Origin and divergence of Afro-Indian Picrodendraceae: linking pollen morphology, dispersal modes, fossil records, molecular dating and paleogeography

Abstract The pantropical Picrodendraceae produce mostly spheroidal to slightly oblate, echinate pollen grains equipped with narrow circular to elliptic pori that can be hard to identify to family level in both extant and fossil material using light microscopy only. Fossil pollen of the family have been described from the Paleogene of America, Antarctica, Australia, New Zealand, and Europe, but until now none have been reported from Afro-India. Extant pollen described here include representatives from all recent Picrodendraceae genera naturally occurring in Africa and/or Madagascar and south India and selected closely related tropical American taxa. Our analyses, using combined light microscopy and scanning electron microscopy, show that pollen of the Afro-Indian genera encompass three morphological types: Type 1, comprising only Hyaenanche; Type 2, including Aristogeitonia, Mischodon, Oldfieldia and Voatamalo; Type 3, comprising the remaining two genera, Androstachys and Stachyandra. Based on the pollen morphology presented here it is evident that some previous light microscopic accounts of spherical and echinate fossil pollen affiliated with Arecaceae, Asteraceae, Malvaceae, and Myristicaceae from the African continent could belong to Picrodendraceae. The pollen morphology of Picrodendraceae, fossil pollen records, a dated intra-familial phylogeny, seed dispersal modes, and the regional Late Cretaceous to early Cenozoic paleogeography, together suggest the family originated in the Americas and dispersed from southern America across Antarctica and into Australasia. A second dispersal route is believed to have occurred from the Americas into continental Africa via the North Atlantic Land Bridge and Europe.

lato (Webster 1994a(Webster , 1994b. The distribution of Picrodendraceae is disjunct across the subtropical and tropical Southern Hemisphere, with some limited representation in cooler temperate zones. Representatives of the family occur in the Americas (five genera), Africa (six genera; Table I), south India and Sri Lanka (one genus; Table I) and in Australasia (13 genera; see Supporting Information). The greatest diversity at both the genus and species levels occurs in Australia with ten genera (seven endemic) and approximately 40 spp (WCSP 2018). The largest genus is Austrobuxus with 22 spp, of which 15 are endemic to New Caledonia (WCSP 2018; see Supporting Information). The extensive vegetative diversity of Picrodendraceae is reflected in the widely different habitats and environments the family occupies (noted from herbarium sheets examined for this study): a range of altitudes from near sea level to over 1800 m; in flooded forests in the Amazon; deserts in the south-western United States and northern Mexico; dry deciduous forests, humid evergreen forests, woodlands, in sand and on rocky hills in Africa, Madagascar, and south India; dry fynbos and savanna in South Africa (Dyer 1975); drier rainforests and woodlands on sandy or serpentine soils in Australia; and rain forests in New Caledonia.
Picrodendraceae includes monoecious or dioecious subshrubs, shrubs, and small to large trees with simple or palmately-compound leaves; tiny, apetalous, mostly unisexual flowers; anthers of male flowers few or to as many as 55; ovaries 2-7locular with biovulate locules; fruit capsular and dehiscent or drupaceous; seeds one or two per locule; and copious endosperm. The floral structure indicates a close relationship to the Phyllanthaceae (formerly fam. Euphorbiaceae s.l., subfam. Phyllanthoideae), which is supported in molecular phylogenetic analyses of the Euphorbiaceae (Wurdack et al. 2004;Wurdack & Davis 2009) and Malpighiales (Soltis et al. 2011;Xi et al. 2012). Morphologically, Picrodendraceae and Phyllanthaceae are united by unspecialised monopodial branching patterns, a lack of laticifers, and biovulate locules in the capsular fruit (Webster 1994a(Webster , 1994bMerino Sutter et al. 2006). In an extensive phylogeny of the order Malpighiales, sequences of 82 plastid genes produced a fully supported tree in which Picrodendraceae and Phyllanthaceae were confirmed as sister families (Xi et al. 2012). The South American genus Podocalyx was retrieved as sister to the other Picrodendraceae, with Tetracoccus + Androstachys, sister to a clade of Petalostigma, Dissiliaria, and Micrantheum + Austrobuxus.
The unique morphology of Picrodendraceae pollen (± spherical, echinate, stephanoporate) and its systematic value was first pointed out by Erdtman (1952) and later by Punt (1962) and Köhler (1965). The pollen morphology of Afro-Indian Picrodendraceae has been studied to some extent (Table II). Among 20 species, eight have been studied using transmission electron microscopy (TEM) (Hayden et al. 1984;, six have been studied using scanning electron microscopy (SEM) (Hayden et al. 1984;Lobreau-Callen & Cervera 1994;, and four have been illustrated using light microscopy (LM) micrographs (Köhler 1965). These studies documented some variation in the pollen morphology among genera or at the species level of Afro-Indian Picrodendraceae that suggested fossil Picrodendraceae pollen grains could be associated with specific intrafamilial lineages.
The current disjunct tropical to subtropical distribution of Picrodendraceae implies that it already had a worldwide distribution in early Cenozoic times. This is supported by fossil pollen records from the Paleocene and/or Eocene of Australia (Martin 1974(Martin , 1978(Martin , 1982, New Zealand (Raine et al. 2011), Europe (Zetter & Hofmann 2008;Hofmann et al. 2011;Zetter et al. 2011) and North America (Tschudy & van Loenen 1970;Tschudy 1973;Graham et al. 2000). Interestingly, there is not a single fossil record of Picrodendraceae from Afro-India.
Here we describe and illustrate pollen from 12 of the 20 (12/20) extant taxa currently accepted in Afro-India (Table III), including: Androstachys (1/ 1), Aristogeitonia (4/7), Hyaenanche (1/1), Oldfieldia (3/4), Stachyandra (1/4), Voatamalo (1/2) and Mischodon (1/1). Pollen of Tetracoccus (1/4) and Piranhea (2/4), from the Americas, are also portrayed as they represent early branching taxa/lineages of the American-Afro-Indian clade. The extant African pollen grains are compared, grouped by morphological type, and their diagnostic features highlighted. Fossil Picrodendraceae pollen from the Miocene of Africa and their Eocene predecessors from Europe are also described. The morphological traits of extant taxa are here used to associate fossil Picrodendraceae pollen grains from Europe and Africa with extant genera and/or lineages. Further, the pollen morphology is discussed in relation to dispersal methods, the fossil record, paleogeography, and the current molecular phylogenetic framework, to unravel the paleophytogeographic history of this family in Afro-India. A dated phylogeny is presented supporting the hypothesis that Picrodendraceae originated in the Upper Cretaceous of the Americas from where they dispersed across the world during the Cenozoic.

Origin and preparation of samples
Extant flower material (see Table III) from the Missouri Botanical Garden (MO), the Royal Botanic Gardens, Kew (K), the South African National Biodiversity Institute (PRE), the National Museums of Kenya (EA), and Naturalis (WAG) was prepared according to the protocol outlined in Grímsson et al. (2017aGrímsson et al. ( , 2018b and Halbritter et al. (2018).  (5) the Elandsfontyn Formation (early Miocene), core sample #114755 from Saldanha Bay, South Africa. For details on the geographic positions, geology, palaeoecology, and previously reported fossil plants from these formations and localities see Table IV and references cited therein.
The sedimentary rock samples were processed and fossil pollen grains extracted according to the method explained in Grímsson et al. (2008). The fossil Picrodendraceae pollen grains were investigated both by LM and SEM using the single grain method as described by Zetter (1989) and Halbritter et al. (2018). SEM stubs with extant and fossil Picrodendraceae pollen produced under this study are stored in the collection of the Department of Paleontology, University of Vienna, Austria.

Tree inference
Sequences for five mitochondrial loci (ccmB, cob, nad1, nad6, mt-rps3), three nuclear loci (18S, EMB2765, PHYC), and four plastid loci (atpB, matK, ndhF, rbcL) from Picrodendraceae were downloaded from NCBI GenBank. Sequences for each locus were aligned using MAFFT v.7.273 (Katoh & Standley 2013) using the L-INS-I algorithm for all loci. Alignments were visually inspected for irregularities and truncated at the end for portions covering data of less than seven genera. The single locus alignments were then concatenated using MESQUITE (Maddison & Maddison 2011). Phylogenetic analyses were run using RaxML v.8.2.10 (Stamatakis 2014), which we used to obtain 1000 (fast) bootstrap replicates, and MrBayes v.3.2.6 (Ronquist et al. 2012), where we conducted two runs with one cold chain and three heated chains for 1 900 000 mcmc generations, with 25% of the generations discarded as burnin. Parameters of the mcmc runs were inspected using Tracer (Rambaut & Drummond 2003) to check for convergence. In both cases, data were partitioned by genome (mitochondrial, plastidial, and nuclear), and an unlinked GTR model with gamma-distributed rate variation was employed. RaxML bootstrap replicates and post-burnin samples from both runs were synthesised using consensus networks (Holland & Moulton 2003) in SplitsTree (Huson & Bryant 2006), using a 20% cutoff.
To obtain a timescale for Picrodendraceae, we conducted a molecular dating analysis with the fossilised birth-death prior (Heath et al. 2014) as implemented in MrBayes v.3.2.6. Fossil pollen taxa compiled for this study were coded as tips, with a uniform prior on their age reflecting uncertainty in the dates. Fossils were constrained in clades with their most probable extant relatives based on informal assessment of their micromorphology. For the molecular data, we employed a relaxed IGR clock, with the same substitution model implemented for the non-clock analysis (see earlier). We set the sampling probability of the extant taxa as 0.26 (based on species count of extant Picrodendraceae), and we set the sample strategy as diversified. We ran two independent runs with four chains (one cold, four heated) for 50 000 000 generations. A consensus tree was generated using all compatible splits.

Systematic palynology
The pollen terminology follows Punt et al. (2007;LM) and Halbritter et al. (2018;SEM). The classification and author names of extant taxa follow WCSP (2018). Genera and species are arranged in alphabetical order with the pollen of each taxon described individually, followed by descriptions of the fossils at the end. Pollen grains of all the extant American-Afro-Indian Picrodendraceae species studied here are compared in Table V. Three morphological types are recognised, referred to here as PT 1-PT 3. Fossil Picrodendraceae pollen grains are compared in Table VI. For practical reasons all the fossil pollen grains are classified as morphotypes (MT) named after the locality where they were found.
Remarks. -The irregular interval and displacement of pori outside of the equator gives the pollen a slightly angular outline in polar and equatorial view. Pollen of this taxon were figured by Köhler (1965, plate 9, figure 9 [LM]) and Simpson and Levin (1994, figures 35, 36 [SEM]). The SEM close-up figured in Simpson and Levin (1994) shows echini that look similar in size and outline to what is observed in the material studied herein, but lacks the resolution needed to identify any sculpture elements in the areas between the echini. Micrographs showing the ultrastructure of Androstachys pollen, from both non-apertural and apertural regions, have also been provided by Simpson and Levin (1994, figures 45, 46 [TEM]).

252
F. Grímsson et al. the areas between the echini can be bigger and more conspicuous than observed in the present material. The ultrastructure of Stachyandra merana pollen is figured by Simpson and Levin (1994, figure 47 [TEM]).
Remarks. -This is an extremely rare element in the Krappfeld palynoflora. The measurements presented earlier are based on a single specimen (versus 20 in extant material) and most likely do not convey the complete natural size ranges of this MT. The Krappfeld MT is not identical to pollen from any extant African Picrodendraceae, but shares most features with the African Hyaenanche and especially the American Piranhea. The Krappfeld MT pollen in size is closer to that of Piranhea than to the larger Hyaenanche pollen, and it also has numerous prominent nanogemmae that seem to encircle perforations as in Piranhea.
In fact, the main difference discriminating the Krappfeld MT pollen grain from those of Piranhea is the shape and outline of echini and especially their prominently striate surface. The pollen morphology suggests that the Krappfeld MT represents an extinct early diverging taxon of the American-Afro-Indian clade, positioned close to the Piranhea lineage, and is ancestral form leading to the Afro-Indian clade (including Hyaenanche). The fossil Picrodendraceae pollen from the Eocene of Austria (Krappfeld MT) and Germany (Stolzenbach MT and Profen MT) are very similar. They clearly represent a closely related stock and might even originate from the same biological taxon.  Stolzenbach MT, pollen close to Piranhea (Figures 16, 17 (Table IV).

Afro-Indian Picrodendraceae Pollen
Remarks. -This is a rare element in the Stolzenbach palynoflora. The measurements presented earlier are based on four specimens (versus 20 in extant material) and most likely do not convey the complete natural size ranges of this MT. For further notes see 'Remarks' of the Krappfeld MT.
Remarks. -This is an extremely rare element in the Profen palynoflora. The measurements presented earlier are based on a single specimen (versus 20 in extant material) and most likely do not convey the complete natural size ranges of this MT. For further notes see 'Remarks' of the Krappfeld MT.   (Table IV).
Remarks. -This is not a rare element in the Mush palynoflora and can be found in various stages, in both perfect preservation or compressed and/or broken. The combined features observed in LM and SEM (± spherical, stephanoporate, elliptic pori, echinate sculpture) clearly place these fossil pollen grains in Picrodendraceae. The fossil Mush MT differs considerably from that of Hyaenanche globosa (PT 1). The fossil pollen grain is larger, it has much fewer echini per 100 µm 2 in the central polar area (4-8 versus 15-30) and the echini are also higher (4.1-5.2 versus 1.0-2.0 µm). The sculpture between the echini is nanogemmate to nanorugulate in the Mush MT, but fossulate, perforate, and nanogemmate in Hyaenanche. The Mush MT also differs in outline from pollen of Androstachys (PT 3) and Stachyandra (PT 3) (circular versus elliptic to slightly angular) and in aperture position (regular intervals and at the equator versus irregular intervals and displaced). The echini are also much higher in the Mush MT (4.1-5.2 versus 0.7-1.3 µm) and of different shape and wider apart (4-8 per 100 µm 2 versus 35-45 per 100 µm 2 ) than in both Androstachys and Stachyandra. Also, the SEM sculpture is nanogemmate to nanorugulate in the Mush MT, but clearly granulate and perforate in Androstachys and Stachyandra (Tables V, VI). The Mush MT shares many features with extant pollen of PT 2 (Aristogeitonia, Mischodon, Oldfieldia and Voatamalo). The pollen body of the Mush MT is very large, between 32 and 35 µm in diameter (excluding echini); similar sized pollen is observed in extant Aristogeitonia monophylla, Mischodon, Oldfieldia, and Voatamalo eugenioides (Table V). All these extant genera also display few (less than 15 per 100 µm 2 ) widely spaced echini in the central polar area, that are relatively high (up to 4.5 µm), a feature also characteristic for the Mush MT. The SEM sculpture observed between the echini in the Mush MT is also very similar to that of PT 2 pollen. There is one prominent difference separating the fossil Mush MT from the extant PT 2 pollen and that is the massive thickness of the solid nexine in the Mush MT (see SEM break in Figure 20D, Locality. -Saldanha Bay drill core, South Africa, early Miocene (Table IV).
Remarks. -Most of the fossil pollen grains are infolded or compressed (flattened). The range of the polar axis is based on only two measurements (versus 20 in extant material) and most likely does not convey the complete natural length between the poles in this MT. Based on the pollen morphology of extant Afro-Indian Picrodendraceae genera it is clear that the pollen should be assigned to genus Hyaenanche (PT 1). The fossil Saldanha MT pollen grains differ in outline from those of Androstachys and Stachyandra (PT 3; elliptic to slightly angular versus circular) and in aper-   times reach 2.5 µm. The SEM sculpture in areas between the echini is nanogemmate to nanorugulate to granulate in Aristogeitonia, Mischodon, Oldfieldia and Voatamalo, but fossulate, perforate and nanogemmate in extant Hyaenanche (Table V). The fossil Saldanha MT pollen grains are extremely similar to those from the extant species, Hyaenanche globosa. The outline and size of the pollen are similar, the position, number, size and outline of the pori are also the same, and they show the same sculpture in SEM. The echini are of similar size, shape and number in the central polar area, the sculpture in areas between echini compares closely, and the preserved parts of the aperture membrane observed in the fossil pollen suggest that it is the same as in the extant species (Tables V, VI).

Taxonomic value of Afro-Indian Picrodendraceae pollen
Extant Afro-Indian Picrodendraceae pollen can be divided into three morphological types, PT 1-PT 3. The pollen of Hyaenanche (PT 1) is clearly unique among the Afro-Indian taxa. The sculpture is echinate, but fossulate, perforate and nanogemmate in areas between echini, and there are 15-30 echini per 100 µm 2 in the central polar area (Figure 6, Table V). The echini are of intermediate size (1.0-2.0 µm high) and the aperture membranes are nanogemmate.
Many of the Afro-Indian genera produce PT 2. This includes, Aristogeitonia (Figures 2-5), Mischodon (Figure 7), Oldfieldia (Figures 8-10), and Voatamalo ( Figure 15). PT 2 is mostly spheroidal (P/E ratio) and circular in outline (polar and equatorial). The pori are placed at regular intervals around the equator. The sculpture is echinate, with tall (2.0-4.5 µm) but few (3-13 per 100 µm 2 ) echini, and nanogemmate to granulate in areas between echini and on the aperture membrane. There are some subtle differences observed among species in the size of the PT 2 pollen grains and their sculpture elements, but there is a complete overlap among the genera (Table V). Taxa from the same geographic region (e.g. Voatamalo versus Aristogeitonia on Madagascar; Oldfieldia versus Aristogeitonia in eastern Africa; Figure 23) are impossible to distinguish on the basis of their dispersed pollen.
PT 3 occurs in two genera, Androstachys ( Figure 1) and Stachyandra (Figure 13). This pollen is characterised by irregularly placed pori, microechinate sculpture (echini 0.7-1.3 µm high and 30-45 per 100 µm 2 ) and nanoechinate aperture membranes ( Table V). The pollen grains of the two taxa studied, Androstachy johnsonii and Stachyandra merana, are so similar that there is no way to distinguish them using LM and/or SEM. The two genera are debatably congeneric (Radcliffe-Smith 2001; Webster 2014).

Origin, divergence, and dispersal of Afro-Indian Picrodendraceae
Previous molecular studies using fossil age constraints suggested that crown Malpighiales began to radiate in the late Early Cretaceous (middle Aptian to middle Albian), at 119.4-110.7 Ma (Davis et al. 2005) or 113.1-106.1 Ma (Xi et al. 2012) (but see also Wikström et al. 2001;Magallón & Castillo 2009;Bell et al. 2010). So far, estimates for the age of the Phyllanthaceae/Picrodendraceae clade fall in the latest Early Cretaceous (Albian), at 114-105.8 Ma (Davis et al. 2005) or in the early Late Cretaceous (Cenomanian to Coniacian), at 101-86.5 Ma (Xi et al. 2012). Estimates for the age of crown-group Picrodendraceae suggest a middle Late Cretaceous (Turonian to Campanian) origin at 92.9-72 Ma (Xi et al. 2012).
Since all the Afro-Indian Picrodendraceae form a monophyletic clade (Figures 24, 25;Wurdack 2008;Wurdack & Davis 2009), it is clear they descended from a single common ancestor that at some point dispersed into Africa. The molecular phylograms of Wurdack andDavis (2009), andXi et al. (2012), with seven genera, place the South American Podocalyx as sister to the other six genera of the Picrodendraceae that were sampled, which form a grade from Tetracoccus to Austrobuxus. The bipartition network presented here ( Figure 24) suggests two major clades, one including all American and African taxa, and the other including all Australasian taxa, with a potential divergence (or origin) point to be placed somewhere close to Podocalyx.
It is interesting that PT 1, found in the African Hyaenanche, is extremely similar to pollen occurring in the early branching Tetracoccus (second branching American lineage; Figure 24) and in Piranhea, that is part of the sister clade to the Afro-Indian taxa. PT 1 is clearly a basal or primitive (plesiomorphic) pollen occurring in early diverging/branching Picrodendraceae. This suggests that PT 1 is the 'original' pollen type of the lineage that dispersed into Africa from the Americas; therefore PT 1 is ancestral to PT 2, a pollen type occurring in most of the extant Afro-Indian taxa (Aristogeitonia, Mischodon, Oldfieldia and Voatamalo) and showing the widest extant distribution ( Figure 23). Lineages with PT 2 subsequently led to the evolution of PT 3, occurring in both Androstachys and Stachyandra.

264
F. Grímsson et al. Kairothamnus, Longetia, Micrantheum, Mischodon, Neoroepera, Paradiodendron, Piranhea, Podocalyx, Pseudanthus, Sankowskya, Scagea, Stachyandra, Stachystemon, Tetracoccus, Voatamalo, Whyanbeelia) resulting in very restricted autochorous dispersal (up to a few metres), believed to be primitive in the family (Webster 1994a(Webster , 2014. About half of the genera produce carunculate seeds (Androstachys, Austrobuxus, Hyaenanche, Longetia, Micrantheum, Neoroepera, Oldfieldia, Petalostigma, Pseudanthus, Sankowskya, Scagea, Stachyandra, Stachystemon, Tetracoccus, Voatamalo, Whyanbeelia) that are dispersed by ants (myrmecochory ;Webster 1994aWebster , 2014. Only three genera have fleshy or drupaceous seeds/fruits (Oldfieldia, Petalostigma, Picrodendron) that are primarily dispersed by birds and/or mammals (endozoochory/ornithochory), a trait believed to be derived in the family (Webster 1994a(Webster , 2014. The African Oldfieldia are dispersed by mammals, mostly primates (Koné et al. 2008), the Australian Petalos-tigma by emus (Clifford & Monteith 1989), and the West Indies Picrodendron by birds and iguanas (Alberts 1999;Hines 2016). According to Webster (1994aWebster ( , 2014 the basal Phyllanthaceae, Picrodendraceae and Euphorbiaceae typically bear capsular fruits with dry seed-coats excluding long distance dispersal across wide ocean barriers. The predominant autochorous and myrmecochorous dispersal of basal Picrodendraceae and their current distribution suggests they could have dispersed across the Southern Hemisphere prior to later stages of the Gondwana breakup when ocean barriers became too wide. The Gondwana landmass started to fragment during the Early to Middle Jurassic, at 180- posed of Madagascar, India, Antarctica and Australia. Contemporaneously, Madagascar/India started to drift apart from Antarctica/Australia. This scenario continued until c. 90 Ma, when India started to break away from Madagascar, and India then continued its northward journey towards Asia. From that time on Madagascar became part of the African plate (Chatterjee et al. 2013). Also, eastern South America and west/central Africa were closely aligned until 90-80 Ma (Gaina et al. 2013), with potential pathways for plant dispersal between the two continents. Furthermore, tectonic models of East Gondwana (Gibbons et al. 2013) suggest that Africa, Madagascar and India were close until 80-70 Ma, allowing for plant dispersal among these landmasses until at least the latest Upper Cretaceous (Campanian/Maastrichtian). Additionally, paleogeographic reconstructions and terrestrial fossil records suggest that southern South America and West Antarctica were connected by a land bridge (Weddellian Isthmus) believed to have been functional until the late Paleocene (Reguero et al. 2014, terrestrial vertebrates) or the middle Eocene (Ghiglione et al. 2013, terrestrial vertebrates), or even the early Oligocene (Graham 2018b, plants). Based on the paleogeography and taking into account dispersal mechanisms of the family it is clear that if Picrodendraceae dispersed across a southern route into Africa they must have done so prior to the end Cretaceous. Also, since Mischodon has an advanced pollen type (PT 2) and is nested within the African clade (Wurdack 2008) that scenario suggests they must have dispersed from Africa/Madagascar into India in the latest Cretaceous (c. 80-70 Ma;Gibbson et al. 2013) or the early Cenozoic. But that does not fit with the fossil record of the family.
Fossil records of the Phyllanthaceae/Picrodendraceae clade can be traced back to the Upper Cretaceous. Webster (2014) summarised the macrofossil record of Euphorbiaceae s.l. citing fossil woods, fruits and inflorescences. However, neither Webster nor previous authors affiliate any macrofossil directly to the Picrodendraceae (or Oldfieldioideae of the Euphorbiaceae s. l.). Long before, Metcalfe and Chalk (1950) presented a clear overview of the wood anatomy of Euphorbiaceae s.l., in which they recognised three groups within the so-called 'Phyllanthoideae' (= today's Phyllanthaceae and Picrodendraceae): Group A (Aporosa or Aporusa

266
F. Grímsson et al. Figure 25. Maximum Clade Credibility consensus timetree of Picrodendraceae obtained using the fossilised birth-death prior. All fossils (see Table VII) have been trimmed off after the construction of the consensus tree. Pollen types for the Afro-Indian taxa are indicated. Bars indicate 95% highest posterier density (HPD) intervals.

Afro-Indian Picrodendraceae Pollen 267
Type), Group B (Glochidion Type) and 'Other genera of the Phyllanthoideae', which mainly represent Picrodendraceae (Hayden 1994 Mädel 1962;Prakash et al. 1986). Of these, only Paraphyllanthoxylon (taxa of species group B sensu Herendeen 1991) with numerous uniseriate rays and longer marginal rows, or more pronounced ray heterogeneity, can be labelled as 'phyllanthoid'. Paraphyllanthoxylon is a very broadly defined fossil genus with numerous species of various ages (Upper Cretaceous to late Cainozoic) described from many parts of the world (for a summary see Gryc et al. 2009;Nunes et al. 2018 (Table VII) are inflorescences and flowers from the late Paleocene of Tennessee (as Protoarecoidea buchananensis ;Feldman 1990). These flowers also contain in situ pollen that represent the earliest known pollen records of this family, along with dispersed pollen described from the late Paleocene of southeast Australia (Harris 1965;Stover & Partridge 1973). Following the Paleocene, fossil Picrodendraceae pollen (Table VII) are known from the Eocene of Europe (this study; Zetter & Hofmann 2008;Zetter et al. 2011), the Americas (Tschudy & van Loenen 1970;Tschudy 1973;Graham et al. 2000), Australasia (Stover & Partridge 1973) and Antarctica (Truswell & Macphail 2009).
The fossil record supports an American origin of Picrodendraceae (Table VII). The fossils at hand suggest that the family was well established in the Americas during the late Paleocene. The fossil record also supports a southern dispersal route, from South America via Antarctica into Australasia, during the late Paleocene, a migration route believed to have been open for plant dispersal during most of the Paleogene (Ghiglione et al. 2013;Reguero et al. 2014;Graham 2018aGraham , 2018b. There are, however, no fossil records from Africa, Madagascar or India from the late Cretaceous or Paleocene to demonstrate the presence of Picrodendraceae in this part of the world during that time. Thus, the European Eocene Krappfeld, Stolzenbach and the Profen fossil pollen MTs become important as the only known representatives of an extinct early diverging lineage of the American-Afro-Indian clade, positioned close to Piranhea. The current Picrodendraceae fossil records indicate the family dispersed into Africa from Europe via a second dispersal route, namely from the Americas via the North Atlantic Land bridge during periods of elevated temperatures in the Paleocene and/or Eocene. The North Atlantic Land Bridge is known as a gateway for numerous thermophilic plants, even some of today's tropical to subtropical plant groups, that were able to disperse freely between the Americas (Greenland) and Europe until the late Eocene (e.g. Tiffney 1985Tiffney , 2000Grímsson et al. 2018a). More cold tolerant plants were able to follow this route until the middle to late Miocene (Denk et al. 2010(Denk et al. , 2011Graham 2018aGraham , 2018b. Such a 'northern route' has recently been proposed for another predominantly southern hemispheric group, the Loranthaceae, believed to have conquered Africa from Asia during the Eocene (Grímsson et al. 2017b(Grímsson et al. , 2018c. During the Eocene, Europe was influenced by hot and humid climate (Zachos et al. 2001;Mosbrugger et al. 2005), comparable to that of present day tropics to subtropics. In Europe, this paleoclimate sustained a special thermophilic flora, so-called Paratropical Rainforest (e.g. Mai 1995), that disappeared at the end of the Eocene. With tropical and subtropical climate equivalents and corresponding vegetation reaching from the equator towards mid latitudes and even high latitudes, temperatures would not have been a barrier to a southward dispersal of Picrodendraceae from Europe into Africa during the Eocene. The African fossil records show that Picrodendraceae had diverged into at least two different lineages prior to the early Miocene. This is evident by the early Miocene Saldanha MT pollen, that is clearly of the extant PT 1 lineage, and the early Miocene Mush MT, that represents an extinct ancestral taxon of the extant PT 2 lineage. Studies by some of the present authors on late Oligocene to early Miocene palynofloras from Ethiopia and South Africa have yielded no additional fossil pollen that truly represent the extant PT 2 pollen. The only pollen discovered so far are the primitive basal PT 1 from Saldanha Bay (South Africa) and the extinct ancestral type to the PT 2 lineage from Mush (Ethiopia). Therefore, the current fossil data suggests that both the extant PT 2 and PT 3 pollen had not evolved during the latest Oligocene or the earliest Miocene, but are post earliest Miocene taxa. Still, this could easily change in the future as more African palynofloras are studied and screened for Picrodendraceae pollen.
The respective ages of the different Picrodendraceae lineages in the dated phylogeny ( Figure 25) support a primary dispersal from South America to Australasia via Antarctica. The Australasian clade supposedly originated between the Late Cretaceous and the Paleo-

268
F. Grímsson et al. Truswell & Macphail 2009 (Continued )  Figure 25). The age of the Afro-Indian crown group spans between the Paleocene and the Eocene (61.8-51.1 Ma). These relatively young ages are in agreement with the fossil records and suggestive of a boreotropical distribution with dispersal across the North Atlantic Land Bridge. However, since the earliest fossil records representing the Afro-Indian clade are mostly from European localities, that could bias the inferred splits towards younger ages by missing potential diversity in Afro-India. Similar biases have been reported for other groups of southern origin with mostly European records, like Solanaceae (Wilf et al. 2017).
If results are accepted from the analysis of the current African fossil records and the dated phylogeny ( Figure  25) indicating PT 2 pollen (occurring in present day Aristogeitonia, Mischodon, Oldfieldia, Voatamalo) did not evolve until the early to middle Miocene, then dispersal of that lineage from continental Africa to Madagascar and other islands on the eastern side of Africa, and especially to India must have taken place via island hopping and/or long-distance dispersal. It is speculative, but possible that birds transported seeds from the continent, or that seeds along with ants (their primary disperser) and/or even mammals or reptiles, were transported eastwards on floating debris, by one or more successive dispersal events.
Alternatively, the lack of fossil Picrodendraceae pollen from the Eocene to Oligocene of continental Africa may not be because these plants did not thrive in these parts of the world during that time, but more due to the fact that not enough samples have been studied or that they have simply been overlooked or misidentified due to their superficial similarities to e.g. Asteraceae, Malvaceae and Arecaceae. The fossil pollen record to date (Table VII) has been compiled by the Vienna working group, routinely using SEM, and the Australasian palynological community. Palynologists working in Africa may not yet be fully aware of the presence of this pollen group or of its affiliation. This is supported by the current identification of the Miocene Mush MT pollen in the Mush Valley palynoflora (Ethiopia), where this pollen type was most likely previously referred to Myristicaceae (Danehy 2010), and the Saldanha MT (Hyaenanche lineage) pollen in the Saldanha Bay palynoflora (South Africa), a well-studied Miocene flora, where this type of pollen was previously referred to Baumannipollis with affiliation to Malvaceae (Roberts et al. 2017).

Conclusion and outlook
Picrodendraceae produce unique pollen grains that are easily identified using the combination of LM and SEM. The extant Afro-Indian pollen can be grouped into three morphological types (geographical lineages). The PT 1 occurring in Hyaenanche is clearly the basal most surviving Afro-Indian pollen type that gave off subsequent pollen types (PT 2 and PT 3).
Combining evidence from the pollen morphology of Picrodendraceae along with their fossil records, the current phylogenetic framework, a new dated intrafamilial phylogeny, and relevant paleogeographic scenarios from the Late Cretaceous to early Cenozoic, we conclude that: (1) Picrodendraceae originated in the Americas, (2) the family dispersed from Southern America across Antarctica and into Australasia during the late Paleocene, (3) a second migration (of a different lineage) took place from the Americas across the North Atlantic Land Bridge and into Europe during the Paleocene and/or Eocene, (4) the Picrodendraceae reached their maximum north-south distribution prior to the end of the Eocene, (5) the family dispersed into Africa from Europe during the Eocene, (6) and dispersed from continental Africa to islands east of the continent and to India in the Neogene via island hopping or long distance dispersal.
Still, to fully understand the early divergence and evolution of Afro-Indian Picrodendraceae, palynological studies need to focus on Late Cretaceous to early Cenozoic floras from the Americas and Afro-India using the morphological data presented here. This will enable scientists to correctly affiliate pollen of this group to particular lineages and to fill some of the gaps in the family's early history. Also, the expansion of the molecular phylogeny is needed to more completely understand the historical evolutionary relationship of all the Afro-Indian, American and Australasian taxa of this family.