Phylogenetics of Serjania (Sapindaceae-Paullinieae), with emphasis on fruit evolution and the description of a new species from Michoacán, Mexico

Serjania is among the largest Neotropical genera of Sapindaceae and comprises ∼240 species. Traditionally, the genus has been defined by its schizocarpic fruits separating into three distally winged mericarps. However, recent phylogenetic studies have revealed that fruit type is not consistent within the genus and that it also includes a few species having capsular fruits that were previously placed in other genera. A phylogenetic analysis is presented based on nuclear ITS and chloroplast trnL-F sequences. The ingroup consisted of 48 species broadly sampled from across the geographic range and taxonomic spectrum of Serjania and focusing on species with atypical fruits. An ancestral character state reconstruction of fruit type was performed and demonstrated that the ancestral fruit type of the genus is strongly supported to be a schizocarp, but there are at least five independent derivations to capsular fruits and at least one transition back from capsules to schizocarps. Also, transitions from winged to wingless mericarps have occurred at least twice. Infrageneric classification is problematic and all of the sections that were represented by more than one species come out as poly- or paraphyletic. Our limited sample of taxa precludes the possibility of a new classification at present and a broader phylogenetic sample of the genus will be needed to understand relationships and determine which lineages warrant formal recognition. In addition to the phylogenetic study, Serjania frutescens, a new capsular-fruited species from Michoacán, Mexico, is described, illustrated, and compared with its putative closest relatives. A key is provided to differentiate this species from other Mexican Serjania with capsular fruits, and new synonymies and lectotypifications associated with capsular-fruited Serjania tortuosa and S. sonorensis are established. Phylogenetic results suggest that these two are cryptic species and at least superficially indistinguishable by morphology.


Introduction
Serjania Mill. is one of the largest Neotropical genera of the family Sapindaceae, and like many other taxa of angiosperms, its circumscription has changed considerably during the past decade due to a better understanding of the relationships among its species and those of closely related genera. It comprises $240 species (Ferrucci & Steinmann, 2019) and belongs to the tribe Paullinieae, which includes climbers or climber-derived shrubs with stipulate, compound leaves and a pair of tendrils at the base of the inflorescence rachis (Acevedo-Rodr ıguez et al., 2017). Recently, these authors confirmed the position of Serjania in the tribe and placed it in the supertribe Paulliniodae. Most species of this genus are woody vines that form dense masses on their supporting vegetation, although a rare few are erect shrubs or subshrubs, which may be low climbers, such as Serjania rosalindae Miller (1754) first proposed Serjania as independent of Paullinia, and his approach was further sustained by Schumacher (1794), who recognized fruit morphology as an essential feature. Since its conception the genus has been defined by its schizocarpic fruits (Figs 10,12,16,17) separating into three distally winged mericarps. Plants with divergent fruit types were placed into other genera of Paullinieae, e.g., Cardiospermum L., Urvillea, Paullinia, Houssayanthus Hunz., Balsas J. Jim enez Ram. & K. Vega, Chimborazoa H. T. Beck, and Thinouia Triana & Planch. This circumscription was followed until a broad phylogenetic study of Paullinieae, including 20 species of Serjania, revealed that fruit type was not consistent for the identification of this genus, and the segregate genera Houssayanthus, Balsas, and Chimborazoa were all nested within a paraphyletic Serjania (Acevedo- Rodr ıguez et al., 2017). Two species previously placed in Cardiospermum, C. urvilloides (Radlk.) Ferrucci and C. integerrimum Radlk., form a grade basal to Serjania. They may also belong to the genus, but were not included in our phylogenetic analysis.
So far, only one study in Paullinieae has tested the evolutionary dynamics of fruit morphology evolution in a phylogenetic context. Chery et al. (2019) presented the results of a molecular phylogeny of Paullinia including 11 molecular markers and 64 species. They found that the fruit type of the genus is generally a septifragal capsule, with the exception of a few species in sect. Castanella (Spruce ex Hook.f) Radlk. that have indehiscent fruits dispersed by water. However, the pericarp is highly variable with regard to the presence of wings and spines, and there have been repeated evolutionary transitions and reversals.
The purpose of the current study is to expand the phylogenetic sampling of Serjania, focusing on species with unusual fruits in order to better understand relationships within the genus and the evolution of fruit types and their morphology. Specifically, the fruit typesschizocarps and capsulesare here hypothesized to be dispersion-related adaptations. Among the morphological traits, the position of the locule (apical vs. central or basal) and the presence or absence of wings were considered. Fruit type is mapped on a molecular phylogeny to observe the patterns of fruit evolution within the genus. In addition, a new species of capsular-fruited Serjania from Michoac an, Mexico is described and illustrated, and taxonomic information is provided for two other species closely related to the new one.

Phylogenetic analysis
Taxon sampling and outgroup selection. We sampled a broad representation of Serjania from across its distributional range and taxonomic spectrum. All species with capsular fruits were included, as too were the species initially placed in segregate genera now considered within Serjania by Acevedo-Rodr ıguez et al. (1 sp.), and Urvillea (2 spp.). The ingroup consisted of 29 previously sampled species (from Acevedo-Rodr ıguez et al., 2017 andChery et al., 2019) and 19 newly sequenced species. All but one accession were determined to species. In total, the ingroup contained 48 species, and 13 of these were represented by multiple accessions for a total of 69 OTUs or 80 when also considering the outgroup. We included all of the 12 sections recognized by Radlkofer (1931Radlkofer ( -1934, and most of these were represented by more than one taxon. Sectional names are found in Fig.  18, as too are the names of the genera for species previously recognized outside Serjania. Material of the newly sequenced species came from either field-collected material or specimens housed in the herbaria ARIZ, IEB, MEXU, and RSA. All newly sequenced representatives included in the phylogenetic reconstruction are associated with an herbarium voucher number and a GenBank code, and information from the previously sequenced representatives is also included (see Appendix 1).
DNA extraction, amplification and sequencing. For the newly included species, total genomic DNA was extracted following the protocols of Doyle and Doyle (1987) and Cullings (1992) for DNA isolation from dried leaves or herbarium specimens. Two regions were selected for this study: ITS for nuclear and the trnL intron for chloroplast. Both of these are non-coding regions and were chosen because they have been used successfully in previous studies of Sapindaceae (e.g., Acevedo-Rodr ıguez et al., 2017). Genomic DNA was diluted to a concentration of 5-10 ng per ll in order to reduce inhibition of enzymes during the polymerase chain reactions (PCR), but some samples that did not amplify at these concentrations were successfully amplified at considerably higher concentrations using stock DNA.
Molecular work was conducted at the laboratory of California Botanic Garden in Claremont, CA, USA. PCR reactions were carried out using GoTaq Flexi DNA Polymerase (Promega, Madison, WI, USA) in a T100 Thermal Cycler (Bio-Rad, Hercules, CA, USA). A negative control was used in each reaction run, replacing DNA with nuclease-free water to test for contamination. The PCR mixture included 16.38 ll of nucleasefree water, 2.5 ll of 5Â colourless GoTaq flexi buffer, 1.25 ll of dNTP (250 ll/2 mM), 1.25 ll of MgCl 2 (25 mM), 0.125 ll of 5 u/ll Taq polymerase, 1.25 ll of each primer (10 lmol/l) and 1 ll of template DNA.
The ITS region was amplified using the primers ITS4 and ITS5 from White et al. (1990), and the trnL intron was amplified with the primers C and F from Taberlet et al. (1991); for those samples with degraded DNA, we used internal primers E and D from Taberlet et al. (1991). The PCR program for both regions was as follows: 95 C for 4 min for initial denaturation; 35 cycles of 95 C for 45 s for denaturation, 60 C for 45 s for annealing, and 72 C for 2 min for extension; a final 7 min extra extension at 72 C was also included. The PCR products were examined in agarose gels (1%) by gel electrophoresis using SYBR Green TM (Thermo Fischer Scientific, Waltham, MA, USA). The unpurified PCR products were sent for sequencing using the same PCR amplification primers to Functional Biosciences (Madison, WI, USA). Sequences were obtained using an ABI 3730xl DNA Sequencer. Previously published sequences were obtained either directly from the authors or by downloading the data sets accompanying the supplemental information of these publications. Sequences were assembled and edited using SequencherV R v.5.2 (Gene Codes Corporation, Ann Arbor, MI, USA).
Phylogenetic and character state reconstruction. Both the ITS region and the trnL intron were concatenated using SequenceMatrix v.1.8 (Vaidya et al., 2011), and they were aligned with MUSCLE v.3.8.31 (Edgar, 2004) using default parameters and subsequently manually adjusted in PhyDE v.0.9971 (M€ uller et al., 2006). Maximum likelihood phylogenetic analysis was conducted using IQ-TREE multicore v.2.0.3 (Minh et al., 2020). The best evolutionary model was determined using the ModelFinder (Kalyaanamoorthy et al., 2017) built-in option in IQ-TREE. TIMe þ R3 was the best-fit model according to Bayesian Information Criterion. Support was determined by the Shimodaira-Hasegawa approximate likelihood ratio test (SH-aLRT) with 1000 replicates (Guindon et al., 2010) and ultrafast bootstrap (UFBoot) (Minh et al., 2013). The Bayesian analyses were conducted under the GTR model using MrBayes v.3.2.6 (Ronquist et al., 2012) in CIPRES (Miller et al., 2010) with two simultaneous runs, each with four Markov chains (three cold and one heated) for 5,000,000 generations. Trees were sampled every 1000 generations, and for the summarization, the first 25% of generated trees were discarded, followed by reconstruction of the majority consensus tree. A simple ancestral character state reconstruction of fruit type (capsule vs. schizocarp) was performed with Mesquite v.3.61 (Maddison & Maddison, 2019) using likelihood as the estimation method under the Mk1 model, assuming an equal probability of character states at the root node. Fruit type was determined by examining the voucher specimens or for instances in which the specimen lacked fruit, by reviewing other material of the species or pertinent literature. Ancestral states were mapped in the main phylogeny tree.

Biogeography
To understand biogeographic patterns, the 48 species included here are broadly classified by distribution, and this feature is mapped on the phylogeny. The following categories are used: South America (SA), Mexico (MX), Central America (CA), United States (US), and the Antilles (A).

Taxonomic treatment
A detailed review of the published literature of Serjania was conducted, including the consultation of the protologues of all capsular-fruited species as well as the only monograph of the genus (Radlkofer, 1931(Radlkofer, -1934. Herbarium material from A, BCMEX, BM, CTES, F, G, GH, K, M, MEXU, MO, NY, QMEX, UC, US, and WU was examined. Additional high-resolution images of type specimens on JSTOR Global Plants (2020) were viewed for material at C, M, and NDG. The acronyms of the cited herbaria follow Index Herbariorum (Thiers, 2020). For species delimitation, we adopt the Unified Species Concept of De Queiroz (2005), which defines species as separately evolving metapopulation lineages. Leaf samples were fixed in FAA (formalin/70%/ alcohol/acetic acid), immersed in CO 2 for critical-point drying, and sputter coated with gold-palladium for scanning electron microscope (SEM) studies. Pollen grains were obtained from anthers of one collection. Samples for light microscopy (LM) were acetolysed according to the procedure of Erdtman (1966) and mounted in glycerin jelly, a mixture of gelatin and glycerine (Johansen, 1940). Permanent slides were deposited at the Palynological Laboratory of the National University of the Northeast, Corrientes, Argentina (PAL-CTES). Polar axis and equatorial diameter were measured on 20 grains using a Leica DM LB2 microscope. The terminology used to describe the grains follows Erdtman (1966) and Punt et al. (2007). Scanning electron images (SEM) were taken from leaves and acetolysed pollen grains. The equipment used was a JEOL 5800 LV operating at 20 KV. The conservation status assessment of Serjania frutescens was based on IUCN Standards and Petitions Subcommittee (2017) Red List criteria recommendations. Extent of occurrence (EOO) and area of occupancy (AOO) were estimated utilizing GeoCAT (Bachman et al., 2011), for the latter utilizing grid cells of 4 km 2 , as recommended by the IUCN Standards and Petitions Subcommittee (2017), since it relates to thresholds of criterion B.

Results
Phylogenetic analysis DNA sequencing. The entire ITS region was obtained and analysed for all species and all accessions, whereas the trnL intron was analysed for all but six species and eight accessions (see Appendix 1). The ITS sequences ranged from 577 to 652 bp. The aligned matrix was 761 bp, with 241 (31.7%) parsimoniously informative sites. Approximately 40% of the trnL intron sequences were partial. They varied from 537 to 990 bp, with the shorter length representing a partial sequence. The aligned matrix was 1098 bp, with 46 (4.2%) parsimoniously informative sites. The alignment and resulting trees are archived on Zenodo (https://doi.org/10.5281/zenodo.5593388).
Phylogenetic and character state reconstruction. The Maximum likelihood tree (Fig. 18) is well resolved and moderately well supported with many nodes having greater than 90% SH-aLRT and UFBoot values. Results from the Bayesian analysis are shown in Supplemental  Fig. S1, and the 50% majority-rule consensus tree is similarly well resolved with many supported nodes. Cardiospermum urvilleoides is sister to Serjania, but not closely related to the other species of Cardiospermum included here. The reconstruction retrieves Serjania as monophyletic, but only with the inclusion of the taxa mentioned below that were transferred by Acevedo-Rodr ıguez et al. (2017). We designate nine well-supported clades (SH-aLRT ¼ 85-100, six of them also well-supported by UFBoot ¼ 90-100, see Fig. 18), that correspond to the primary lineages recovered in the analysis. Clade 1 contains S. ampelopsis, S. cardiospermoides, and S. communis (SH-aLRT ¼ 98, UFBoot ¼ 90) and is sister to the remainder of the genus (SH-aLRT ¼ 84, UFBoot ¼ 65). All of these species were previously placed in sect. Platycoccus, but the section is here supported to be polyphyletic with S. cuspidata of Clade 2 being more closely related to S. caracasana (sect. Eucoccus) than to the other members of sect. Platycoccus. In fact, all of the sections that were represented by more than one species come out as poly-or paraphyletic. As proposed by Acevedo-Rodr ıguez et al. (2017), the segregate genera Balsas, Houssayanthus, and Chimborazoa are nested within Serjania, as too are the following previously unsampled species that had been placed in Cardiospermum or Paullinia but recently transferred to Serjania based on morphology: Serjania sonorensis (Paullinia), S. cristobaliae (Cardiospermum), S. dissecta (Cardiospermum), and S. tortuosa (Cardiospermum).

Biogeography
Concerning distribution, the earliest two diverging clades (1 and 2) contain species that occur mostly in South America as well as Central America and Mexico. The next clade (3) has only South American species. Of the remaining lineages, some are composed entirely of Mexican and Central American species (clade 7), others entirely of South American species (clades 6 and 8), but most have Mexican, Central American, and South American species (clades 4, 5, and 9).

Phylogenetic analysis
There is considerably less variation in the trnL-F region (4.2%) than in the ITS region (31.7%). When analysed separately, trnL-F provides little resolution, thus our decision to present only the results from a combined analysis. Twelve species are represented by more than one accession. In all instances, the species form a single monophyletic lineage with the exception of S. paucidentata being nested within S. pyramidata (clade 5).
One unsuspected and enigmatic result is the distant relationship between Serjania tortuosa and S. sonorensis. Both of these species occur in arid north-western Mexico, with the former endemic to Baja California and the latter restricted to the states of Sonora and Sinaloa. Although they were traditionally placed in separate genera (Cardiospermum and Paullinia, respectively), they are remarkably similar in morphology. In fact, we are unable to ascertain characters that consistently differentiate the two. However, mainland plants and peninsular plants, both represented by more than one accession, belong to distinct lineages with vastly different evolutionary histories. They are not even closely related with S. tortuosa forming part of clade 7 (SH-aLRT ¼ 99, UFBoot ¼ 92) and S. sonorensis being a member of clade 9 (SH-aLRT ¼ 96, UFBoot ¼ 99). More study of the two would be beneficial to determine any potential differences that would support their placement in different clades of Serjania, and S. tortuosa in itself would benefit from research on its variability to determine if one or more species might be involved.
Fruit evolution. No other character has played a more important role in the classification of the tribe Paullineae than fruit type. Since its conception, Serjania has been defined by the possession of schizocarps that separate into three winged mericarps. This is certainly an adaptation to wind dispersal, but Acevedo-Rodr ıguez (1993) mentions that the mericarps also can be secondarily dispersed by water.
The ancestral fruit of Serjania is strongly supported to be a schizocarp, and the lineages just outside the genus possess capsules. Within the tribe Paullinieae schizocarps are otherwise only found in the distantly related Lophostigma and Thinouia, which have a basal position in the tribe, showing that this trait is homoplastic in Paullinieae. More than 95% of the species of Serjania possess schizocarps, but there are at least five reversals to capsules: (1) S. cristobaliae of clade 3; (2) S. biternata and S. rosalindae of clade 4; (3) S. tortuosa of clade 7; (4) S. dissecta of clade 7; and (5) the grade including S. sonorensis, S. frutescens, S. crucensis, and S. guerrerensis of clade 9. Interestingly, there is also evidence of one transition back to a schizocarp from a capsular fruit, this supported by the schizocarpic S. schiedeana being nested within a clade that contains capsular-fruited species.
Although there have been many independent derivations of capsular fruit from the typical fruit of Serjania, in all instances, there has been no subsequent radiation of species. Four times there is but a single species resulting from the reversal. The most diverse group of capsular Serjania is represented by the grade of S. sonorensis, S. frutescens, S. crucensis, and S. guerrerensis in clade 9. Still though, there are only four species involved. The lack of subsequent diversification despite various derivation capsular fruits suggests that transitions are all either very recent and/or that this change results in an evolutionary dead end due to the loss of the highly adaptive wind dispersal mechanism.
The capsules of some Serjania are winged, e.g., the longitudinal locular wings of S. rosalindae (Fig. 14) or the alate stipe of S. frutescens (Fig. 15). In S. tortuosa, there is a reduction to exhibiting wingless capsules, or exceptionally only a carina or rarely a reduced wing is observed in the proximal half, at the dorsal vein level of each carpel (Fig. 11). However, in all these species the wings are insignificant and unlikely to aid in dispersal. In fact, the capsular fruited species appear to be atelechorous, and further research is needed to understand their mechanisms of dispersal. Atelechory is generally associated with arid regions and deserts in particular (Fllner & Shmida, 1981), and traits constraining seed dispersal are common in arid and semiarid floras (Van Rheede van Oudtshoorn & van Rooyen, 1999). The occurrence of capsular fruits having especially large seeds has been documented in some species of Serjania (e.g., Ferrucci & Steinmann, 2019;Ferrucci & Urdampilleta, 2011), and there appears to be a correlation between these two traits. However, a statistical analysis of more species will be necessary for confirmation. The larger size of the seed would be a mechanism to avoid being transported over long distances but also give rise to larger seedlings and these better withstand environmental hazards like deep shade and drought (Bruun & Ten Brink, 2008). This character is observed not only in species of Serjania native to arid lands, but also in some occupying the interior of deciduous tropical forests, e.g., S. biternata.
The only two species that are found in true deserts both possess capsular fruits, S. tortuosa of the Sonoran Desert and S. dissecta of the Chihuahuan Desert. As mentioned above, these taxa represent independent transitions from the typical schizocarpic fruit. The capsular fruited Serjania frutescens and S. crucensis occur in the Infiernillo region of Michoac an, Mexico, which is the most arid portion of the Balsas Depression. Similarly, S. cristobaliae is restricted to rocky outcrops in a semiarid portion of Minas Gerais, Brazil. There are some other interesting fruit trends in Serjania. Independent transitions from winged to wingless mericarps have occurred at least twice, once in S. lachnocarpa (clade 6) and again in Serjania rzedowskiana (clade 9). This feature was used to justify the segregate genus Chimborazoa to accommodate the latter species. However, Acevedo-Rodr ıguez (1998) relegated Chimborazoa to synonymy under Serjania based on fruit morphology, and his decision was supported by Acevedo-Rodr ıguez et al. (2017) and is further reinforced here. The other species possessing shortly winged or wingless mericarps at maturity are S. cissoides Radlk., S. herteri Ferrucci, and S. macrococca Radlk. However, none of these were assessed in our study. The mericarps of Serjania incana Radlk. are noteworthy in having a central locus and a circumferential wing that extends longitudinally (Fig. 13). That trait was used to define the segregate genus Houssayanthus, but molecular evidence supports that Houssayanthus is nested within Serjania, and the change represents a relatively minor modification from the typical fruits of Serjania with their distal locus and proximal wings.

Biogeography
Our sample represents fairly well the geographic distribution of Serjania, which ranges from Texas in the USA to northern Argentina and the Caribbean. Twenty species included here are restricted to South America, and 14 are endemic to Mexico. Of the remainder, five are shared between Mexico and Central America, one occurs in Texas and northern Mexico, one is endemic to Central America, and an additional seven range from South America to Central America, Mexico and/or the Antilles. No endemic Antillean species were included in our analysis. Acevedo-Rodr ıguez (1993) suggests a South American origin of Serjania, and our results support his proposal in that the earliest diverging lineages are primarily South American. Clades 1, 2, 4, 5, and 9 contain species from both South America and Mexico. Clades 3, 6, and 8 are strictly South American, whereas only clade 7 is comprised of entirely Mexican and Central American species. In the clades with species from Mexico/Central America and South America, complex patterns of distribution occur. For example, all the species of clades 2 and 5 occur in South America, but there is at least one widespread member that extends as far north as Mexico. Within clade 9, the lineage containing the South American species S. inflata, S. deltoidea, and S. rhombea is nested in a group of Mexican and Central American species. Of the four species in clade 4, S. marginata is restricted to Brazil, Bolivia, and Argentina, S. mexicana is widespread from northern South America to Mexico and the Antilles, and S. biternata and S. rosalindae are narrow endemic species known from small areas in the Balsas Depression of central Mexico. As more species are included in the phylogeny, more complex patterns will certainly emerge and a more complete picture of the biogeography of Serjania will be revealed. For now, we explain the existence of widely distributed clades by the fact that the fruits of most Serjania are wind dispersed and presumably able to travel long distances.

Infrageneric classification
Our study further supports the conclusion of Acevedo-Rodr ıguez et al. (2017) that Serjania, as traditionally defined by schizocarpic fruits with winged mericarps, is highly paraphyletic. Monophyly is achieved only after the inclusion of some capsular fruited species that were previously placed in closely related genera (Cardiospermum cristobaliae, C. dissectum, C. tortuosum, and Paullinia sonorensis), as well the synonymization of Houssayanthus and Balsas. As a result, Serjania losses much of its taxonomic predictability and becomes more difficult to define.
The last complete treatment of Serjania is Radlkofer's (1931Radlkofer's ( -1934 monograph for Engler's Das Pflazenreich, in which he recognized 12 sections and 208 species. Since then, numerous additional species have been proposed, and the most significant post-Radlkofer treatment is Acevedo-Rodr ıguez's 1993 revision of Serjania sect. Platycoccus. In addition to his account of the section, Acevedo-Rodr ıguez provides a detailed synopsis of the genus and proposes substantial modifications to Radlkofer's infrageneric classification, which he describes as 'untenable' and 'difficult to use'. Acevedo-Rodr ıguez, in contrast to Radlkofer, recognizes only six sections, describing a new one (sect. Confertiflora), correctly applying the name sect. Serjania (¼ sect. Dictyococcus sensu Radklofer), and reducing seven of Radlkofer's sections to synonymy under sect. Serjania (sects. Eucoccus, Pachycoccus, Holcoccous, Simococcus, Oococcus, Phacococcus, and Syncoccus). Both Radlkofer and Acevedo-Rodr ıguez emphasized fruit characters as defining features of their sections. Furthermore, Radlkofer used numbers of sepals, cambial variants, and presence of mucilaginous idioblasts in the epidermis of the leaflets, whereas Acevedo-Rodr ıguez placed importance on seed shape. Coulleri et al. (2012) applied a cytogenetic approach to evaluate infrageneric classification of Serjania. The authors also considered some important reproductive and vegetative morphological traits. Their results did not support either of the two classifications and demonstrated a lack of diagnostic characters specific to each section in Radlkofer's proposal and, on the other hand, the reduced number of traits under consideration, only carpological, in Acevedo-Rodriguez's classification.
Our sample contains $20% of the genus, and all of the 12 sections proposed by Radlkofer (1931Radlkofer ( -1934 are represented. However, we were unable to include Acevedo-Rodr ıguez's monotypic sect. Confertiflora. Our results clearly demonstrate that most of the previously defined supraspecific taxa recognized by Radlkofer (1931Radlkofer ( -1934 or Acevedo-Rodr ıguez (1993) are not monophyletic. In fact, the classification of Serjania can be summarized in two words … a mess! For example, three members of sect. Platycoccus (S. communis, S. cardiospermoides, and S. ampelopsis) form the earliest diverging lineage within the genus, clade 1 (SH-aLRT ¼ 98, UFBoot ¼ 91). However, a fourth species assigned to the section is strongly supported (SH-aLRT ¼ 100, UFBoot ¼ 95) to belong in clade 2 with S. caracasana of sect. Eucoccus (following Radlkofer) or sect. Serjania (following Acevedo-Rodr ıguez). To further complicate matters, an additional species assigned to sect. Eucoccus by Radlkofer, S. pyramidata, is not closely related to S. caracasana but instead forms part of clade 5. Other examples of polyphyletic assemblages of Radlkofer's sections that were later reduced to synonyms of sect. Serjania by Acevedo-Rodr ıguez (1993) include sects. Oococcus (clades 5, 6 and 8), Syncoccus (clades 5 and 9), Pachycoccus (clades 4), and Holcococcus (clade 5). Section Eurycoccus is among the sections circumscribed similarly by both Radlkofer and Acevedo-Rodr ıguez. However, our results show that it is monophyletic only after the inclusion of S. cristobaliae, which was initially described as a species of Cardiospermum. Physococcus is another section recognized by both Radlkofer and Acevedo-Rodr ıguez that is clearly polyphyletic with its species coming out in three independent portions of clades 7 and 9. With regard to Acevedo-Rodr ıguez's expanded sect. Serjania, it too is highly paraphyletic, not only as a result of containing many species previously placed in other genera, but also because members of sects. Ceratococcus and Physococcus are nested within it.
Fruit characteristics have been of monumental importance in the classification of the tribe Paullinieae, but in the case of Serjania, they are not good indicators of relationship, and previous attempts to construct a natural infrageneric classification based on them have been unsuccessful. The recent inclusion of a small group of Serjania with capsular fruits makes the infra-generic classification even more difficult, and our limited sample of taxa precludes the possibility of a new system at present. A considerably broader phylogenetic sample of the genus will be needed to understand relationships and determine which lineages warrant formal recognition. Also, much more research is needed on the individual species to determine morphological, anatomical, cytological, and/or palynological features that could define future infrageneric taxa because little is gained by simply recognizing infrageneric taxa that are not predictable and supported by other evidence. Diagnosis. Species morphologically similar to the polymorphic Serjania tortuosa and S. sonorensis but distinguished by a completely shrubby habit; 3-foliolate leaves, subchartaceous, laticifers absent from lower epidermis; tendrils absent; the two posterior nectary lobes ovoid, obtuse at the apex and the anterior ones circular in outline; pistillode and gynoecium glabrous; and capsule globose stipitate, stipe narrowly winged, ± 7 mm long, glabrous.
Etymology. The specific epithet refers to the shrubby habit.
IUCN conservation status. Serjania frutescens is known from six collections, representing four locations. Although two of these are inside the Zicuir an-Infiernillo Biosphere Reserve, there are still threats of deforestation and extensive cattle raising. The northern-most population along the R ıo Marqu es was lost in 2011 when the area was flooded after the construction of a dam. The initial extent of occurrence (EOO) was calculated to be $186.6 km 2 , but is now less with the loss of the northern-most population. The area of occupancy (AOO) is estimated at 16 km 2 . Therefore, following the IUCN criteria ( IUCN Standards & Petitions Subcommittee, 2017), and if a formal analysis were performed this species would probably be considered as endangered B1ab(ii,iii,iv)þ2ab(ii,iii,iv). Notes. The leaves of S. frutescens are characterized by the presence of stomata on the lower surface and few stomata near the midrib on the upper surface, a condition called hypoamphistomatic (Mart ınez Quesada, 2009). Approximately 35% of the species of Serjania are narrow endemics known from few collections (Acevedo-Rodr ıguez, 1993), and S. frutescens is another one of these.
Serjania frutescens is characterized by its shrubby habit, apterous capsules with winged stipe, 3-foliolate leaves lacking laticifers on the lower epidermis, deciduous stipules, and the absence of tendrils. Among its congeners, the most similar taxa are two cryptic species, S. tortuosa and S. sonorensis, and the differences are listed in Table 1. Both S. tortuosa and S. sonorensis were previously treated in Cardiospermum or Paullinia, and their placement in Serjania is the result of the molecular phylogenetic analysis conducted by Acevedo-Rodr ıguez et al. (2017). A synopsis of these two species is presented below, including a description of the fruit, as well as data on distribution, ecology, phenology, and list of specimens examined. Lectotypifications for both species and new synonyms for S. tortuosa are also included. Finally, a key to identify this small group of Mexican Serjania species with capsular fruit is presented.
Serjania    Only the collection site of Cardiospermum tortuosum is provided in the protologue, but two syntypes with duplicates were located. These correspond to the collections of R.B. Hinds s. n. and G.W. Barclay 3078. The duplicates detected coincide with those detailed by Raven (1964) in his very useful paper about George Barclay and his collecting in California. The sheet on which K 000037463 is mounted also contains material from the Barclay collection which was given to Bentham by Hooker in 1842 and is separately barcoded. The specimen here selected as lectotype is a branch in good condition and with a fruit.    The pencil annotations on the sheet indicate that Bentham had both specimens available when describing the species and preparing the excellent figure that accompanies it. In the protologue of Serjania californica, the author does not indicate in which herbarium the material studied by him is deposited, and we were unable to find a specimen in M. Among the three duplicates of the type collection located, we choose as lectotype the specimen deposited in K that was identified by Radlkofer in his handwriting in October 1885.
The protologue of Cardiospermum spinosum (Radlkofer, 1895) states 'Affinis Cardiospermo tortuoso' and mentions a collection made by Palmer in La Paz, Lower California. We have located specimens in GH, K, UC, and US that coincide with the collector and morphology. Among these specimens, we designate the one at GH as lectotype because it is a very well-preserved specimen.
The protologue of Serjania albida states 'Affinis videtur Serjaniae (?) californicae' and cites a collection made 'In California inferiore ad Santa Agueda'. Three duplicates were found that match perfectly with the protologue in locality, collector, and collection number (M 0212268, MEXU 00018883, NY). The lectotype at M that we here designated is the best preserved, although it is still very poor. Also, it is identified by Radlkofer in his own handwriting.
The following carpological and seminal characters are described for the first time. Serjania tortuosa shares these traits with S. sonorensis, a very close species that cannot be distinguished by morphological characters.
Distribution, ecology, and phenology. Mexican endemic of Baja California Sur, where it occurs along volcanic rocks and forms part of the thornscrub vegetation, at elevations of 0 to 1200 m. Flowers from August to April and fruits from September to March.
Notes. Serjania tortuosa and S. sonorensis are cryptic species and at least superficially indistinguishable by morphology, although more study would be beneficial. They are mainly recognized by having leaves that are 3or 5-foliolate or biternate. The leaflets are obovate or ovate, incise-dentate or dentate-serrate, less often pinnatisect, and abaxially they have conspicuous laticifers. Their fruits are with inconspicuous venation or little marked. The abundant material reviewed indicates that it is a fairly polymorphic species with xeromorphic adaptive characters such as: development of spinescent tendrils, the abundance of stomata in the adaxial leaf surface, and the degree of division of the blade.