The rediscovery of Passiflora kwangtungensis Merr. (subgenus Decaloba supersection Disemma): a critically endangered Chinese endemic

Abstract Passiflora kwangtungensis is a critically endangered Chinese species known from Guangxi, Guangdong, and Jiangxi Provinces. The species belongs to Passiflora subgenus Decaloba, supersection Disemma, section Octandranthus. Field observations decreased rapidly during the 1970s to 1980s, and it was suspected that this species might have been extirpated due to repeated deforestation events throughout southern China. In recent years, however, small isolated populations of this species have been rediscovered in Hunan Province, representing new locality records for Passiflora kwangtungensis. New herbarium collections, color photographs, and silica gel collections have provided an unexpected opportunity to examine the evolutionary significance of this species. The current study presents a revised morphological description of Passiflora kwangtungensis based on fresh material, along with an updated distribution map. Using nrITS sequence data, preliminary insights into the phylogenetic position of Passiflora kwangtungensis are presented. Molecular data support the placement of Passiflora kwangtungensis within supersection Disemma section Octandranthus. However, the exact placement of Passiflora kwangtungensis within this lineage is unclear. The nrITS data suggest that Passiflora kwangtungensis may be sister to a clade containing Passiflora from China, Nepal, India, and Southeast Asia. Morphologically, Passiflora kwangtungensis displays the most similarity Passiflora geminiflora (Nepal, India) and Passiflora henryi (China). Lastly, conservation status and recommendations are made for Passiflora kwangtungensis following the IUCN Red List Criteria, where this species is classified as CR C1+C2a(i); D.


Introduction
Th e genus Passifl ora L. consists of ca. 526 species (Feuillet and MacDougal 2003) with native ranges throughout the southern United States, Mexico, Central, and South America. In addition, there are 24 species of Passifl ora endemic to the Old World. Th e Old World species are recognized within two subgenera: subgenus Tetrapathea (DC.) Rchb. (Krosnick et al. 2009) and subgenus Decaloba (DC.) Rchb.. Subgenus Tetrapathea consists of three species restricted to Australia, New Zealand, and Papua New Guinea. Subgenus Decaloba supersection Disemma (Labill.) J.M. MacDougal & Feuillet contains the remaining 21 species found throughout Asia, Southeast Asia, and the Austral Pacifi c (Krosnick and Freudenstein 2005). Supersection Disemma contains three sections: section Octandranthus Harms , with 17 Asian and Southeast Asian species, section Disemma (Labill.) J.M. MacDougal & Feuillet , with three Australian endemics, and lastly, the monotypic section Hollrungiella Harms from Papua New Guinea . Th e largest section, Octandranthus, has its center of diversity in China with 13 of the 17 species in this clade distributed there. Th ese species are found in Yunnan, Guangdong, Guangxi, Jiangxi, Hunan, and Hainan Provinces (Wang et al. 2007). Th e native Passifl ora in China exhibit high levels of endemism, seldom display overlapping distributions, and are in general extremely rare. Of those 13 species found in China, Passifl ora wilsonii Hemsl., P. eberhardtii Gagnep., P. jugorum W.W. Sm. and P. tonkinensis W.J. de Wilde are the only species found in surrounding countries as well, and these primarily represent narrow range expansions beyond the borders of China south into Vietnam, or west into Myanmar or India.
Th e Chinese Passifl ora are typically associated with limestone-rich soils and are most often found in wet, sunny openings within subtropical rainforest, along humid forest margins, or among large boulders on moist hillsides. Th ese species generally require primary forests and are rarely found in secondary regrowth or disturbed habitat. Th e Asian Passifl ora are found at elevations from 50 to 2000 meters but are most frequently associated with mid to upper elevations (1000-1500 meters). Population sizes are often quite small, with only a single plant observed over several kilometers (Krosnick 2006). Th is geographical isolation is compounded by the fact that the majority of Passifl ora are self-incompatible (Ulmer and MacDougal 2004), which may eff ectively decrease population size even further. Th eir specialized habitat preferences and limited population size have undoubtedly contributed to the overall rarity of the Chinese Passifl ora.
While not often discussed in the literature, a signifi cant factor aff ecting the distribution of the native Chinese Passifl ora has been deforestation that has occurred within the forests of China over the past 60 years. With the establishment of the People's Republic of China in 1949, country-wide deforestation and forest degradation accelerated rapidly (Zaizhi 2001). Th is was due to intense logging for timber as well as fuelwood needs brought on by several important governmental initiatives (Zaizhi and Chokkalingam 2006). Th e fi rst major degradation episode was from 1958-1961 during the Great Leap Forward and Iron-and-Steel Making campaigns, where communities set up large furnaces to make steel and used primary forest wood to make charcoal to feed these furnaces (Lang 2002, Zaizhi andChokkalingam 2006). Between 1966 and1976, the Great Cultural Revolution and governmental campaigns for self-sufficiency led to more deforestation for cultivation of corn and wheat, as well as additional fuelwood collection (Harkness 1988, Zaizhi andChokkalingam 2006).
Because the Chinese species of Passifl ora require primary forest and undisturbed habitats, deforestation and deterioration of forests throughout the subtropical southern provinces of Guangdong, Guangxi, Yunnan, Hainan, Jianxi, and Hunan would have been especially detrimental to these species. One species that appears to have been vulnerable to the eff ects of rapid deforestation is Passifl ora kwangtungensis Merr.. Th is species, originally described by Merrill in 1934, has since been documented in Guangdong (23 herbarium records), Guangxi (8 records), and Jiangxi Provinces (5 records). Th ese are all provinces that experienced intense deforestation during the 1960's and 1970's. A total of just 35 specimens of P. kwangtungensis were collected between 1924 and 1987, after which point all new collections ceased for this species. No additional collections of P. kwangtungensis were obtained for 13 years, until a single specimen was observed and collected by Ye Huagu ( Ye 3381 , IBSC) in Guangdong Province in 2000. Even with this recent collection, when Krosnick and Deng performed fi eldwork in 2003 visiting all recorded localities for P. kwangtungensis in Guangdong Province, the forest habitats in each location had been cleared or heavily disturbed and the species was not located. At that time, it was assumed that this species was extremely rare, nearing extirpation in Guangdong Province and possibly near extinction throughout its entire range. Fortunately, in 2007, P. kwangtungensis was reported by Yu in Hunan Province, a province where P. kwangtungensis was not previously known to occur. Between 2007 and 2010, Yu observed approximately 14 plants in total across four localities in Hunan Province. His later collection of a single plant in 2010 ( Yu & Tan s.n ., MO) represents a new locality record for P. kwangtungensis , and quite possibly documents one of the last extant individuals of this species.
Th e recent high quality herbarium collection and photographs of fresh material that Yu made of P. kwangtungensis, used in conjunction with herbarium material collected over the last 80 years, allow for the revision of Merrill's original description to more accurately refl ect this species with regard to morphology, ecology, and geographical distribution. Fresh DNA material collected from this specimen provides a new opportunity to examine the phylogenetic position of P. kwangtungensis within supersection Disemma using ITS sequence data. In addition, conservation status assessments and recommendations are made for P. kwangtungensis based on current distribution information according to ICBN criteria.

Field observations
In 2004, botanical fi eld work in Guangdong Province was completed by Krosnick and Deng. All known localities for Passifl ora kwangtungensis in Guangdong were visited based on available herbarium specimen information at the time. Between the years of 2007-2010, Yu and accompanying students conducted fi eld studies in the Nanling Mountains spanning four counties in south Hunan Province: Rucheng (Jiulongjiang National Forestry Park), Shuangpai (Wuxinling Forest Farm), Jingzhou (county nature reserve), and Jiangyong (provincial nature reserve), where they observed ca. 14 individual plants of P. kwangtungensis . Th e greatest number of plants were observed at Rucheng (10 individuals), with just one or two individual plants seen at the Shuangpai, Jingzhou, and Jiangyong locations. Due to the rarity of the species, photos of P. kwangtungensis were taken in lieu of herbarium specimens. A single herbarium specimen was collected in May 2010 from Jiulongjiang National Forestry Park ( Yu & Tan s.n., MO), as a voucher for morphological study and to provide tissue for DNA analysis.

Morphological description
Krosnick (2006) examined 29 herbarium specimens representing material from the major herbaria with strengths in China to create a species description for P. kwangtungensis . Eight additional specimens from IBSC, IBK, and LBG were examined by Deng and Krosnick for the current study, including the 2010 collection of Yu & Tan s.n. (MO). Th us, a total of 37 herbarium specimens from the following herbaria were examined: A, IBK, IBSC, KUN, L, LBG, MO, NY, PE, US. Extensive color photographs accompanying the Yu & Tan s.n . (MO) specimen were used to assist with color details in the species description.

Geographical distribution
As none of the herbarium specimens examined contained primary GPS coordinates, an updated species distribution map was generated by inferring latitude and longitude coordinates using GOOGLE EARTH (Google 2012) or GeoNames Search (National Geospatial-Intelligence Agency 2012). Coordinates were inferred only where locality data was suffi ciently detailed at the level of city, town, or village; thus, only 30 of the 37 herbarium specimens were used for the distribution map. Th ree additional points were added from populations observed directly by Yu during 2007-2010. Appendix 1 includes all herbarium specimen information with inferred latitude and longitude coordinates.

Taxon sampling and outgroup selection
Th e monophyly of supersection Disemma was established using molecular data by Krosnick and Freudenstein (2005) and Krosnick (2006), with three monophyletic sections: Disemma, Octandranthus , and Hollrungiella. However, in those earlier analyses, P. kwangtungensis was not included because fresh material was not available. Krosnick (2006) hypothesized that P. kwangtungensis , once sampled, would fall within section Octandranthus based on morphology and geographical distribution . In the current analysis, supersection Disemma was fully represented with all 21 species currently recognized. Representative species from the following supersections in subgenus

DNA extraction, amplification and sequencing
Total genomic DNA was isolated from fresh leaf material or tissue preserved in silica gel and extracted using the CTAB method (Doyle and Doyle 1987) performed in microcentrifuge tubes, or with the DNeasy Plant Mini kit (Qiagen Inc., Valencia, CA). When necessary, DNA samples were further purifi ed using the Elu-Quik DNA Purifi cation Kit (Whatman Inc., Piscataway, NJ), or the QIAquick PCR Purifi cation Kit (Qiagen Inc., Valencia, CA). Th e nuclear ribosomal internal transcribed spacer region (nrITS) including ITS1, the 5.8S gene, and ITS2, was directly amplifi ed using primers 5 and 4 of White et al. (1990). PCR reaction protocols for ITS followed Krosnick and Freudenstein (2005). Amplifi cations were purifi ed by precipitating with 50 μl of 20% polyethylene glycol-2.5 M NaCl followed by two ethanol precipitations or by using Qiagen PCR Purifi cation Kits (Qiagen Inc., Valencia, CA). Dideoxy cycle sequencing reactions were performed using BigDye Terminator version 3.1 chemistry (Applied Biosystems, Foster City, CA) scaled down to quarter reaction volume. Sequencing reactions were analyzed on an Applied Biosystems 3100 automated sequencer at Th e Ohio State University (Columbus, OH), or at Rancho Santa Ana Botanic Garden (Claremont, CA). Bidirectional sequence contigs were assembled and edited using GENEIOUS Pro v. 5.0.3 (Drummond et al., 2011), or by using SEQUENCHER v. 4.1.1 (Gene Codes Corporation, 2000). All sequences were initially aligned using CLUSTAL W (Th ompson et al. 1994), and manually adjusted using SE-AL (Rambout 2000). See Appendix 2 for complete list of taxa sampled, voucher information, and Genbank accession numbers. Appendix 3 contains the fully aligned dataset for all taxa as a NEXUS matrix fi le.

Phylogenetic analyses
Unweighted Maximum Parsimony (MP) analyses were performed using WINCLADA (Beta) ver. 0.9.9 (Nixon 1999). All characters were treated as non-additive. Heuristic searches were performed using NONA ver. 2 (Goloboff 1999) with the following parameters: 10,000 trees held in memory (hold 10000), 5,000 tree bisection reconnection (TBR) replications (mult*5000), and using two starting trees per replication (hold/2). Resultant trees were summarized with a strict consensus. Branch support for the ITS analysis was assessed using 5,000 jackknife replicates in WINCLADA, with random character removal set at 37%. Th e heuristic searches for jackknife analyses utilized two TBR searches per replication (mult*2), using two starting trees per replicate (hold/2). Only clades with a frequency of 50% or higher were retained in the jackknife consensus. Jackknife support values were mapped directly onto the strict consensus for clades retained in both the jackknife and strict consensus topologies.

Conservation status
Conservation recommendations were made following the ICBN guidelines for application of Red List categories and criteria (IUCN Standards and Petitions Subcommittee 2010).

Phylogenetic analysis
Th e aligned ITS dataset consisted of 801 characters, of which 272 were parsimony informative. Th e heuristic searches resulted in two most parsimonious trees (L=1,070 steps, CI=0.54, RI=0.71). One branch collapsed in the strict consensus of the two MP trees (Fig. 1). Results of the phylogenetic analysis support the monophyly of supersection Disemma, though with low jackknife support (63%; Fig. 1, clade A). Monophyly was strongly supported for supersections Cieca (100%) , Hahniopathanthus (99%) , and Bryonioides (100%). Supersection Decaloba is resolved as polyphyletic in this analysis, with one strongly supported (99%) clade containing P. allantophylla Mast., P. mexicana Juss., P. bifl ora Lam. , P. murucuja L. and P. tulae Urb., a second clade consisting of P. citrina J.M. MacDougal and P. cisnana Harms (100%), and a single unresolved P. fi lipes Benth. Supersection Multifl ora is also polyphyletic, with P. holosericea L. resolved as sis-ter to P. multifl ora L. + supersection Disemma (<50%) , and then P. monadelpha P. Jørg. & Holm-Niels. as sister to P. auriculata Kunth (68%). Within supersection Disemma, P. hollrungii K. Schum. is resolved as sister to the rest of the clade, which consists of two lineages, section Disemma (100%; Fig. 1, clade B), and section Octandranthus (<50%; Fig. 1, clade C). Within Octandranthus, two lineages are well supported: a clade of fi ve species (98%; Fig. 1, clade D), and a second clade with the remaining 12 species (97%; Fig. 1, clade E). Although P. kwangtungensis is resolved as sister to the remaining species in clade E, jackknife support for the position of P. kwangtungensis and P. altebilobata Hemsl. relative to remaining species is <50%. To further explore the placement of P. kwangtungensis as sister to the remainder of clade E, another heuristic search using the same parameters was performed with P. altebilobata removed from the dataset (data not shown). In that analysis, Passifl ora kwangtungensis was still resolved as basal within clade E, suggesting that while jackknife support is low for its placement, the position of P. kwangtungensis was not aff ected by the presence of P. altebilobata . Within the remaining 10 species, two subclades appear: a Southeast Asian clade (100%; Fig. 1, clade F), and a Chinese clade (<50%; Fig. 1, clade G) .

Revised species description
Based on the high quality photographs of living material (Fig. 2) and the additional herbarium specimens incorporated in the present study, a morphological description that more accurately refl ects P. kwangtungensis is presented here. Of particular note are color details that were not visible in the older herbarium specimens. Merrill (1934) suggested that the fl owers of P. kwangtungensis were white throughout, but it is now evident that fl owers in this species have greenish-yellow sepals, whitish petals, an outer corona that is bright yellow in the upper half and yellow-green in the lower half, a yellow-green inner corona, and distinct brown fl ecks along the androgynophore and limen ( Fig. 2A). Merrill also described the fl owers as solitary in the axils of the leaves, but additional examination of herbarium material has revealed highly branched cymose infl orescences with up to 6 fl owers per infl orescence. Th e infl orescences, when observed as they are naturally held on the plant (Fig. 2B-D), show a unique arrangement of third order branches on either side of the tendril (where the peduncle and terminal tendril are designated as fi rst order, sensu Krosnick and Freudenstein 2005). Floral pedicels are of equal length and the terminal second order bud is sometimes absent. Merrill's original description made no mention of fruits in P. kwangtungensis. Although fresh samples were observed while immature, fruits in this species (Fig. 2E) appear similar in shape to other species in section Octandranthus in being relatively small (ca. 1 cm in diameter), globose, and paired.
Th e observation of fresh material has provided additional insights into P. kwangtungensis with regard to vegetative characters. Th e petiole has two paired papillate nectaries near the apex of the petiole (Fig. 2F) . While the original description did note that the leaves are 3-nerved (Fig. 2F-G), the fresh material reveals a unique mottled variegation along the veins in younger leaves. Two distinct leaf shapes, lanceolate and ovate, are observed in the fresh material ( Fig. 2G-I). Th e more lanceolate shape is associated with the juvenile growth form, while the ovate shape is observed on older portions of the plant.
Ecology. Passifl ora kwangtungensis is observed most frequently on hillsides in thickets, along roadsides in forest valleys, or along primary forest margins. Th is species prefers wet, sandy soils, and is scandent along the ground, sometimes climbing onto low shrubs or tree trunks. Elevation ranges from 500-1000 m.

Phylogenetic position of Passiflora kwangtungensis
Based on the strict consensus of the ITS data presented here (Fig. 1), both supersection Disemma (clade A) and section Octandranthus (clade C) are resolved as monophyletic, though with low jackknife support. Within section Octandranthus, P. kwangtungensis is strongly supported as a member of clade E, which consists of species from India, Nepal, China, and Southeast Asia. Passifl ora kwangtungensis and P. altebilobata form a basal grade leading to a Southeast Asian clade (clade F) and a Chinese clade (clade G). However, jackknife support values are quite low for several key nodes within the ITS phylogeny presented here, suggesting that alternative topologies may be obtained as more loci are included. Th erefore, while taxon sampling for supersection Disemma is complete with regard to the ITS dataset, it is not yet possible to make strong conclusions regarding relationships within Disemma or about the phylogenetic position of P. kwangtungensis in section Octandranthus . Th e addition of nuclear and chloroplast sequence data for P. kwangtungensis will allow for more thorough insights into the evolutionary position of this species within section Octandranthus .
Supersection Disemma is a diffi cult lineage to study from a morphological standpoint because there are no clear synapomorphies that distinguish these 21 species as a group from the rest of subgenus Decaloba. Moreover, there seems to be a high rate of character transformation in this lineage, such that even closely related species appear quite distinct with regard to key fl oral and vegetative features. For example, within clade D, P. eberhardtii has the smallest fl owers in the supersection (ca. 1 cm or less in diameter), large cordate leaves with scattered abaxial nectaries, and fl attened petiolar nectaries. Th is species is sister to a clade containing P. jianfengensis S.M. Hwang & Q. Huang, P. jugorum, and P. wilsonii, all of which have fl owers 3 cm or greater in diameter, leaves that are more or less truncate, abaxial nectaries in pairs, and petioles with peg-shaped glands. Similarly, placement of P. kwangtungensis within supersection Disemma is challenging because while this species displays characters that might be considered plesiomorphic for clade E , it also exhibits many morphological similarities (infl orescence structure, fl oral coloration, and petiolar nectary shape) to both P. henryi Hemsl. and P. geminifl ora D. Don, both of which occupy relatively derived positions in clade G. Th us, it is useful to consider the similarities of P. kwangtungensis to the remainder of clade E as a whole ( P. altebilobata + clades F, G), as well the similarities of this species to P. henryi and P. geminifl ora.
Considering fi rst the placement of P. kwangtungensis as basal within clade E, a number of features observed in Passifl ora kwangtungensis could be viewed as plesiomorphic for this clade. Passifl ora kwangtungensis has small fl owers that are generally no larger than 2 cm in diameter. Seven of the 12 species in clade E have small fl owers (less than 2.5 cm in diameter), while the fi ve Southeast Asian species (clade F) have much larger fl owers, generally 3-5 cm in diameter. Larger fl ower size could represent a synapomorphy for clade F, while for the rest of clade E fl owers could have remained small. All species in clade E display infl orescence branching through at least the second order, and all species have at least two fl owers per infl orescence. Branching in P. kwangtungensis may be through the third order, with one to four fl owers per infl orescence. However, there is great variation in the extent of branching across the species in clade E. For example, P. altebilobata has branching through the fourth order and up to 11 fl owers per infl orescence, while infl orescences in P. cupiformis may have up to 18 fl owers. Leaves in P. kwangtungensis range from lanceolate to ovate, simple shapes that could easily be modifi ed to create the various forms observed across the clade. Passifl ora altebilobata (clade E) and P. xishuangbannaensis Krosnick (clade G) are perhaps the most specialized with deeply bilobed leaves, but this shape could be readily achieved through truncation of the midvein if starting from an ovate leaf form. Th e leaves of P. kwangtungensis have submarginal abaxial nectaries, a feature which is observed in all species across clade E. Should its basal position continue to be supported as additional loci are sequenced, the morphological features of P. kwangtungensis described here would be consistent with character traits in the other 11 species in clade E, highlighting the notable morphological plasticity in supersection Disemma .
Alternatively, there are three morphological similarities shared among P. kwangtungensis, P. henryi, and P. geminifl ora that are suggestive of a close relationship between these species. First, Passifl ora kwangtungensis (Fig. 2B-D) , P. henryi, and P. geminifl ora display many similarities with regard to their infl orescence architecture. Th ey all have cymose infl orescences branched through the third or fourth order. Within the infl orescence, third order fl owers sometimes appear to be arranged in pairs, caused when the second order bud is aborted. Th is condition is commonly observed in P. geminifl ora and somewhat less commonly in P. henryi . Pedicels within the infl orescence are of more or less equal lengths and held at the widest angle possible from one another, which results in the infl orescences appearing as mirror images of one another on either side of the central tendril. Th is diff ers from other species in section Octandranthus that have fasciculate infl orescences caused by the presence of sequentially shorter pedicels as branching order increases. Second, P. kwangtungensis ( Fig. 2A) , P. henryi, and P. geminifl ora each exhibit narrow fl ecks of brown coloration ca. 1 mm in length along the androgynophore and limen surface. Th ird, Passifl ora kwangtungensis has papillate to narrowly peg-shaped petiolar nectaries (Fig. 2F), which are also observed in P. henryi and P. geminifl ora. Should additional data resolve P. kwangtungensis with P. henryi and P. geminifl ora, these similarities would represent synapomorphies for that clade. Th ese features are strongly suggestive of an evolutionary connection among the three species, or at the very least, an interesting convergence of form.

Geographical distribution
Passifl ora kwangtungensis was originally described based on two herbarium specimens ( Tsang 20609 holotype , Tso 20749 paratype) collected in Guangdong Province. Even in the original description, Merrill (1934) noted the affi nities between P. geminifl ora (as syn. P. nepalensis Walp.) with P. kwangtungensis. In 1940, Chun, in "Flora of Kwangtung and South-Eastern China, III" noted a new collection of P. kwangtungensis in Guangxi Province. Later, in 1972, De Wilde cited three additional specimens from Guangdong. In 1984, Bao cited two new records for Jiangxi and one for Guangxi. Th e geographical distribution of P. kwangtungensis was cited as Guangdong, Guangxi, and Jiangxi by both Krosnick (2006) and Wang et al. (2007). Th e most recent specimens available (Guangdong: 2000, Ye 3381 , IBSC; 2010, Hunan: Yu & Tan s.n., MO) suggest that this species is still extant, though quite rare, and may be found in a narrow range along the border of Guangdong and Hunan Provinces.
Historically, P. kwangtungensis appears to have been most abundant in Guangdong Province, with 23 of the 37 localities from this province. Given the high number of deforestation events that have occurred in southern China since 1958, it seems plausible that the decreasing numbers of collections each year for P. kwangtungensis was correlated to the abundance of suitable habitat available in its native range. It is possible that these declining collections may simply refl ect a decrease in botanical fi eld work in Guangdong, Guangxi, and Jiangxi Provinces. However, given the gradual decline in numbers of P. kwangtungensis specimens collected from the 1960's through the 1980's and the complete absence of collections after 1987, it seems more likely the result of reduced available habitat for an already rare, obligately out-crossing species being pushed to the brink of extinction throughout its range. Th e two most recent collections made in 2000 and 2010 are along the border of Guangdong and Hunan Province in the Nanling Mountain Range (Fig. 3). Based on the fi eld observations of Yu during 2007-2012, it appears that the ca. 14 individual plants observed in Hunan Province may be some of the last remaining extant individuals of P. kwangtungensis.

Conservation
Under the IUCN Red List guidelines (IUCN Standards and Petitions Subcommittee 2010), Passifl ora kwangtungensis should be classifi ed as CR C1+C2a(i); D , or critically endangered, based on two assessment criteria, C and D. With respect to criterion C, small population size and decline: the number of mature individuals known for Passifl ora kwangtungensis is less than 250 in total, with just 14 plants observed in Hunan over three years of surveys. Within category C, P. kwangtungensis should be classifi ed as C1, an estimated continuing population decline of at least 25% in 3 years or 1 generation, because this species is self-incompatible and exists in extremely fragmented environments which restrict gene fl ow. It is important to note the defi nition of population according to the IUCN (2010) is the total number of individuals in a taxon, rather than the number of individuals at a given location. In a traditional sense, population sizes of P. kwangtungensis are even smaller (ca. 1-2 individuals in three observed populations, 10 maximum for the largest population according to Yu's observations). Given these limitations, it is likely there will be a decline of the total population size over the next generation (which could be 5-10 years based on most Passifl ora species). Within category C, P. kwangtungensis can also be classifi ed as C2, a continuing decline and a(i) number of mature individuals in each subpopulation less than 50. Given that only 36 herbarium collections have been made since 1924 and ca. 14 plants are currently known from Hunan, a realistic estimate of the total population size for P. kwangtungensis would be 50 or fewer, optimistically. Under criterion D, very small or restricted population: P. kwangtungensis should be classifi ed as D, number of mature individuals less than 50. Th is species, previously feared to be extinct throughout its native range, is surviving in isolated pockets along primary forest margins, or quality habitat on undamaged hillsides throughout the Nanling Mountains. Th ere is likely very little gene fl ow among the subpopulations, and even if the species are self-compatible, genetic diversity would be assumed to be quite low due to inbreeding. Fortunately, three of the four locations where P. kwangtungensis was observed are in county, provincial, or national park reserves. Th is gives them some protection from habitat destruction but cannot ensure their survival due to reproductive isolation caused by low population numbers.
While the highest Red List conservation status a species qualifi es for should be used, Passifl ora kwangtungensis would also qualify as endangered under criterion A2abc, where A2 specifi es a ≥50% decline over the longer of 10 years or three generations, and where population reduction was observed or inferred to have occurred in the past and the causes of reduction may not have ceased, may not be reversible, and may not be understood. Th e "abc" is determined based on a, direct observation, b, an index of abundance appropriate to the taxon, and c, a decline in the extent of occurrence and habitat quality. If herbarium specimens are taken as evidence, a clear drop off in the number of collections made occurs from the late 1980s forward. Conservatively, the lower number of specimens collected is assumed to refl ect reduced population numbers, as opposed to reduced collecting eff orts by scientists in the region. As several important fl oristic works focused on China have emerged during the 1980's and 1990's (Chen 1987, Lin 1993, Wu 1984, Wu and Raven 1994, it seems the former explanation is more likely than the latter. Moreover, herbarium specimen records indicate a clear decline in the extent of occurrence in P. kwangtungensis throughout its originally described range; much of this is likely due to decreased habitat availability brought on during the deforestation campaigns spanning the 1950s to 1970s. Passifl ora kwangtungensis does not appear to have ever been truly abundant in its native habitat, at least given evidence from specimens collected from 1924 onward. It may be that population numbers of P. kwangtungensis were reduced below a sustainable size during the deforestation eff orts that occurred mid to late-century, thus setting off the observed decline in the late 1980s and onward. Taken together, the evidence suggests that this species has been extirpated, or is nearing extirpation, from Guangdong, Guangxi, and Jiangxi Provinces and is currently surviving only in isolated pockets of refugial habitat in Hunan. In general, Passifl ora grow quite well from stem cuttings. Both in situ and ex situ conservation methods would be recommended with immediate implementation to protect the remaining individuals of P. kwangtungensis from what seems to be near-certain extinction. Seeds, if produced, should be germinated and maintained in cultivation at local botanical gardens (such as IBSC) where soil type and other ecological factors will be most favorable for their survival. Further exploration is needed in Hunan, Jiangxi, and Guangxi Provinces to see if additional refugial populations still exist; if so, particular eff ort should be placed on cultivation of stem cuttings and eventual cross-pollination with the Hunan material to increase the genetic diversity of the material in cultivation. Th e case of Passifl ora kwangtungensis represents a rare opportunity where botanists have the chance to assist in bringing a plant back from the brink of extinction. We hope that the information presented here will facilitate the protection and conservation of this species. Th is manuscript will also be presented as part of the application for placement of P. kwangtungensis as critically endangered on the IUCN Red List , a recognition that will confer additional protection and increased awareness regarding the status of this species.