwana was via the Antarctic Peninsula.This region has been postulated as the link between the mentioned land masses with effect on the angiosperm radiation including Nothofagus (Cantrill and Poole 2002).It is thought that the Antarctic Peninsula and southern South America were linked by a land bridge before the opening of the Drake Passage up until 28 Ma (Dalziel 2014;Eagles et al. 2014).The Late Cretaceous paleofloras of the southern Patagonian basins and those of the Antarctic Peninsula are qualitatively similar suggesting that the peninsular area might have been a centre of origin and diversification of this genus (Barreda et al. 2019;Cook and Crisp 2005;Dutra and Batten 2000;Premoli et al. 2011;Romero et al. 2019;Scasso et al. 2020).However, some works have suggested South America as the most likely ancestral area from the regions currently hosting Nothofagus (Swenson et al. 2000).Molecular phylogenies were calibrated using fossil data mostly pollen and indicated a possible origin of Nothofagaceae around 65-70 Ma with the four subgenera clearly differentiated.This time interval is characterised by the appearance and diversification of the stem lineage of genus Nothofagus (Cook and Crisp 2005;Knapp et al. 2005;Premoli et al. 2011;Sauquet et al. 2012).
Even though the fragmentary specimens or the lack of good taxonomic descriptions, the most ancient fossil leaves of Nothofagus date back to the Late Cretaceous of western Antarctica and provide an excellent evidence for understanding floristic disjunction in the Southern Hemisphere (Hayes et al. 2006;Kvaček and Vodrážka 2016;Leppe et al. 2016;Romero et al. 2019).They suggest that the origin and early diversification of the genus might have occurred in the Antarctic Peninsula.In spite of this, Antarctica was excluded in many former biogeographic analyses because there is no extant species of Nothofagus inhabiting this area today (Swenson et al. 2001a) or due to unclear descriptions and illustrations of the fossil specimens (Barton 1964).In a recent overview of the past flora from Antarctica, the incorporation of this area in biogeographical analysis was suggested by Estrella et al. (2019).A main hypothesis postulated the idea of South America and the Antarctic Peninsula as ancestral areas of genus Nothofagus (Hill 1992;Dutra and Batten 2000;Leppe et al. 2012) while other authors postulated Asia-Australasia by the presence of a Fagalean complex (van Steenis 1971;Hill 1992).In this regard, many areas for the origin, dispersal routes, and diversification have been proposed for Nothofagus but they are not free of complications (Hill 1992).The available fossil evidence including leaves, wood, and pollen grains together with a vicariance-cladistic analysis reaffirmed a southern origin of Nothofagus (Tanai 1986;Poole 2002).Our biogeographic analysis employing macrofossil records from the ancient continent suggests that the diversification and distribution area was the Antarctic Peninsula and South America (Figure 2).The fossil pollen record from western Antarctica is remarkably consistent in these areas and point out a possible centre of evolution and dispersion there or in nearby regions (Barreda et al. 2019;Romero et al. 2019).This idea is also supported by fossil leaves from the same area (Figure 2).
The calibrated age (Late Cretaceous-early Palaeocene) for the subgenera Podocarpus, also present in western Antarctica, suggests an Atlantic-subtropical biogeographical corridor between South America and Africa long after the break-up of Gondwana (Quiroga et al. 2016).However, the absence of Nothofagus in the Cretaceous successions of Africa, India, and Madagascar might be for the existence of an oceanic barrier that separated these landmasses from the rest of Gondwana during the Late Cretaceous, or to the more northerly position of these areas with respect to Australasia and South America (Hill 1992;Dutra and Batten 2000;Heads 2006).

Nothofagus cladogenetic events and biogeographic pathways
The distribution and dispersion of Nothofagus can be explained by assuming land connections or at least a closer proximity in the past, between the land areas of South America, Antarctic Peninsula, and Australasia not necessarily all at one time (Couper 1960;Leppe et al. 2012;Kemp et al. 2014).Vicariance events for the four subgenera were congruent with allopatric speciation and dispersal was significant since the Late Cretaceous, when the subgenera evolved as it was inferred for the fossil record (Heads 2006).
Even though some fragmentary leaves were not incorporated in our analysis, undoubtedly, the first fossil leaves with morphological characters of Nothofagus such as the simple or compound teeth and the venation pattern, come from the Late Cretaceous (Hayes et al. 2006;Leppe et al. 2016;Romero et al. 2019).This genus has been the main angiosperm represented in the paleoflora of austral regions since the Cretaceous (Reguero et al. 2013).Macrofossils were recorded since the early and late Campanian (~77-86 Ma) in the Antarctic Peninsula (Leppe et al. 2016) with a single and incomplete record in the Coniacian (~86-90 Ma) and Santonian-early Campanian sediments of Hidden Lake Formation, James Ross Island (Hayes et al. 2006;Kvaček and Vodrážka 2016).The most ancient leaf impression in South America were discovered in the lower Maastrichtian (~68.9-71.4Ma) of Chilean Patagonia but a detailed taxonomic description was not published yet (Leppe et al. 2016).
On the other hand, the pollen record of Nothofagus in Antarctica and South America is extremely rich and provides the most complete record throughout the geological time (Romero 1986;Dettman et al. 1990;Romero et al. 2019;Scasso et al. 2020;Pujana et al. 2021).The oldest fossil pollen grains were recorded from the Santonian (~85-83 Ma) of the Antarctic Peninsula and the Campanian-Maastrichtian of southern South America (Dettmann et al. 1990;Barreda et al. 2019;Pujana et al. 2021).
The age difference of the findings in Antarctica and South America may indicate the existence of a geographic barrier that avoided the dispersion of Nothofagus.The asynchrony in the presence of leaves could be interpreted as evidence of land discontinuity and latitudinal or climatic differences between the Antarctic Peninsula and southern South America during the Late Cretaceous, that prevented the dispersion from Antarctica to Patagonia.The land barrier probably disappeared during the lower Maastrichtian, enabling colonisation from Antarctica (Leppe et al. 2016).However, biogeographic bridges connecting Patagonia and the Antarctic Peninsula during the last interval of the Late Cretaceous were proposed (Linder and Crisp 1995;Leppe et al. 2012;Reguero et al. 2013;Kemp et al. 2014).The Late Cretaceous to Early Palaeogene (70-41 Ma) was an interval of climatic and biotic changes with fluctuations in the global sea level (Kemp et al. 2014).This fact would demonstrate the existence of a cold pulse that produced the formation of ice at the poles and the drop in the sea level, leaving the arc of islands between Antarctica and southern South America land masses emerging (Leppe et al. 2012(Leppe et al. , 2016)).Fragmentary fossil leaves of Nothofagus were found in the late Campanian sediments of New Zealand, being the oldest records found in Australasia (Pole 1992).These leaves could have arrived by long-distance dispersion at the end of the Cretaceous when ice sheets were probably still present due to the cooling (Leppe et al. 2012;Kemp et al. 2014).Dutra and Batten (2000) postulated that leaves of N. glaucifolia have a high affinity with the extant species of subgenus Lophozonia but these assumptions were only based on simple observations.Our results show the mentioned taxon is closely related to the Eocene N. multinervis from western Antarctica, and with the extant N. obliqua, and N. glauca into Lophozonia.Nothofagus glaucifolia can be regarded as the ancestor of the taxa into Lophozonia (Figures 1 and 2).Nothofagus glauca is only present in South America and it could be considered as a relic of the ancient N. glaucifolia, and together with N. obliqua probably diversified later in the Pliocene-Pleistocene (Figure 1).Probably, members of Lophozonia were later dispersed by long distances and evolved simultaneously in southern South America and Australasia (Figure 2).The extant Nothofagus cunnighamii Oersted (Southeastern Australia and Tasmania) and Nothofagus gunnii (Hooker) Oersted (Tasmania) have a long evolutionary history and possess the largest temporal range (Figure 1), owing to fossil specimens with the same morphological characters (venation pattern, leaf size, teeth shape) were found in Tasmania (early Eocene-Oligocene) and west Antarctica (late Oligocene), respectively (Hill 1984(Hill , 1991)).
The Antarctic fossil taxa N. zastawniakiae, N. betulifolia and Nothofagus cretacea Zastawniak together with Nothofagus subferruginea (Dusén) Tanai (re-classified as Nothofagus hilii Dutra and Batten) from the Antarctic Peninsula, and South America share characters such as the leaf size, apex and base shape, and the number of secondary veins with the extant N. alessandrii all theminto the subgenus Fuscospora (Figure 1).
Nothofagus cretacea is the earliest and best preserved fossil leaf found in Antarctica (Half Three Point Formation) with a Late Cretaceous age (Gao et al. 2018 and references therein).This taxon was also found in the Zamek Hill Formation (early-middle Eocene) (Dutra and Batten 2000;Mozer et al. 2015).According to the morphological features observed in this taxon such as teeth shape and secondary venation, its botanical affinity was assigned to Fuscospora by Zastawniak (1994), and the position into that subgenus is reaffirmed with our results (Figure 1).A fossil leaf with similar morphological features assigned to Nothofagus sp. was found in sediments of the Vega Island (Late Cretaceous) by Romero et al. (2019), who indicated a possible affinity to N. cretacea.
Currently, representatives of Fuscospora are more diverse in New Zealand than other areas (Heads 2006).However, the fossil evidence indicates that the subgenus was present in western Antarctica since the Late Cretaceous and it was later dispersed to South America, Australia, New Zealand, and Tasmania (Figures 2  and 3a-b).Fossils of Fuscospora were common in the Eocene, but became increasingly rare throughout the Oligocene and Neogene (Cantrill and Poole 2012).It was widespread in Gondwana during the Early Tertiary, but is now much more restricted probably due to changes in the photoperiod as the landmasses moved and the climate conditions (Hill 1991).Its distribution and decreasing range may indicate this subgenus as ancestral with its origin in the Antarctic Peninsula (Figure 2).After the gradual isolation of Antarctica, during the Eocene-Miocene, multiple passageways opened in the region and climate conditions began to change, producing an effect on vegetation distribution (Cantrill and Poole 2012).Species of Nothofagus probably decreased in a forest transition that produced their extinction in Antarctica after the Oligocene (Figure 3b).
Several investigations about the biogeography of Nothofagus considered that the dispersion and distribution reflect vicariance as a consequence of the break-up of Gondwana (Swenson et al. 2001a, Swenson et al. 2001b;Knapp et al. 2005).Yet, the results of the DEC+J model herein described suggest a biogeographical history where dispersal and vicariance have been fundamental in shaping the current distribution pattern in Nothofagaceae (Figure 2), especially for the early diversification of the genus.This result is congruent with the findings made by Cook and Crisp (2005) who also indicated that both events have clearly played a role in the current distribution patterns.
The diversification of Nothofagus is often attributed to the final break-up of Gondwana at the end of the Cretaceous, but this event probably occurred too late to have been involved in the evolution of the extant subgenera because they already had an extensive pollen record in the Late Cretaceous (Dettmann et al. 1990;Heads 2006).The calibrated relaxed molecular clock performed by Knapp et al. (2005) indicated that the biogeography of Nothofagus is a little more complex.The authors highlighted the potential study of the mechanism of long-distance dispersal (trans-oceanic) of Nothofagus seeds and concluded that its evolution involved both dispersal and vicariance.Although past long-distance dispersal events result highly improbable due to the scant evidence, some authors have supported this idea based on the Nothofagus distribution (Hill 1992;Martin and Dowd 1993;Hill and Dettmann 1996;Swenson et al. 2001a, Swenson et al. 2000b).For example, according to the results found by Swenson et al. (2001a) N. gunnii and N. menziesii were probably dispersed by long distances.However, the long-distance dispersion of nuts appears to have limitation to move and have been assumed to be incapable of trans-oceanic dispersal (Hill 1992;Heads 2006).
Based on the sequence of Gondwana break-up, a hypothesis of vicariance predicts that Australian Nothofagus species should be most closely related to South American species when continents were connected via Antarctica until around 35 Ma (Knapp et al. 2005).Northeastern Australia may have had some early Nothofagus or an ancestor to it, but the subgenera arrived late, after land-based dispersal from southern Australia.Subgenus Brassospora spread to New Guinea through eastern Australia (Hill 2004).At the moment, there are no documented macrofossils of this subgenus in Antarctica or southern South America (Manos 1997;Hill 2001;Pujana et al. 2021).
On the other hand, molecular phylogenies showed that the crown group of Nothofagus has a recent origin during the Miocene as it was pointed out by Cook and Crisp (2005).Our timecalibrated phylogeny reveals the earliest presence in southern South America of subgenus Nothofagus in the Palaeocene with a clear diversification in the middle Eocene-Oligocene (Figures 1 and 2).Leaves of N. lobata from Tasmania placed in subgenus Nothofagus represent the single fossil record of this subgenus outside South America (Hill 1991) and it is closely related to the extant N. nitida (Figure 1).Dispersion and vicariance events probably separated these species.Nothofagus lobata and its relationship with representative subgenus Nothofagus will confirm an allopatric speciation from a common ancestor probably placed in an ancestral area of the Antarctic Peninsula (Figure 2).A combination of large-scale disturbance and an increase in seasonality probably caused the extinction of this subgenus from that area and the concentration of its biogeographical range to Southern South America today (Scriven and Hill 1996).
The subgenus Brassospora includes the Oligocene-Miocene fossil taxa Nothofagus kiandrensis Paull and Hill from southeastern Australia, Nothofagus serrata Hill, Nothofagus mucronata Hill from Tasmania, and Nothofagus palustris Carpenter, Bannister, Lee and Jordan from New Zealand (Figure 2).The fossil N. palustris is the unique well-preserved fossil of Brassospora discovered in New Zealand and probably a link between Australia and New Guinea (Carpenter et al. 2014).According to the fossil pollen record, diversification of this subgenus occurred during the Eocene-Oligocene in Australia and New Zealand (Hill 1992).It was dominant in southeastern Australia during the Tertiary, and most of the Oligocene representatives were morphologically similar to the extant species found in New Guinea.Brassospora appears as a most recent clade (Figure 1) with a middle Eocene age and its diversification in Australasia was probably by dispersion and vicariance events (Figure 2).The subgenera Brassospora and Nothofagus represent clades that evolved in endemism areas with a later diversification (Figure 2).The geographic connection of southernmost South America and the Antarctic Peninsula, currently disjointed by the Drake Passage, may partly explain the vicariant pattern and areas of endemism (Leppe et al. 2012).The separation of the land masses and the origin of islands in Australasia might have contributed to an allopatric speciation (Figure 3a-c).The wood record found in western Antarctica also reaffirms the hypothesis that modern subgenera of Nothofagus had begun to diverge in southern South America and the Antarctic Peninsula by the late Campanian (Poole 2002).
Several works have explained the biogeography of genus Nothofagus exclusively by vicariance.The model based only on this requires that the four subgenera, which match the four pollen types, were widespread along the Weddell coast of Antarctic Peninsula before the final the break-up of Gondwana and spread over the land masses in the Late Cretaceous (Linder and Crisp 1995;Setoguchi 1997).Cook and Crisp (2005) analysed vicariance scenarios that can be rejected because the radiation of the extant crown group was too recent.They postulated that the divergence between Australian and New Zealand in Fuscospora and Lophozonia occurred less than 50 Ma.

Conclusions
The exclusive Southern Hemisphere family Nothofagaceae was a diverse group of angiosperms that existed before the final break-up of Gondwana and later proliferated in southern South America and eastern Australasia with the subsequent extinction in the Antarctic Peninsula.Fossil leaves of genus Nothofagus from Antarctica, South America, and Australasia included in our phylogenetic and biogeographic analysis contributed to a more complete model of its evolutionary history.Our results support Antarctic Peninsula as the most likely ancestral area of origin for this family in the Late Cretaceous with the oldest subgenera Fuscospora.The available fossil evidence (pollen and leaves) indicates that the first representatives arrived in southern South America during the Late Cretaceous.After our analysis, we concluded that both dispersion and vicariance have driven the diversification of species of Nothofagus.Migration and speciation took place after the separation of the Antarctic Peninsula and South America by temporary ice bridges at the end of the Cretaceous due to a cooling event.The addition of fossil leaves from the Antarctic Peninsula allow us to elucidate that N. cretacea is the oldest and best preserved macrofossil record until now.The temporal distribution range in the Antarctic Peninsula indicates that the youngest leaf records in that area correspond to the Eocene-Oligocene.
Nothofagus cunninghamii and N. gunnii preserve their leaves almost without morphological changes since their origin at the end of the Palaeocene and they can be considered as extant fossils.In addition, the endemic N. alesssandrii from southern South America is the oldest extant taxa that has a close relationship with the fossils N. zastawiniakiae, N. betulifolia, N. subferruginea, and N. cretacea into the ancient subgenus Fuscospora from the Antarctic Peninsula.The fossil N. glaucifolia from the Antarctic Peninsula may be considered as the ancestor of the extant N. glauca that evolved in southern South America.
The subgenus Nothofagus is the youngest clade with a Palaeocene-Eocene age in southern South America and Brassospora in Australasia with an Eocene-Oligocene age.Currently, both clades represent endemism processes of the crown group with the consequent diversification.The taxon N. lobata is the unique fossil of subgenus Nothofagus discovered in Tasmania, probably from a common ancestor from the Palaeocene-early Eocene of southern South America.

Figure 1 .
Figure 1.Time calibration tree from genus Nothofagus.The tree is time-scaled (Gradstein et al. 2012) using DatePhylo function and timeSliceTree from paleotree package (Bapst 2012).The bold type letter indicates fossil species.The time scale is in millions of years.

(
Figure 2).The ancestral area of subgenus Lophozonia (node 84) is indicated as the Antarctic Peninsula (57%).There is a clear differentiation of Lophozonia species in two clades.The southern South America and Antarctica clade, and Tasmania and a few taxa from New Zealand and Southeast Australia clade.Within Lophozonia, there are dispersal and vicariance events (node 75) for the fossil taxa Nothofagus multinervis Haomin and Zhekun from western Antarctica which is grouped with the extant Nothofagus obliqua (Mirbel) Oersted and Nothofagus glauca (Philippi) Krasser (currently endemic of the southernmost Chile (Moya et al. 2017) from southern South America.The Antarctic Peninsula is indicated as the ancestral

Figure 2 .
Figure 2. Reconstructed ancestral distributions from DEC+J analysis in RASP using dated phylogeny including fossils and extant species of Nothofagus.The bold type letter indicates fossil species.Pie charts represent the marginal probabilities for each alternative ancestral area.Black sections of pie charts (*) designate uncertainties or areas with marginal probability below 5%.The ancestral areas at each node are indicated by letters.Arrows indicate the most probable direction of dispersal events during the Late Cretaceous.The time scale is millions of years ago (Ma) .