Subspecies of Hypolepis rugosula ( Dennstaedtiaceae ; Pteridophyta ) around the world : morphological and biogeographic perspectives

The “Hypolepis rugosula complex” has been the subject of great debate among pteridologists: some have considered H. rugosula a single subcosmopolitan (or circum-Antarctic) species, whereas others have considered it a species-complex, encompassing several species. In the 1920s and 1930s, four geographically distinct varieties of H. rugosula were recognized. In this work, we present a new taxonomy (with new combinations and statuses, as well as typification and full synonymy), together with complete distribution data for the species, with an infraspecific classification based on morphological and biogeographic perspectives. Hypolepis rugosula occurs in southern temperate regions and high-elevation tropical regions of the Americas, Africa (including Madagascar), Oceania and the Philippines, as well as in some isolated oceanic volcanic islands (e.g., Saint Helena and Tristan da Cunha). Here, 15 geographically distinct subspecies are recognized. All subspecies are geographically segregated from each other, except in New Zealand, where two occur sympatrically—possibly due to two different arrival and colonization times. Four patterns of “indument” (referring to catenate and glandular hairs collectively) are distinguished. Different lineages are successful in their respective habitats; we observed two lineages with different ploidy levels (tetraploid and octoploid). Although long-distance dispersal is the best explanation for the extant distribution of H. rugosula; we do not exclude vicariance as a possible explanation for their occurrence on the land masses that were once united as Gondwana. Therefore, we are assuming that a fern species could remain unchanged for more than 70 Myr, and we are adopting the refugia theory, albeit with a different focus.


Introduction
Hypolepis rugosula (Labill.)J. Sm. was originally described as Polypodium rugosulum by Labillardière (1806), based on a plant collected in van Diemen's Land (now Tasmania) during the Voyage à la Recherche de la La Pérouse (La Pérouse's Research Expedition), made between 1791 and 1793 (Labillardière 1800(Labillardière , 1802(Labillardière , 1806;;Stafleu & Cowan 1979;Brownsey & Chinnock 1984, 1987;Schwartsburd & Prado 2011).Kaulfuss (1824) attributed the concept of cosmopolitanism (or at least circum-Antarctic distribution) to this species, also identifying plants from Chile as Polypodium rugosulum Labill.He was then seconded by Hooker & Arnott (1841), Gay (1853), andBrackenridge (1854).Brackenridge (1854: 18) also stated, ". . .we had some doubts at first, as to the identity of our New Zealand plant with the one from Chili [sic]; but a careful comparison has convinced us that they are one and the same species . . ." This concept was subsequently complemented by other authors, who gradually increased the knowledge of the range of Polypodium rugosulum (= Hypolepis rugosula): Moore (1857) included several names under the synonymy of Pol.rugosulum; Hooker (1862) considered it subcosmopolitan, recording it from Tasmania, Australia, New Zealand, Norfolk Island, India, Java, China, Ceylon, Chile, Ecuador, Peru and the Juan Fernandez Islands, as well as from the islands of Norfolk, Santa Helena, Tristan da Cunha, Bourbon (Réunion) and Fernando Po (Bioko); Mettenius (1858; 1865-as Phegopteris rugosula (Labill.)Fée) also recorded it from Ostindien ("East Indies"-i.e., Southeast Asia); Cordemoy (1891) recorded it for Réunion Island; Dobbie & Crookes (1951) recorded and described it in detail from New Zealand, as well as from South Africa and Japan; Brade (1956-then as Hypolepis rugosula) recorded it from the highlands of southeastern Brazil;Tardieu-Blot (1958-as H. villoso-viscida (Thouars) Tardieu) recorded it from the highlands of Belgian Congo;and Tryon & Tryon (1982-again as H. rugosula) recorded it from "the Americas".Consequently, the name H. rugosula has been adopted by others (Cronk 1989;Agnew & Agnew (code, SI); the Botanical Institute of São Paulo, in São Paulo, Brazil (code, SP); the University of São Paulo Herbarium of Phanerogamae, in São Paulo, Brazil (code, SPF); the Utrecht National Herbarium of the Netherlands, (now) in Leiden, Netherlands (code, U); the Federal University of Paraná, in Curitiba, Brazil (code, UPCB); Uppsala University, in Uppsala, Sweden (code, UPS); the Federal University of Viçosa, in Viçosa, Brazil (code, VIC); the Natural History Museum, in Vienna, Austria (code, W); and the Museum of New Zealand Te Papa Tongarewa, in Wellington, New Zealand (code, WELT).Living specimens of H. rugosula subsp.pradoana were studied in nature, at four different locations in Brazil.The complete list of the material examined appears in Appendix 1.The isolated populations are defined here as subspecies.Below each subspecies listed in the main text, only representative specimens are cited.Drawings were made (Fig. 1).The map shown in Fig. 2 was constructed with Diva-Gis, version 7.5 (http://www.diva-gis.org/download),the coordinates having been estimated with Google Earth (http://www.google.com/earth/download/ge/agree.html).

Results
In analyzing specimens of Hypolepis rugosula from nearly everywhere it occurs, we observed the following: • Hypolepis rugosula is a subcosmopolitan species, with isolated populations throughout the southern temperate regions and tropical regions (at the upper elevations), as well as on isolated oceanic volcanic islands.• Hypolepis rugosula is a unique species characterized by a combination of morphological characters (some almost exclusive) that distinguish it from all other species in the world.• The subspecies are best defined geographically, and none present any striking exclusive morphological character.• There are four well-defined patterns of "indument" (referring to catenate and glandular hairscollectively), which can be noted as tendencies ascribed to the subspecies.• Some morphological characters are linked to local ecological factors and can vary within each subspecies, especially the frond size; the composition and thickness of the lamina; and the indument (thus the divisions among the four indument patterns are sometimes blurred).• There are at least two successful lineages of Hypolepis rugosula: one tetraploid and one octoploid.The ploidy level seems unrelated to the indument pattern.• Long-distance dispersal is the most likely explanation for the extant distribution of Hypolepis rugosula around the world, especially on the volcanic islands.However, we do not exclude the possibility of vicariance among the land masses that were once united as Gondwana.

Characters used in the delineation of the taxa
Morphology was the main basis of the species concept, from macroscopic features (e.g., size of the frond, color of the axis, laminar architecture) to anatomical features (e.g., stele and vascular bundles), microscopic features (e.g., development of the pseudoindusia) and the indument pattern as the finest feature (especially, the types of hair, their sizes, in which lamina region they are present, etc.)This species concept is almost in total accordance with the actual state of art (e.g., Bronwsey & Chinnock 1984;1987;Brownsey 1987;Moran 1995;Mickel & Smith 2004;Schwartsburd 2012b;Schwartsburd & Prado unpublished data).
The sensitive difference of the present paper in relation to those cited above (and others) is that it accepts a morphological plasticity within Hypolepis rugosula-not only among the subspecies but especially within each subspecies.Although the indument is a reliable taxonomic character for most Hypolepis species, it must be carefully analyzed in H. rugosula.Individuals of H. rugosula typically bear both catenate-acicular and catenate-glandular hairs.However, the density of those hairs, as well as their number and position in the laminar regions, and even the co-occurrence of the two types of hair, are sometimes linked to local ecological factors.
Hypolepis rugosula is easily recognized by the color of the main axes (petioles, rachises, and pinna-rachises), which are burgundy (see Schwartsburd 2012b: Fig. 3D).This is almost an exclusive character, shared only with H. distans Hook.(a quite unique species); in all other species, the color is brown, ranging from straw-colored to dark brown (in a few species, they are purple, although only at the base of the petioles).Another feature that is probably exclusive is the vascular bundle of the petioles, which is proximally pi-shaped, and distally split into two S-shaped bundles.The other species present mostly omega-shaped vascular bundles proximally, distally splitting into many bundles (details in Schwartsburd 2012a; Schwartsburd & Prado unpublished data).In addition to these two features, the combination of the following characters serves to distinguish H. rugosula from all other species in the world: fronds small to medium in size-(15.0-)60.0-140.0× (4.0-)20.0-50.0(-60.0)cm, main axes rugose and inermous, proximal pinnae equilateral and distant from the next pair , sori submarginal, pseudoindusia absent (the lamina margins occasionally revolute, protecting the sori, but never differentiated into hyaline flaps), and indument typically composed of catenate-acicular and catenateglandular hairs (plus some brown hairs similar to those of the rhizome present throughout the main axes)-Fig.1, A-R; see also Schwartsburd (2012b: Fig. 2).

Morphological plasticity within an isolated population
Sizes and degree of lamina division.Possibly, the frond sizes and degree of lamina division are related to the length and width of the rhizomes, and thus to the age of the plants.For example, mature fertile fronds of Hypolepis rugosula subsp.pradoana can be found ranging from 15.0 cm long, with pinnate-pinnatisect laminae (Labiak 4778 [UPCB]), up to 95.0 cm long, with bipinnate-bipinnatisect lamina (Schwartsburd 2310 [VIC])-see also Schwartsburd (2012b: Fig. 2).Similar variation can be observed on plants from Tristan da Cunha, for example: fronds 25.0 cm long, with pinnate-pinnatisect lamina (Rogers [K-000369143]); and fronds up to 60.0 cm long, with pinnate-bipinnatisect lamina (Mejland 1158 [K]).
Lamina width and indument.Specimens from caves, or rocky/sheltered sites usually have thinly membranous laminae, almost flaccid, sparsely pilose to glabrous or completely glabrous.This can be seen, for example, in Tristan da Cunha (Rogers [K-000369141]), southeastern Brazil (Brade 15540 [RB]-Fig.1, C and D) and the Juan Fernandez Islands (Sparre s.n.[K], Meyer 9536 [K]).However, specimens from open sunny areas tend to present laminae that are thicker, subchartaceous and much more densely pilose-see also the discussion of Christensen & Skottsberg (1920).

Indument patterns
The four basic indument patterns are easily distinguished, although the divisions between them can be blurred in specimens from sheltered places, which tend to present glabrescent or glabrous laminae, or in juvenile fronds, which tend to be much hairier.Apart from those exceptions, the four indument patterns are constant for most specimens of each subspecies of Hypolepis rugosula: Typical.In the typical pattern, the laminar tissue between the veins is abaxially glabrous or nearly glabrous; laminar margins with catenate-acicular and catenate-glandular hairs sparsely throughout (Schwartsburd 2012b: Fig. 2), or with only catenate-acicular hairs, especially notable near the soral region (Fig. 1, C-H, O and P).This type is found in most subspecies: colorata, poeppigiana, pradoana, pulcherrima, rugosula and subantarctica.
Lacteal.In the lacteal pattern, the laminar tissue between the veins, abaxially, has conspicuous short catenateglandular hairs; laminar margins copiously furnished with only catenate-glandular hairs (Fig. 1K, L).This pattern is found only in the subspecies lactea.
Africana.In the africana pattern, the laminar tissue between the veins is abaxially glabrous; laminar margins glabrous (Fig. 1, M, N and Q-T).This pattern is found in the subspecies africana, brownseyana, pichi-sermolliana, rudis and viscida.
Regarding the Hypolepis rugosula complex, Brownsey & Chinnock (1984) counted n = 52 for two specimens from New Zealand, which present the rufobarbata indument pattern (Brownsey NZ 1435 from North Island) and the typical indument pattern (Brownsey s.n.[WELT-P11518 A and B], from Campbell Island).Following Brownsey (1983, x = 26), those numbers would represent tetraploid lineages (4n = 104).Because the two specimens present two distinct indument patterns, it is reasonable to suggest that the ploidy level is unrelated to the indument pattern.
Although it is too early to draw definitive conclusions, there are apparently at least two successful lineages within the Hypolepis rugosula complex (one tetraploid and one octoploid), and the ploidy level is probably unrelated to the indument pattern.

Biogeographic hypotheses and the adoption of subspecific rank
As many authors have argued (Brownsey 2001;Parris 2001;Wolf et al. 2001;Moran 2008), the distribution of fern taxa can be inferred to have happened by vicariance, dispersal or a combination of the two.Dispersal is the more widely accepted explanation, because ferns famously disperse their microspores by wind.Long-distance dispersal is also well accepted by the scientific community, being a common explanation for the origins of island floras (Tryon 1970;Moran 2008).However, vicariance tends to be avoided as an explanation for phenomena such as the Gondwana relationships.That is primarily due to our limited knowledge of the relationship between extant and extinct genera; often, "dispersal obscures evidence of vicariance" (Wolf et al. 2001).Because vicariance is avoided for explaining the distribution of families and genera, it is almost completely rejected at the species level.
The genus Pteridium is thought to contain only one species (Tryon 1941); two/three species, as well as two hypothetical allotetraploid species (Der et al. 2009;Thomson 2012); or up to a dozen species (Mickel & Smith 2004).There are also many infraspecific classifications that, in the interest of brevity, will not be discussed here.Nevertheless, the genus Pteridium is cosmopolitan and widely distributed throughout its range, and its isolated populations (= morpho-taxa: species, subspecies, or varieties) are interconnected.Therefore, it is reasonable to assume that its current distribution occurred by vicariance (see extensive discussion in Wolf et al. 2001: 267).
What about Microlepia strigosa, Histiopteris incisa and Hypolepis rugosula?Although they do present current subcosmopolitan distributions, their populations are all isolated from each other (Fig. 2).In addition, they seem quite rare in nature (our personal observation throughout Brazil).Certainly, no one has ever seriously considered these distributions to be a result of Gondwana vicariance but rather of long-distance dispersal.Would one species remain unchanged for more than 70 Myr?Or, even if speciation has occurred, would the same "successful design" be maintained unchanged for 70 Myr in "cryptic species"?And furthermore, could three species within the same family share a coincidental "cryptic pattern"?One way to support the vicariance hypothesis for these three species is to invoke and adapt the Haffer-Vanzolini theory of refugia (Haffer 1992).These three Dennstaedtiaceae species currently occur in relatively southern temperate regions, including tropical regions, although their distribution in the latter is mostly limited to the upper elevations-commonly at ≥ 1000 m in subtropical regions, and at ≥ 2000 m in tropical regions (Brade 1956;Mickel & Smith 2004;Schwartsburd & Prado 2011;Schwartsburd 2012b).This is a common biogeographic pattern in which there is a similarity between upper elevations and the upper latitudes.It was conceived by the earliest biogeographers, such as Tournefort and Linnaeus (Papavero et al. 2013: Chap. 6), who termed it "altitudinal climatic zonation".Therefore, even though these species inhabit tropical regions, their occurrence is limited to temperate conditions.With this in mind, we can speculate that these species were more widely distributed during times when the Earth was cooler (e.g., the last ice age) and are now relicts, restricted to cold refugia, as a result of climatic vicariance.Obviously, to interpret the current distribution of these three species as resulting from long-distance dispersal still seems much more reasonable, especially because they occur sympatrically on isolated oceanic volcanic islands (e.g., Tristan da Cunha, Gough and Saint Helena).Given that volcanic islands are "young", their colonization by these species almost certainly occurred through long-distance, the spores traveling thousands of kilometers across the oceans to germinate and establish new populations.These plants seem adapted to long-distance dispersal.How many times such dispersal could have occurred is a question that cannot yet be answered.Did it occur during isolated events, such as the passing of super-hurricanes or El Niño?Do the populations often disperse their spores across the oceans on air masses or via jet streams?It is still too early to draw any conclusions about the origins of their distribution, which could be attributable to a combination of vicariance and dispersal.
Our basic hypothesis is that vicariance, dispersal or both occurred a long time ago and (in the case of dispersal) only a few times, thus providing a sufficient amount of time for the isolated populations of Hypolepis rugosula to start the process of speciation but not enough time for them to actually differentiate into separate species (given that they are still so similar).With this hypothesis in mind, we have recognized the geographically isolated populations as subspecies of H. rugosula.

The New Zealand hypothesis
Apart from the local ecological plasticity of indument, most subspecies of Hypolepis rugosula are quite homogeneous.The only place where two distinct patterns of indument (rufobarbata and lacteal) can be found is in New Zealand (on the North and South Islands, respectively).For New Zealand, we are adopting a secondary hypothesis: H. rugosula arrived there at two different times (on the South Island and on the North Island) and then became two well-defined isolated populations (subspecies), having now encountered each other and bred, creating "hybrids" (Brownsey & Chinnock 1984).

Taxonomic treatment
Hypolepis rugosula (Labill.)Plants terrestrial, humicolous or epipetric.Rhizomes solenostelic and siphonostelic, (1.2-) 2.0-4.0 mm diam., lanose, the hairs catenate-acicular, yellowish to reddishbrown, 1.0-3.0(-4.0)mm long, 15-30-celled.Fronds erect or arched, with continuous growth, (15.0-)60.0-140.0cm long; petioles (6.0-)20.0-50.0cm × (0.8-)1.5-4.0 mm, proximally with one pi-shaped stele, distally with two Sshaped steles, burgundy, inermous, rugose, abaxially and adaxially sparsely villous with three or four kinds of hairs, the first kind of hair catenate-acicular, hyaline with reddish cross-walls, 0.3-0.7 mm long, 5-10-celled, the second catenate-glandular, hyaline with reddish cross-walls, 0.3-0.7 mm long, 5-10-celled, the third similar to those from the rhizomes, sparse and caduceus, the fourth (only in the rufobarbata pattern) catenate-glandular, entirely yellowishbrown, 1.0-2.5 mm long, 10-20-celled, scattered; laminae lanceolate or ovate, proximally pinnate-pinnatisect with segments pinnatifid up to bipinnate-bipinnatisect with pinnatifid segments, medially less decompound, distally pinnatisect, (10.0) 40.0-95.0× (4.0-)20.0-50.0(-60.0)cm, with the two basal pair of pinnae much more distant from each other than from those above; rachises straight, burgundy throughout, or proximally burgundy and straw-colored above, inermous, rugose, the indument similar to the petioles but the hairs more numerous, especially on the adaxial groove; proximal pinnae (2.0-)12.0-30.0× (1.5-)4.0-12.0cm, equilateral; costae abaxially sparsely villous, adaxially copiously villous, the hairs similar to those on the petioles and rachises but the third (and fourth) kind(s) usually absent or very scattered; costules abaxially and adaxially with the first and second kinds of hairs, or rarely with only the first or the second kind, shorter, 0.2-0.5 mm long, 4-to 7-celled; veins with the indument similar to the costules; laminar tissue between the veins abaxially glabrous or nearly glabrous (typical and africana-types), with the first kind of hair (rufobarbata pattern), or with the second kind (lacteal pattern), adaxially with both the first and second kinds of hair, ca.0.2-0.3mm long, 3-or 4-celled; lamina margins glabrous, with the first or second kind of hair, occasionally with both, similar in size to those on the laminar tissue between the veins; sori submarginal, unprotected or slightly protected; pseudoindusia absent, the lamina margin sometimes slightly revolute, protecting the sori, but never differentiated into a hyaline flap.Distribution and ecology: Subcosmopolitan.The populations are segregated from each other (Fig. 2): Tasmania and Australia; New Zealand; the highlands of New Guinea; Luzon (the Philippines); Santa Helena Island; Ascension Island; Madagascar, Réunion Island, and Mauritius; Bioko and the highlands of central Africa (the Democratic Republic of Congo, Ethiopia, Uganda, Kenya, Tanzania and Rwanda); the islands of Tristan da Cunha and Gough; central and southern Argentina and Chile (including the Juan Fernandez Islands); the highlands of southern and southeastern Brazil; the highlands of Costa Rica; and the highlands of Mexico, Jamaica and Hispaniola.At sea level and above in temperate regions; above ca.1000 m in subtropical regions; and above ca.2000 m in tropical regions, having been reported at 3450 m and 3560 m, respectively, in Uganda and New Guinea.

Final conclusions and future prospects
From the extant geographic distribution data, it is possible to comprehend how widely distributed Hypolepis rugosula is, as well as to speculate how it was distributed in ancient times.It is also possible to notice the high number of isolated populations around the world (= subspecies richness).There might also be more subspecies, the herbaria material for which was not accessible, especially throughout other oceanic islands and possibly in other places in the Philippines.
Based on the morphological and geographic data alone, the present paper is unique and important in that it aggregates all of the Hypolepis rugosula subspecies names (under the same species-complex) and provides a detailed account of the distribution.The taxonomy proposed here certainly is not final-it is actually the starting point for future works.
Given that we are entering into a collaboration with Drs. Patrick Brownsey and Leon Perrie from the Te Papa Museum of New Zealand, we strongly encourage future works involving molecular investigation and comparison of Hypolepis rugosula subspecies.Such studies should attempt to answer the following questions: Is there one center of origin of H. rugosula?;Which were (or are) the routes of migration?;Could the distribution in the Americas, Africa and Oceania be attributable to Gondwana vicariance?; Which matrix areas are involved in long-distance dispersal to oceanic islands, and how many such areas are there?;Do the various indument patterns constitute natural subgroups?;Is it possible to speculate on the age of the species and subspecies, and on why speciation is taking so long?; Does the taxonomy presented here best represent H. rugosula?;Are some of the subspecies better classified as species?; and Should varietal level also be incorporated as a third taxonomic level?Cytology will be also very useful to determine which subspecies are tetraploid and which are octoploid.Do they represent two or more consistent and distinct lineages-or are the two ploidies based on random individuals, not representing lineages?Can both occur together in the same isolated population?Are there ploidies other than tetraploid and octoploid?

Dubious records HAWAII-see above in Dubious names.
[SAMOA].Chatam Island, s.d., Travers s.n.(K, W)-these materials resemble Hypolepis rugosula, showing a pattern of indument similar to the lacteal pattern, but are too poorly preserved to be accurately identified.Dobbie & Crookes (1951) also recorded H. rugosula from Chatam, but their concept does not totally agree with ours (they listed it also for Japan, for example).Further collections are needed in order to determine whether H. rugosula does actually occur in Chatam (possibly as an undescribed subspecies) or not.SOUTH AFRICA- Roux (2001) recorded Hypolepis villoso-viscida from South Africa.Although we saw no material of H. rugosula from there, it is likely that it occurs there.Possibly, the best name to apply to this taxon is H. rugosula subsp.africana.
[VENEZUELA].Caracas, s.d., Bredemeyer s.n.(B, W)-these materials clearly belong to the Hypolepis rugosula complex.In the materials at Herbarium W, the locality "Caracas" is overwritten with "St.Helena".According to Schwartsburd (2012a) and Schwartsburd & Prado (unpublished data), H. rugosula does not occur in Venezuela.This collection is most probably from Santa Helena, but since such mislabeling seems so strange, these materials are better left as dubious.