Gymnosiphon syceorosensis (Burmanniaceae), the second new species for the Philippines

Abstract A new holomycoheterotrophic member of Burmanniaceae, Gymnosiphon syceorosensis, is described from Mt. Hamiguitan located on the island of Mindanao, Philippines. This species differs from the recently named G. philippinensis from Cebu in a number of quantitative and qualitative characters. Phenetic (neighbor-joining) and phylogenetic (maximum parsimony) analyses of characters from Asian and Australian Gymnosiphon species were conducted and diagnostic taxonomic features were discussed. This new species appears to be most closely related to G. affinis J.J. Sm. from New Guinea but differs in a number of floral features including inner perianth lobe shape, stamen position in floral tube, and anther connective shape.


Introduction
Mycoheterotrophs are plants that obtain nutrients from mycorrhizal fungi that are attached to the roots of vascular plants as well as from saprophytic fungi (Leake 1994). This trophic form occurs in ten angiosperm families, some of which are green and photosythetic and are called partial mycoheterotrophs whereas others have little or no photosynthetic activity and are called full mycoheterotrophs (Merckx et al. 2013). These conditions are analagous to hemi-and holoparasitic angiosperms. Because both life forms can be found in various genera of Burmanniaceae (Dioscoreales) and even among different species of Burmannia, this family provide an opportunity to study the evolutionary origins of these trophic modes. Among the eight genera in the family (Merckx et al. 2013; Burmanniaceae s. str., i.e. without five genera now placed in Thismiaceae), Campylosiphon Benth., Burmannia L. and Gymnosiphon Blume have both Neo-and Paleotropical members. The holomycoheterotrophic genus Gymnosiphon was listed by Merckx et al. (2013) as containing 16 Neotropical, 8 African + Madagascan, and 9 Asian species. Since that publication, two species have been added: G. philippinensis Pelser, Salares & Barcelona (Philippines) and G. queenslandicus Gray, Mahyuni & Low (Australia) bringing the total number of species to 35 (Gray et al. 2019;Pelser et al. 2019). As pointed out by Pelser et al. (2019), most Asian species are rarely collected, reflecting either true rarity or the fact that these plants are often overlooked.
Field work on the Philippine island of Mindanao was conducted during June 2019 as part of a project funded by the National Science Foundation entitled "Plant Discovery in the Southern Philippines". One excursion included the Mount Hamiguitan Range Wildlife Sanctuary, a UNESCO World Heritage site that contains many Philippine endemic and endangered plant species such as Nepenthes copelandii Merr. ex Macfarl., Paphiopedilum adductum Asher, Rhododendron kochii Stein, and Shorea polysperma Merr. (Amoroso and Aspiras 2011). During the course of general collecting, specimens initially identified as Burmannia were obtained. Part of this collection was later determined to be Gymnosiphon by the presence of an unwinged ovary with prominent locular glands, parietal placentation, and a deciduous perianth limb. Merrill (1924, p. 251) excluded G. aphyllus from the Philippines and no other species were listed for the archipelago. The recently named G. philippinensis Pelser, Salares & Barcelona (Pelser et al. 2019), collected on limestone substrate in southern Cebu, is apparently the first documentation of this genus in the Philippines (Pelser et al. 2011 onwards). Those authors concluded that this taxon represented a new species based on a manual comparison to the morphologies of other Asian Gymnosiphon species. A similar method was used in Gray et al. (2019) for G. queenslandicus Gray, Mahyuni & Low. The present study presents results from cladistic and phenetic analyses that were used to assist the process of determining the distinctiveness of the Hamiguitan taxon as compared to previously described species. Moreover, these data also provided information about species boundaries in Gymnosiphon section Gymnosiphon Urb. This was deemed necessary given the different taxonomic concepts for Asian and Australian Gymnosiphon species as published by different authors over the past century (Fig. 1). The matrices constructed included continuous (quantitative) and categorical (qualitative) characters and these were analyzed separately and in a concatenated matrix. The separate analyses were conducted to determine if similar relationships emerged no matter what the partition or method of analysis. Ideally, these morphological data should be examined in the context of a molecular phylogeny of the species, but the only published studies of this type (Merckx et al. 2006(Merckx et al. , 2008 sampled just one Asian species (G. aphyllus Blume) among the 11 accessions included. Names shaded with the same color represent synonyms of the same species according to Jonker (1938). The two taxa named by Tuyama (1940) were not included in this study. The synonymy of G. nana (nanus) is based upon World Plants Online (2020).

Field work
Flowering individuals of the Mt. Hamiguitan Gymnosiphon taxon were photographed in situ. Collections were dried and pressed as herbarium vouchers (no. 1314) and tissue was dried in silica gel for later DNA extraction and sequencing. A few individuals were also placed in bottles containing 70% ethanol for later examination. Dissection and photography of the fixed tissues was accomplished with an Olympus SZH-10 stereomicroscope fitted with a Leica MC190HD digital camera.

Character scoring
For comparisons with other Asian and Australian Gymnosiphon species, descriptions and illustrations from the primary literature were examined. The taxa G. nanus (Fukuy. & T.Suzuki) Tuyama from Orchid Island and G. okamotoi Tuyama from Republic of Palau were not included because their protologues were obtained after manuscript submission. From these a list of characters that appeared taxonomically useful was compiled. The original observations for this study as well as information from the literature were compiled in an Excel spreadsheet (Suppl. material 1: Table S1). These were mainly from (Jonker 1938(Jonker , 1948 and Schlechter (1913) as well as the descriptions of G. aphyllus by Beccari (1878) and Smith (1922). Three species (G. minahassae Schltr., G. oliganthus Schltr., and G. pauciflorus Schltr.) were only collected once thus the taxon descriptions of Schlechter and Jonker (who examined the same specimens) were consolidated. For G. affinis J.J.Sm., G. aphyllus Blume and G. papuanus Becc., the descriptions and taxonomic views differed among Blume, Schlechter and Jonker, thus, they were considered separately (see Fig. 1).

Analyses
Sizes reported in the literature as ranges were converted to median values. Means were calculated from original observations from the Hamiguitan samples as well as measurements taken from the BGBM photographs. Data matrices containing the untransformed data were constructed in Mesquite (Maddison and Maddison 2018) and exported as Nexus and TNT (Tree Analysis using New Technology) files for downstream analyses (Goloboff et al. 2008; Suppl. material 3: File S3). The categorical characters were used "as is" in later analyses whereas the continuous character mean values were natural logtransformed [ln(x+1)] and range-standardized [x s = (x -min/max -min) × 10] as outlined in Thiele (1993) with Microsoft Excel. All characters were treated as unordered.
Uncorrected distances for the transformed continuous character matrices were generated using Mesquite. Neighbor-joining (NJ) was performed separately on this matrix and the "as is" categorical character matrix using PAUP* (Swofford 2002). Maximum parsimony (MP) analyses of the categorical data were conducted with PAUP*. The neotropical species Gymnosiphon suaveolens was chosen as the outgroup for all analyses because ancestral area analyses suggests the genus originated in the New World and took a boreotropical migration route to the Old World (Merckx et al. 2008). MP analyses of the continuous and concatenated (continuous plus categorical) data matrices were conducted with TNT. The log-transformed standardized continuous characters were optimized as additive with Farris (1970) optimization. The search routine as implemented in "aquickie.run" finds optimal scores 20 times independently by using defaults of "xmult" plus 10 cycles of tree-drifting (Goloboff 1999). For strict consensus calculation, TBR (tree bisection reconnection) collapsing was used (Goloboff and Farris 2001). The direction and magnitude of change for the continuous characters was determined by using the TNT command blength for each of the 11 characters. Bremer (1994) support values (decay indices) were calculated which represent the difference (number of steps) between the score of the most parsimonious tree and the next most parsimonious tree where the node in question is lost.

Results and discussion
The MP strict consensus tree from the concatenated continuous and categorical data matrices analyzed with TNT (Fig. 2) contains clades with varying degrees of support as measured by Bremer decay index values. The four Gymnosiphon aphyllus terminals are present as a grade at the base of the tree. A well-supported clade (Bremer decay index > 5 steps) is composed of G. aphyllus (G. borneensis Beccari) and all remaining species. A clade composed of G. minahassae, G. oliganthus, G. pauciflorus and the two G. papuanus terminals is present, but with a Bremer index of < 1 step. Bremer support was higher (> 3 steps) for the sister relationship between G. affinis (G. torricellensis Schlechter) and G. philippinensis. This clade was then sister to G. queenslandicus, but with lower Bremer support. Finally, a well-supported clade (> 5 steps) was recovered containing G. neglectus, G. affinis Jonker and Gymnosiphon sp. 1314.
The above results can be compared to those obtained when the continuous and categorical characters are analyzed separately using phenetic and MP methods (Suppl. material 4: File S4). A number of relationships shown in Fig. 2 are recovered as clusters (Suppl. material 4: File S4A) or clades (Suppl. material 4: File S4B) when the continuous characters are analyzed separately. These include the grade of G. aphyllus, the clade of G. papuanus Becc. and G. pauciflorus, the clade G. affinis (G. torricellensis) and G. philippinensis, and the clade of the new taxon Gymnosiphon sp. 1314 with G. affinis J.J. Sm. and G. neglectus. The taxa G. minahassae, G. oliganthus, and G. papuanus (G. celebicum Schlechter) emerged as a grade in the TNT analysis ( Fig. 2) but as a cluster or clade (Suppl. material 4: File S4A, B, respectively) when continuous characters were analyzed alone. Analysis of the categorical characters separately (Suppl. material 4: File S4C, D) resulted in topologies that differed substantially from the TNT results (Fig. 2).
Overall, it appears that a greater contribution to the tree shown in Fig. 2 came from the continuous, not categorical characters. Both NJ and MP analyses of this data partition recovered similar groupings as compared with the concatenated dataset analyzed with MP in TNT. When apomorphies are plotted on the strict consensus tree, the majority (60%) are autapomorphic (confined to the terminals). Of the 66 synapomorphies, only 11 are categorical characters. This study demonstrates that greater resolution can be obtained by including continuous characters, which has been shown in empirical studies (e.g. Hardy et al. 2008) as well as simulations (Parins-Fukuchi 2017).
The type species for the genus, Gymnosiphon aphyllus, may be the earliest diverging member among the Asian species. The variation in descriptive data (Suppl. material 1: Table S1) probably results from different taxonomic terminology and interpretation of the morphology as well as real polymorphism that exists among populations of this widespread taxon. The first explanation can be demonstrated by comparing the descriptions of Schlechter and Jonker who both examined the same specimens. The four G. aphyllus terminals did not form a clade with MP or a cluster with NJ in any of the analyses. Jonker (1938) lumped G. borneensis Beccari and G. pedicellatus Schlechter into G. aphyllus Blume (Fig. 1). Because of their weak association, further study G. papuanus Becc. (G. celebicum Schlechter 1913) G. papuanus Becc. (Jonker 1938) G. minahassae Schlechter

G. aphyllus Blume
of original material from all of these taxa is required to justify any decision regarding lumping or splitting. Because species boundaries cannot easily be determined from morphology alone, a molecular phylogenetic study of all these taxa is required.
The clade composed of G. minhassae, G. oliganthus, G. papuanus and G. pauciflorus (Fig. 2) has nearly the same composition as the group formed from NJ of continuous characters (Suppl. material 4: File S4A), with the exception that the latter contains G. queenslandicus. Gymnosiphon pauciflorus is sister to G. papuanus and that clade sister to G. papuanus Becc. that was considered G. celebicum by Schlechter (Fig. 2). To avoid a paraphyletic G. papuanus, one could lump G. pauciflorus into G. papuanus or recognize three species. Gymnosiphon pauciflorus shares several features with G. papuanus (Suppl. material 1: Table S1), including characters states such as entire outer perianth lobe margins, linear inner perianth lobes, the position of stamens in the floral tube and forked connectives. Differences that were used in the key by Jonker include the number of flowers in the inflorescence, a meristic character not used here because of extreme variation and overlap. For these two taxa, Jonker (1938) indicates "3-many" for G. papuanus and "1-3" for G. pauciflorus. Because the latter was collected only once (Schlechter 16653), this taxon could represent a few-flowered variant of G. papuanus. Interestingly, the species nearest to G. pauciflorus in the Jonker key was G. neglectus that occurs in a distant clade in Fig. 2. Jonker (1938) combined G. torricellensis with G. affinis, describing the type of the former as "incomplete material but very probably belonging to this species." With reference to G. torricellensis, Schlechter (1913) wrote (translated from German): "Of all the species hitherto known from the monsoon area, the present one is well differentiated by the broad, slightly three-lobed petals, and by the anthers." The description in that work was complete (Suppl. material 1: Table S1) and analysis of this taxon with the description of G. affinis from Jonker (1938) results in the two being present in two different clades (Fig. 2). This result was consistent across all partitions and analytical methods. The two taxa differ in several taxonomic characters including outer perianth lobe margin, inner perianth lobe shape and apex, position of stamen in floral tube, connective shape, and ratio of floral tube to outer perianth lobe length. For this reason, it seems prudent to maintain these two taxa as different species. The taxon Gymnosiphon okamotoi Tuyama was not included in these analyses; however, after examining its description and illustration (Tuyama 1940), it is clear this species has strong affinity with G. affinis and may even be conspecific with it.
The results of the present study agree with the assessment by Gray et al. (2019) that Gymnosiphon queenslandicus is closely related to G. affinis s. lat., which is reflected with the categorical but not the continuous characters. Although G. torricellensis occurs together with G. philippinensis and G. queenslandicus, Bremer support for that clade is relative low. Gray et al. (2019), who examined one of the type collections of G. affinis (Versteeg 1425) noted that the position of the stamens was illustrated incorrectly in Smith (1909) and that they actually occur below the middle of the floral tube. This character state also occurs in G. neglectus and G. philippinensis but not G. torricellensis and Gymnosiphon sp. 1314. Whether this feature changes through floral developmental (bud through anthesis) should be investigated, although for G. affinis, Jonker (1938 p. 31) says that even in very young buds there is still "lowly" insertion of the stamens.
Gymnosiphon sp. 1314 is clearly not conspecific with G. philippinensis because it differs in many character states such as outer perianth lobe margin, flower color, inner perianth lobe shape, position of the stamens in the floral tube, and length of the floral tube relative to the outer perianth lobe. NJ and MP of the continuous characters and the combined data analyses place it as sister to G. affinis with good Bremer support, but this relationship was not seen with analyses of the categorical characters. The unique combination of character states justifies describing the Hamiguitan taxon as a new species. Diagnosis. Similar to G. affinis J.J. Sm. s. str. but differing in the outer perianth lobe color (white and violet vs. pure white), inner perianth lobe shape (cuneate vs. obovate), stamen position in floral tube (just below inner lobe vs. below middle of perianth), connective shape (elliptical vs. quadrangular), and connective apex (not apiculate vs. apiculate).
Distribution, habitat, and conservation. Gymnosiphon syceorosensis is only known from the type collected in the tropical upper montane rainforest of Mt. Hamiguitan, Mindanao. The plant was found along the trail at 1184 m elevation, ca. 1 air km south of the summit of Mt. Hamiguitan. The substrate was predominantly ultramafic. This forest has the highest number of endemic and threatened plant species among the five vegetation types surveyed by Amoroso and Aspiras (2011). The habitat where this plant was found also contained other mycoheterotrophs such as Burmannia lutescens (a new record for this species for the Philippines) and Sciaphila sp. (Triuridaceae). Association of different mycoheterotrophs in one local area was mentioned by Schlechter (1913) and Pelser et al. (2019). This phenomenon may reflect the ecological requirements of the fungi or the association of different plant species with one fungus (Maas-van de Kamer 1998). The latter seems to be supported for Burmanniaceae where that family as well as Gentianaceae and Triuridaceae have been found associated with Glomerales and Diversisporales (Hynson and Bruns 2010). Because only one population of G. syceorosensis was discovered, no estimation of its abundance or overall distribution can me made. It, like most Gymnosiphon species, is likely rare in nature, but because it is inconspicuous, it is likely undercollected. Until further work can be undertaken to determine how many populations of G. syceorosensis exist, the conservation status of this species should at this time be considered Data Deficient (DD) according to the IUCN (2019). Note that the DD category does not imply that the taxon is not threatened.
Etymology. The specific epithet commemorates the Mt. Hamiguitan Range Wildlife Sanctuary. The word "hagímit" is Cebuano for "a small tree of primary forest with rough leaves: Ficus sp." (Wolff 1972). Apparently the "g" and "m" consonants were switched (a common occurrence in Cebuano), thereby producing "hamigit". Adding the suffix "-an" which mean "a place of " gives hamigitan, i.e. "a fig tree place" or "a place with a fig tree". When constructing the specific epithet for Gymnosiphon, the goal was to express "from fig-mountain". Fig-tree is translated to Latin as "syce" (συκη, feminine) and mountain as "oros" (όρος, masculine), thus giving "syceoros" (Stearn 1992). Using one of the recommended adjectival endings for geographic epithets with a masculine termination yields "syceorosensis".
It should be pointed that generic names derived from Greek that end in "-on" are often interpreted as neuter, however, according to ICN Art. 62.2, compound generic names take the gender of the last word in the nominative case in the compound. In this example, the Greek word element -siphon (σίφων) is masculine, thus the gender for all specific epithets of Gymnosiphon should be masculine. The type species was originally published by Blume (1827) as G. aphyllum (neuter), but this should be corrected to G. aphyllus (masculine).