Arrival of Passiflora in the Old World. In order to place supersection Disemma within the proper context for biogeographic discussion, some time must be spent on the geographical origins of the Passifloraceae and close relatives Malesherbiaceae and Turneraceae. As many of the genera in Passifloraceae are small and endemic to remote locations, limited information is available for relationships at the familial level. Krosnick et al. (2005) used ndhF sequences to reconstruct relationships within the Passiflorinae and found that these data supported two monophyletic tribes within the Passifloraceae: Paropsieae and Passifloreae. Paropsieae consists of ca. six genera that are found exclusively in Africa. Passifloreae contains ca. 10 genera, four of which (including Passiflora) are endemic to Central and South America. Six of the genera in Passifloreae are African and Asian, including the second largest genus in the family, Adenia, which has ca. 110 species in Africa, Madagascar, India, and Southeast Asia.

The families Malesherbiaceae and Turneraceae are closely related to Passifloraceae, though various analyses provide conflicting resolution of relationships between these families. In some analyses, Malesherbiaceae and Turneraceae are sisters to one another (Chase et al., 2000; Chase et al., 2002; Sosa et al., 2003) and in others, Malesherbiaceae is sister to Passifloraceae plus Turneraceae (Davis et al., 2005; Krosnick et al., 2005). Malesherbiaceae is a small family of just 24 species endemic to Pacific coastal desert regions in Chile, Peru, and Argentina (Gengler-Nowak, 2002). Turneraceae is more widespread, consisting of 10 genera throughout Central and tropical South America and central and southern Africa and Madagascar.

If Malesherbiaceae is sister to Passifloraceae and Turneraceae, Passifloraceae likely originated in South America. Alternatively, if Malesherbiaceae and Turneraceae are sisters, then Passifloraceae could have originated in either Africa or South America. Fossil evidence for these families is scant, but a few unconfirmed records do exist for the Passifloraceae. Rásky (1960) described Passifloraephyllum from the upper Eocene in Hungary based on leaf morphology. Passiflora kirchheimeri and P. heizmannii were described from the Miocene in Eastern Europe and Germany based on seeds (Mai, 1967; Gregor, 1982). An unnamed Passiflora was described from Veracruz, Mexico, during the Pliocene based on pollen (Graham, 1976; Porter-Utley, 2003). A species of Passiflora was described from the Eocene in New Zealand based on a leaf (Pole, 1994).

While Passifloraceae does not have a strong fossil record, the Malpighiales, the order to which Passifloraceae belongs, is well-represented by fossil data. Recently, Davis et al. (2005) used 15 palynofossils and macrofossils to determine the time of divergence for the order and the type of habitat in which they originated. In this analysis, the fossil data were used to calibrate divergence times for a molecular analysis of all major clades within the Malpighiales. They estimated that the Malpighiales originated between the late Aptian to the mid-Albian, depending on the age constrains they used for the basal node within their cladogram (either 109 or 125 Ma). They suggested that 24 of the 29 clades underwent a rapid radiation between 114-89 Ma, with the emergence of Passifloraceae between 96.5 and 113.9 Ma. Interestingly, the age estimates presented in Davis et al. (2005) are much older than those presupposed elsewhere: Davis et al. (2002) suggested that Malpighiaceae itself was 63-70 Ma, then 74-80 Ma in Davis et al. (2004), and finally in Davis et al. (2005), rate estimates suggest both Passifloraceae and Malpighiaceae are of similar ages, being 89.0-113.2 Ma old. Gengler-Nowak (2002) suggested that Malesherbiaceae was only 40 myr old based on the existence of appropriate habitat in South America for the extant species. Taking all of these estimates into consideration, Passifloraceae might have a minimum age somewhere between 40 and 113 Ma.

Passiflora is the most speciose genus in the Passifloraceae, with more than 520 species currently described (Feuillet and MacDougal, 2003). It is probable that the genus converged on a suite of characters, specifically the androgynophore, corona, and extensive floral and extrafloral nectaries, that enabled it to radiate quite rapidly relative to the other genera in Passifloraceae. As already discussed, of all the species in Passiflora, only 24 are endemic to the Old World. Twenty-one species in the Old World belong to supersection Disemma, and three others belong to subgenus Tetrapathaea (see Chapter 3). The question of exactly how Passiflora arrived in the Old World has never been answered. This is due primarily to the fact that no satisfactory phylogeny has existed for the supersection Disemma or for subgenus Tetrapathaea. In Chapter 3, the monotypic genera Tetrapathaea and Hollrungia were recognized as species within Passiflora, which requires that two separate radiations occurred in the Old World. The first radiation would have involved subgenus Tetrapathaea, based on its phylogenetic positionwithin the genus (Chapter 3, Figures 3.2, 3.3). The second radiation (supersection Disemma) took place within a relatively derived clade of subgenus Decaloba (Figure 4.7D).

There are three possible hypotheses that can account for the present-day distribution of Passiflora. First, the Passifloraceae might have been distributed according to the Gondwanan aborigine hypothesis sensu Davis et al. (2004), which would explain the current distribution of the family through vicariance of Western Gondwana. According to this idea, the ancestral distribution of the family could have been in both Africa, where Adenia might have diversified, and Central and South America, where Passiflora would have diversified. The intercontinental connections might have facilitated the migration of Passiflora into the Old World, and then as the continents began to separate, those species in Asia and the Austral Pacific would have become isolated. Both Adenia and Passiflora have a small number of species endemic to the Old World that could be explained by Gondwanan vicariance. While this hypothesis might explain the distribution of Passiflora, it requires the assumption that the genus is very old and had diversified and migrated across Gondwana while the continents were still in close proximity to one another. Western Gondwana (South America and Africa) was already broken apart by 105 Ma. If Passifloraceae itself is between 40 and 113 Ma, it is unlikely that it Passiflora would have had enough time to diversify and reach eastern Gondwana before the distance between South America and Africa was large enough to make dispersal unlikely. Further, if Passiflora was old enough to diversify across Gondwana, it might be expected that some relictual species of Passiflora might be present on the African continent, which is not the case.

The second explanation for the geographical distribution of Passiflora is found in the Boreotropics hypothesis (Wolfe, 1975; Tiffney, 1985; Lavin and Luckow, 1993). Following this hypothesis, the center of origin for Passiflora would have been Central or South America, with the genus then becoming widespread throughout North America and Eurasia via an oceanic land bridge. Two possible routes would have existed for migration across Laurasia, either Beringia across the Pacific, or the North Atlantic Land Bridge (Tiffney, 1985). Beringia was located between 69° and 75°N, and was likely too cold to support tropical-adapted genera (Davis et al., 2002). The North Atlantic, however, was at a thermal maximum between the Eocene/Oligocene (McKenna, 1972; Wolfe, 1975), and the Northern Hemisphere as a whole was at its warmest during the Paleocene/Eocene (Davis et al., 2004). The North Atlantic Land Bridge present during this period was only 45°N, and was thought to be warm enough to support broad-leaved evergreen plants (Tiffney and Manchester, 2001). Given the possible age of Passifloraceae, this is certainly a viable route through which migration could have occurred. As global temperatures dropped in the Oligocene, species might have been extirpated from colder climates, expanding their ranges into warmer regions farther south.

The North Atlantic Land Bridge hypothesis is an attractive explanation for the distribution of Passiflora for three reasons. First, the presence of Passiflora-like fossils in Central and Eastern Europe, while unconfirmed, might suggest that this genus was present in the once-warmer climates of Laurasia. The dates of these fossils range from the Eocene through the Pliocene (54-1.8 Ma), and if they are truly representatives of the Passifloraceae, the minimum age of 54 Ma based on these fossils is also within the 113-40 Ma conservative estimate for Passifloraceae. Secondly, at least two other families within the Malpighiales exhibit a similar biogeographic pattern. Malpighiaceae has a nearly identical geographic distribution to the Passifloraceae (Davis et al., 2002; Davis et al., 2004; Davis et al., 2005) and the minimum age of 113-63 Ma for the Malpighiaceae is congruent with the age of Passifloraceae. Achariaceae is another closely related family to Passifloraceae (Bernhard and Endress, 1999; Krosnick et al., 2005), and the tribe Pangieae (Achariaceae) has a similar boreotropical distribution to Passiflora (Sosa et al., 2003). The genus Chiangiodendron is the only New World member of an Asian and Australian lineage. Several other genera have similar boreotropical patterns, in particular, Hedyosmum in Chloranthaceae (Todzia, 1988), Exacum in Gentianaceae (Yuan et al., 2005), and Styrax section Valvatae (Fritsch, 2001).

Lastly, the wide time span for the North Atlantic Land Bridge hypothesis would be plausible in that it allows for multiple migrations from South America to Asia and Australia. Because there are two unrelated clades endemic to the Old World in Passiflora, any hypothesis of their biogeographic history must accommodate multiple radiations to the Old World. The timeframe for the Gondwanan hypothesis would provide a short window of only eight million years (113 Ma as oldest date for origin of Passifloraceae, 105 Ma before Western Gondwana is no longer connected) for the evolution and expansion of the family. The timeframe postulated in the North American Land Bridge hypothesis allows a window for evolution and expansion of 33 my, which would be more favorable for multiple radiations.

The third explanation for the distribution of Passiflora is simply that the genus was widespread throughout South America, Laurasia, Southeast Asia and Australia during the late Cretaceous and early Paleocene periods. The opening of the Tethys Seaway caused a warming of the global climate between five and eight degrees (Fluteau, 2003; Jenkyns, 2003), which would have enabled the tropically adapted genus to become widespread throughout northern latitudes. The earth’s climate began cooling during the last 45 Ma (Mosbrugger et al., 2005), causing extinctions within the more widely distributed taxa and restricting species to warmer habitats such as those in tropical and subtropical regions of Asia, Australia, and Southeast Asia. This hypothesis is the simplest as it requires the least number of ad hoc hypotheses about dispersal and vicariance events for Passiflora. It also accommodates the distribution of the entire Passifloraceae, which has a significant tropical African component (all of tribe Paropsieae and part of Passifloreae). This hypothesis is closely related and congruent with the Boreotropical hypothesis and the North Atlantic Land Bridge; however, it does not require the dispersal of Passiflora from the New to Old World. According to this idea, the current distribution of supersection Disemma is relictual, based primarily on habitat availability.

Radiation of supersection Disemma throughout Southeast Asia and the Austral-Pacific Region. Southeast Asia originated through pre-Cenozoic break up of Gondwana, the fragments of which eventually collided with Asia, forming Sundaland (Hall, 2001). Australia and India separated from Gondwana during the Cretaceous, and moved northwards. India collided with the Asian continent ca. 50 Ma (Lee and Lawver, 1995), and during the last 25 million years Australia and Papua New Guinea collided with Sundaland (Li and Powell, 2001). Within Southeast Asia is Wallacea, a biogeographic region of transition with elements of both Australian and Asiatic floras and faunas, but with exceptionally high degrees of endemism (Hall, 2001). There have been several lines drawn between the Australian and Asian biogeographic areas, e.g. Wallace’s Line, Weber’s Line, and Lydekker’s Line (Metcalfe, 1996). Interestingly, the distribution of plants versus that of animals is dissimilar across Wallacea; while faunal differences are well characterized as either purely Oriental or Australo-Papuan, the distribution of plants is ubiquitous and has been viewed as a continuous ecoregion (Erdelen, 2001).

Given that the distance between the Asiatic and Australian elements was prohibitively large until the Oligocene, the question of how migrations occurred across Southeast Asia and the Austral-Pacific must be answered either of two ways. First, it could be explained by relatively recent events, i.e. within the last 35 Ma, or by an older, widespread distribution that contracted resulting in a relictual distribution of extant species. The hypothesis of recent events will be examined first. If the North Atlantic Land Bridge hypothesis is adopted to explain the distribution of Passiflora in the Old World, a recent radiation of the genus in the Old World is congruent with the developing connections between Asia, Australia, and New Guinea. The first migration event of Passiflora would have involved subgenus Tetrapathaea, which is distributed in New Zealand, Eastern New Guinea, and tropical Queensland, Australia.

The phylogenetic position of P. tetrandra as sister to P. aurantioides and P. kuranda suggests that their ancestral geographical distributions would have been within close proximity of one another. Sanmartin and Ronquist (2004) suggested that close affinities between New Zealand and Australia are best explained by trans-Tasman Sea dispersal events. The Tasman Sea opened at around 85 Ma (Li and Powell, 2001), and New Zealand began its rotation northeastward towards Australia. New Zealand was in close proximity to Australia by the Eocene (55-37 Ma; Hall, 2001), which is congruent with both fossil and rate-estimated ages for Passifloraceae. Further, the seeds of Passiflora are dispersed by birds and large mammals (MacDougal, 1994; Ulmer and MacDougal, 2004) which could have been favorable for long distance dispersal between Australia, New Guinea, and New Zealand. Based on these assumptions, the phylogenetic position and geographical distribution of subgenus Tetrapathaea could constrain the evolution of supersection Disemma at an upper age of 55 Ma.

Supersection Disemma is resolved into two major clades, an Austral-Pacific clade including section Hollrungiella and Disemma, and an Asiatic clade Octandranthus (Figure 4.8A-B). As these two clades are sisters, they must be of equivalent ages to one another. Furthermore, this requires that the ancestor to supersection Disemma diverged in two directions, one leading to Austral-Pacific, and one toward mainland Asia and Southeast Asia (Figure 4.8). De Jong (2001) noted that migration from Asia to Australia is much more common than the reverse scenario. However, he suggested that north-south migratory events would not have been possible at all before 25 Ma, as this was when the first land emerged between mainland Asia and Australia in East Sulawesi. The area of exposed land between Asia and Australia was at its greatest point between 15 and 5 Ma.

Section Octandranthus has four major lineages that are primarily found on mainland Asia. Area two (Figure 4.8C) is interesting because of the disjunct distribution of P. geminiflora in Nepal and northeast India with P. cupiformis and P. henryi in Yunnan and Guangxi Provinces, China. Sister to the rest of section Octandranthus is P. papilio, which is also endemic to Guangxi Province. Section Octandranthus could have been primitively distributed across India, South China, Myanmar, North Vietnam and Laos in regions two, three, and four (Figure 4.8C). The presence of two widespread species in region four, P. wilsonii and P. jugorum, suggests that movement across this region was unrestricted. Further, a member of the same clade, P. leschenaultii, is found in south India. This species may have been widespread across India when the temperatures were more tropical, and then extirpated in central India as the climate became drier (Briggs, 1989; Ghosh et al., 1995; Briggs, 2003). The collision of India with south Tibet began around 50 Ma, and continued through 20 Ma (Lee and Lawver, 1995). The separation of P. geminiflora from the rest of its clade is suggestive of vicariance caused by the uplift of the Himalayas. Other plant species have a similar distribution (Yuan et al., 2005). This disjunction might also support a recent time frame for the evolution of Disemma of 50 Ma.

Areas four and five (Figure 4.8C) show a connection between mainland China and Southeast Asia. Area four extends into Southeast Asia because of P. wilsonii, which is perhaps the most widespread species in supersection Disemma. The species in area five (Southeast Asia) belong to clade U (Figures 4.3, 4.4), one of the most terminal clades within the cladogram, implying that they are one of the most recent lineages to emerge within the Old World species. If their date of emergence is more recent (25 to 5 Ma), the Southeast Asian species might have moved across the region using the emerging islands within the region as stepping stones (de Jong, 2001). This pattern of movement might have accelerated evolution within this clade due to factors such as geographical isolation, long distance dispersal, and founders effects as they moved from island to island. This hypothesis might explain the longer branches observed in this clade relative to those in the sister lineage to clade U, clade R.

The Austral-Pacific lineage in area one (Figure 4.8C) resolves P. hollrungii from southeastern Papua New Guinea as sister to the Australian species, indicating that the ancestor to this lineage came from the north and moved southward to Australia. A recent stepping stone diversification across Southeast Asia into New Guinea and then into Australia seems plausible, but the shorter length of the branches in clade I (Figures 4.3-4.5) seems to suggest a more direct route to this region via long-distance dispersal. The latter hypothesis is unlikely given the extreme distance between Asia and Australia that persisted until at least 30 Ma. However, the history of New Guinea is especially complex; it is a composite region that contains several pieces of island arcs (De Boer, 1995; Ladiges et al., 2003). It is possible that the ancestor to sections Hollrungiella and Disemma was distributed throughout Southeast Asia into Wallacea, and then joined with New Guinea as arcs collided. De Jong (2001) explained a similar distribution of Taractrocera, a genus of butterflies distributed throughout Asia and Australia, by suggesting that these species lived in the lowlands of the dry Arafura Sea. They then reached New Guinea and were extirpated from the Arafura Sea when the region resubmerged. This hypothesis necessitates a much more recent time frame for divergence of only 10 Ma, which may or may not be congruent with patterns observed in Passifloraceae.

A second interesting aspect of the distribution of supersection Disemma in the Austral-Pacific is evident through the resolution of phylogenetic relationships within section Disemma. Passiflora cinnabarina is resolved as sister to P. aurantia and P. herbertiana. Passiflora cinnabarina is fire-adapted (Ellison, 1999), and grows in coastal scrublands in New South Wales and Victoria Provinces of southern Australia. Passiflora herbertiana is endemic to forest margins of New South Wales, Lord Howe Island, and Queensland. Passiflora aurantia is a rainforest gap specialist endemic to Queensland, Papua New Guinea, and Samoa. The geographical distribution of section Disemma is counterintuitive; a relationship where P. aurantia is sister P. cinnabarina and P. herbertiana might be expected based the geographical ranges of these species, and the distribution of P. hollrungii in Papua New Guinea.Instead, P. cinnabarina is sister to P. herberitiana and P. aurantia. Crisp et al. (1995) examined the distribution of several Australian and New Guinean endemic plant species for congruence. Interestingly, they recovered area cladograms that supported a relationship of Papua New Guinea to the monsoon regions of Australia (The Kimberley, Arnhem Land, Cape York Peninsula, and Northern Territory). This clade was then sister to a clade that resolved the Southwest group (southwest region, Western Australia) as sister to an East/South group (Atherton, East Queensland, Adelaide, Victoria, Southeast New South Wales, McPherson-Macleay). Their area cladogram resolves Atherton as sister to a clade in which Victoria is sister to New South Wales. While their area cladograms are not directly congruent with the distribution of section Disemma, they do illustrate a fundamental separation in the evolutionary histories of the Papua New Guinea and East/South Coast regions of Australia. They concluded that a monsoon climate might have been present in northern and central Australia until the mid-Tertiary. Based on these data, the ancestor to section Disemma could have had a much broader distribution, which would have contracted as the arid zone of central Australia expanded during the Miocene and Pliocene (Crisp et al., 1995). Further, their area cladogram supported a closer relationship between the East/South Coast region and the Southwest group. Therefore, the diversification of Passiflora across Australia need not have followed a north-south pattern leading directly from New Guinea downwards along the east coast.

The alternative explanation for the biogeographical history of supersection Disemma is one already mentioned, where an older widespread distribution contracted, leaving the current relictual distribution observed today. This explanation is attractive because it is congruent with the topology supported by the combined dataset as well as the alternative topology found in some of t he trees in the molecular dataset. The molecular data resolve P. hollrungii as sister to section Disemma in one topology, i.e., the same result recovered in the combined analysis of molecular and morphological data. The other topology resolved by molecular data alone places P. hollrungii as sister to a clade containing Disemma plus Octandranthus. If the second topology represents the actual relationships in supersection Disemma, the ancestor to this lineage could have been located in either New Guinea or Australia, and diversification ould have moved from south to north. While movement in this direction is rare (de Jong, 2001), distributions such as the one postulated in the alternative topology could also be explained by a once widespread distribution becoming restricted, resulting in a relictual distribution of species that may or may not correlate with the geological history of the region.

The ideas presented here on the historical biogeography of Disemma are intended to be a first attempt at formulating hypotheses about the diversification of this lineage. Without strong fossil evidence, none of these hypotheses are verifiable. Having accurate minimum dates for the emergence of Passifloraceae is absolutely critical for synchronizing the rates of divergence within Passiflora supersection Disemma (Heads, 2005). It is possible that the current distribution of supersection Disemma might be explained by an ancestral Gondwanan pattern. However, fossil evidence and geographical patterns observed in the Malpighiales, the Malpighiaceae, and Achariaceae argue for a much more recent origin for the family. Obtaining an accurate time frame for the origination of Passifloraceae affects the entire evolutionary history of the family; if the date of origin for the Passifloraceae is shown to be later than the minimum age of 113 Ma, a Gondwanan distribution would become an equally viable explanation of the diversification of the family.

Supersection Disemma is most closely related to species in Central and northern South America. Independent of how Disemma arrived in the New World, species in Disemma continued to evolve as the geology of the Australasian region changed beneath them. Because of this, it is possible that some portion of the phylogeny of supersection Disemma is incongruous with the geographical history of the region. The extreme morphological variation observed within supersection Disemma is likely due to extended periods of isolation and diversification throughout a rapidly changing landscape. The hypotheses presented in this section will be tested as more fossils are identified and when those already identified are verified for their phylogenetic affinities. Efforts are underway to attain a well-resolved phylogeny for the Passifloraceae, which will help greatly in calibration of dates of divergence for Passiflora (R. Yockteng, Muséum National D’Histoire Naturelle, pers. comm.). Only then will it be possible to obtain a well-founded understanding of the evolutionary and biogeographic history of this diverse group of plants.

Material from: