The neotropical genus Witheringia L′Her. encompasses about 20 species, with a geographic range extending from Mexico and the Antilles to Bolivia and Brazil (Hunziker 2001; Bohs 2015). Taxonomic confusion has accompanied Witheringia from the outset. It was first described in 1788 based on a single specimen grown in cultivation, and named W. solanacea L′Her. (L′Heritier 1788). Sixty years later, Miers (1849a) transferred the type species to the genus Saracha Ruiz & Pav. He retained the name Witheringia but applied it instead to species within the recently described Athenaea Sendtn. For the next twelve decades, specimens of Witheringia were assigned to a plethora of genera, including Acnistus Schott, Athenaea, Bassovia Aubl., Brachistus Miers, Capsicum L., Cuatresia Hunz., Lycianthes Hassl. (Dunal), and Saracha. Hunziker (1969) reinstated the genus based on its original type species. Since that foundational monograph, Witheringia has been considered only in regional treatments (e.g. D′Arcy 1973, 2001; Gentry and Standley 1974; Bohs 2015) and in a Ph.D. thesis (Sousa-Peña 2001). The focus on regional floras has led to confusion about species circumscriptions. Various authors have designated new species, proposed merging species, or resurrected previously synonymized species. Therefore, the number of species and their delimitation are currently unclear. For example, Hunziker (1969) considered W. coccoloboides and W. fuscoviolacea to be distinct species whereas D′Arcy (1973) synonymized them in his Flora of Panama, and Bohs (2015) in the Manual of Plants of Costa Rica was unsure whether the primarily Panamanian W. correana should be considered a synonym of W. macrantha.

Phylogenetic studies of the Solanaceae establish that Witheringia lies within the tribe Physalideae and subtribe Physalidinae [formerly subtribe Physalinae; Olmstead et al. 2008; Särkinen et al. 2013; Deanna et al. 2019]. Deanna et al. (2019) examined phylogenetic relationships in the tribe Physalideae using four gene regions and extensive taxonomic sampling. Cuatresia, a genus previously thought to be closely related to Witheringia, is instead part of a clade with Aureliana Sendtn. and Deprea Raf. in a recircumscribed subtribe Withaninae. The subtribe Physalidinae forms a clade that includes Witheringia, Physalis L., and several small “physaloid” genera [e.g. Leucophysalis Rybd., Chamaesaracha (A.Gray) Benth. & Hook.f., and Tzeltalia E.Estrada & M.Martínez], whose relationships are still unclear.

A major systematic problem with Witheringia above the species level is its relationship with the genus Brachistus. Miers (1849b) created Brachistus and included fifteen species within it. All of these are now assigned to different genera (Capsicum, Cuatresia, Physalis, and Witheringia) except B. stramoniifolius, which has been placed in Brachistus by many authors. Species of Brachistus have been distinguished from Witheringia by their lobed and often accrescent calyces, whereas those of Witheringia are truncate or shallowly lobed and not or slightly accrescent in fruit (Hunziker 1969; D′Arcy et al. 1981; Bohs 2015). Hunziker (1969) originally considered Brachistus to be a section within Witheringia and recognized three species within it, although he later elevated it as a separate genus (Hunziker 2001), as did D′Arcy et al. (1981) and Bohs (2015). A molecular phylogeny of the Solanaceae based on cpDNA markers placed B. stramoniifolius nested in a clade containing four species of Witheringia (Olmstead et al. 2008). Subsequent analyses (Särkinen et al. 2013; Zamora-Tavares et al. 2016) have found B. stramoniifolius to be sister to Witheringia, but these analyses sampled only two to three Witheringia species and did not include other Brachistus species. Deanna et al. (2019) included two Brachistus species in their phylogenies; these emerged in a clade with multiple Witheringia species and Schraderanthus viscosus (Schrad.) Averett. Whether Witheringia and Brachistus are each monophyletic has been unknown (Bohs 2015).

The center of diversity for Witheringia is currently considered to be Costa Rica (Hunziker 2001). According to the time-calibrated phylogeny of Särkinen et al. (2013), the clade including Brachistus and Witheringia diverged from Leucophysalis ca. 3.72 million years ago, although subsequent discoveries such as Physalis fossils from the early Eocene of Argentina dated at ca. 52 million years suggest that the Physalideae diversified much earlier (Wilf et al. 2017; Deanna et al. 2020, 2023). Therefore, the origin and diversification of the genus is largely contemporaneous with the uplift of the northern Andes ca. 15–35 Ma and the formation of the Central American land bridge ca. 3–20 Ma (Coates and Obando 1996; Blandin and Purser 2013; Bacon et al. 2015), providing an interesting example of diversification in a geologically dynamic context.

Our goal in this study was to create a well-supported phylogenetic analysis of Witheringia and Brachistus, including three of the three to four currently recognized species of Brachistus and 13 of the approximately 20 species of Witheringia, as well as multiple accessions of widespread species. The molecular phylogeny will provide support for a thorough taxonomic treatment. Increased resolution of the phylogeny of Witheringia will also permit investigations of character evolution and biogeographic questions.

Materials and Methods

Taxon Sampling—The ingroup consists of 31 accessions belonging to 16 putative species, including three species of Brachistus and 13 species of Witheringia, as well as multiple accessions of widespread species and several putative new species (Appendix 1). All but two of the ingroup accessions were collected or propagated by the authors. For B. nelsonii, DNA was extracted from an herbarium specimen collected in 2000 (Appendix 1). DNA for W. correana from Panama was shared by R. Olmstead from the BIRM Solanaceae collection. All new specimens gathered for the study were collected legally with all necessary valid permits; see Acknowledgments for permit information.

Several of the Witheringia accessions used here are provisionally identified pending increased taxonomic study of the genus. Witheringia cf. asterotricha refers to a Colombian collection with branched hairs that is morphologically similar to W. asterotricha accessions from Costa Rica. The accession labeled W. sp. 3858 from southwestern Colombia was identified in the field as W. killipiana; upon further study, we think it probably does not belong to W. killipiana, but its identity is currently unknown. Witheringia cf. solanacea refers to accessions from Costa Rica that are similar to W. solanacea but have 5-merous flowers, are self-compatible, and may represent a distinct taxon (Stone and Jenkins 2008; Bohs 2015). Finally, the name W. coccoloboides is used provisionally here. Although material sequenced here from Costa Rica and Panama is identified as W. coccoloboides using Bohs (2015), the type of W. coccoloboides is from Colombia; ongoing studies indicate that there may be more than one species represented in this morphological complex.

Outgroup taxa were chosen with reference to Särkinen et al. (2013), focusing on species within the tribe Physalideae: six from subtribe Physalidinae (Chamaesaracha, Leucophysalis, Physalis, and Tzeltalia) and two from subtribe Iochrominae (Iochroma and Saracha). Most outgroup taxa were previously sequenced for all three loci (Appendix 1) and these sequences were retrieved from GenBank. For the waxy locus, full-length sequences in GenBank were available only for Leucophysalis and Saracha.

DNA Extraction, Amplification, and Sequencing—Genomic DNA was extracted from fresh, silica-dried, or herbarium leaf tissue using the Qiagen DNeasy Plant Mini Kit (Qiagen, Valencia California), the CTAB procedure (Doyle and Doyle 1987), or the Qiagen kit with CTAB substituting for the first buffer. PCR was carried out in 25 mL volumes in standard buffer with 1.5 mM MgCl₂, 0.8 mM dNTPs, 0.5 mM each primer, 2 U Taq polymerase, and 10–20 ng genomic DNA. For recalcitrant samples from herbarium tissue, BSA was added at a final concentration of 0.5 g/L.

We used primers c and f and standard cycling conditions to amplify the chloroplast trnL-trnF region (Taberlet et al. 1991). Amplification of the ITS region was done using primers ITSleu1 (Andreasen et al. 1999) and ITS4 (White et al. 1990), and standard cycling conditions (Bohs and Olmstead 2001). For a subset of taxa, these conditions produced either fungal or pseudogene amplicons, and a more specific primer was required. Primer ITSSol1 was designed based on GenBank records for Solanaceae sequences of the 18S region from Capsicum frutescens L. (JF766710), Lycium barbarum L. (JN8354580), Physalis heterophylla Nees (KT179698.1), and Solanum carolinense L. (KT179700). In each case, the sixth base was a T rather than a C as in ITSLeu1. The ITSSol1 primer sequence is therefore 5′-GTC CAT TGA ACC TTA TCA TTT AG-3′. In addition, the annealing temperature was increased to 55°C, and 5% DMSO was added to the PCR reaction. Amplification of the granule-bound starch synthase gene (GBSSI or waxy) was carried out as in Levin et al. (2006), amplifying the gene in two sections. Primers waxyF and 1171R (Walsh and Hoot 2001) were used to amplify from exons two to seven; primers 1058 and 2R were used to amplify from exons six to ten (Levin et al. 2006). In cases where spurious bands appeared, the magnesium concentration was reduced to 1.5 mM and annealing temperatures for the touchdown PCR were raised two degrees, to begin at 54°C and end at 50°C.

PCR products were cleaned using the QIAquick PCR purification kit (Qiagen, Valencia California). Sequencing was done in both directions using an ABI 3130 automated sequencer (Applied Biosystems, Foster City California). For trnL-trnF and waxy, primers used for sequencing were the same as those used in initial amplifications. For ITS, internal primers ITS2 and ITS3 (White et al. 1990) were also used in sequencing. In addition, we designed a new internal primer based on sequences from a range of Solanaceae species for the 5.8S region. The ITSSol3 primer differs from ITS3 in having a T rather than a C in the fourth-to-last position (5′-GCA TCG ATG AAG AAC GTA GC -3′). Sequences were obtained for all accessions at all loci except for ITS for Brachistus knappiae (Appendix 1). In this case, repeated efforts produced pseudogene amplicons as revealed by their GC content < 57%, in contrast to the 62–69% GC content found for ITS amplicons in this group. Accordingly, the ITS sequence for B. knappiae was coded as missing data in the matrix. The data matrix from the combined analysis as well as the individual trees from each of the three loci are deposited in Dryad (see Stone et al. 2024).

Phylogenetic Reconstruction—Sequences were aligned using Geneious 11.1.4 (Kearse et al. 2012) and adjusted by eye. Phylogenetic analysis was carried out using Bayesian inference (BI; Mr. Bayes 3.2; Ronquist et al. 2011), with gaps treated as missing data.

The appropriate nucleotide substitution rate model for each locus was evaluated by the Akaike Information Criterion (AIC) as implemented in JModeltest2 (Guindon and Gascuel 2003; Darriba et al. 2012). The GTR + I model had a delta AIC < 2 for all three loci and was applied to all, with each locus treated as a separate partition. Two simultaneous runs of four MCMC chains were run for 20 million generations, each starting with a random tree and saving every 1000 generations. Convergence among independent runs was indicated by standard deviation of split frequencies < 0.01 and by Potential Scale Reduction Factors near 1.0. The first 25% of the trees were discarded as burn-in before creating a majority-rule consensus tree. Trees were displayed using IcyTree (Vaughan 2017).

Incongruence between the individual trees from each locus and the combined tree was evaluated using posterior probability (PP) values of ≥ 0.95 indicating well-supported incongruence.

Results

In total, 84 new sequences were obtained: 28 for ITS, 26 for trnL-trnF, and 30 for waxy, with the remaining sequences obtained from GenBank (Appendix 1). The combined data matrix contained 3640 aligned characters: 649 for ITS, 1216 for trnL-trnF, and 1775 for waxy. According to Bayesian Inference, most branches were supported by PP of 1.0 (Fig. 1). There were a few well-supported conflicts between phylogenetic hypotheses constructed from the three individual loci and the combined tree (Figs. 1, S1–S3). These concerned the positions of W. mexicana (sister to a well-supported clade of 24 taxa in the waxy tree and part of a larger clade containing 30 taxa in the combined tree) and B. nelsonii (sister to a well-supported clade of 23 taxa in the waxy tree and part of a larger clade containing 28 taxa in the combined tree). Conflicting positions were also seen in the relationships among the three accessions of W. meiantha and among the accessions of W. solanacea from Costa Rica and Colombia in the trnL-trnF vs. waxy and waxy vs. combined trees. We present the individual and combined data trees for comparison but consider the combined tree (Fig. 1) as the most complete phylogenetic hypothesis for relationships within Witheringia and Brachistus at this time.

Fig. 1.

Majority-rule consensus tree for Witheringia, Brachistus, and outgroup taxa based on the combined data set from trnL-trnF, ITS, and waxy sequences. Thick branches represent Bayesian posterior probabilities greater than or equal to 0.98. Country code is indicated for species accessions spanning geopolitical boundaries (BO = Bolivia; BZ = Belize; CO = Colombia; CR = Costa Rica; MX = Mexico; PN = Panama). Accession numbers are indicated for species with more than one accession per country.

The phylogeny recovered previously documented relationships among Witheringia and outgroup taxa published by Olmstead et al. (2008), Särkinen et al. (2013), Zamora-Tavares et al. (2016), and Deanna et al. (2019), with Witheringia + Brachistus included with other members of the subtribe Physalidinae. Where sampled, Tzeltalia and Schraderanthus viscosus (Schrad.) Averett [as Leucophysalis viscosa (Schrad.) Hunz. in some trees] were closely related to the two genera, although their positions varied depending on the taxa sampled and regions sequenced.

In our trees, Witheringia and Brachistus together form a monophyletic group (the Witheringia clade). Brachistus species appear to have originated multiple times within the Witheringia clade, but several of these branches have low support. The Mexican endemic species W. stellata and W. mexicana are the earliest diverging lineages of the Witheringia clade. Beyond these, B. knappiae, B. nelsonii, and B. stramoniifolius form a paraphyletic group with W. cf. solanacea. The remainder of the sampled species form two well-supported clades. The first includes W. asterotricha, W. solanacea, W. mortonii, W. sp. 3858, and W. cf. asterotricha from Colombia. Within this clade, all accessions of W. asterotricha and W. solanacea form a well-supported group (the Solanacea clade of Fig. 1). The second clade includes W. coccoloboides, W. correana, W. macrantha, W. maculata, and W. meiantha. All accessions of W. coccoloboides, W. maculata, and W. meiantha form monophyletic groups, but accessions of W. correana and W. macrantha are interspersed throughout this clade.

Discussion

Taxonomic Implications—The phylogenetic framework established here provides a useful construct for the taxonomic treatment of this challenging genus. Most importantly, it resolves the longstanding debate about the status of Brachistus. Brachistus was initially described as a genus distinct from what is now Witheringia based on its lobed (versus truncate) calyx (Miers 1849b). Our phylogeny shows that Brachistus is nested within Witheringia, and, although some of these nodes have only moderate support, it is clear that Brachistus is not a monophyletic group. At best, it is paraphyletic and should be merged within Witheringia. The characters used to separate Brachistus from Witheringia (lobed and accrescent calyces in Brachistus vs. unlobed and non-accrescent calyces in Witheringia) have arisen several times within this clade, as was also shown in the analyses of Deanna et al. (2019). A list of Brachistus species now included in Witheringia is given below, with one new combination, as well as an enumeration of Brachistus names now attributed to other genera.

The phylogeny verifies the similarity of W. asterotricha and the widespread W. solanacea, which together occupy a well-supported clade. These taxa are interfertile, and several authors have suggested that perhaps they should be synonymized (Hunziker 1969; Bohs 2000). Nevertheless, the two W. asterotricha sampled here form a well-supported monophyletic group. Furthermore, W. asterotricha and W. solanacea are morphologically and perhaps geographically distinct, so we advocate keeping them separate at this time. Witheringia solanacea is exceedingly widespread, ranging from Mexico to Andean South America. It is also widely collected and displays substantial morphological variation across its range. In our phylogeny, accessions from Costa Rica and Colombia form a single clade, but the placement of the accession from Bolivia is ambiguous. More sampling will be required to determine whether southern accessions belong to the same species as northern collections.

The phylogeny sheds partial light on the persistent confusion among W. maculata, W. macrantha, and W. correana. These taxa differ in several characteristics but are similar in all having an elongated peduncle, which is lacking in the other taxa of the Witheringia clade. In our phylogeny, the sampled accessions of W. maculata form a well-supported clade distinct from W. macrantha, thus contradicting their synonymization by D′Arcy (1973). Witheringia maculata is also clearly morphologically distinct from these other two taxa by its possession of a frilled calyx visible both in flower and in fruit. However, the three accessions of W. macrantha do not cluster together on the tree and distinguishing between W. macrantha and W. correana morphologically is difficult. Bohs (2015) treated these taxa as possible synonyms, differentiating between them by the size of the flowers, specifically the calyces, which are easily compared in most herbarium specimens. A major challenge in delineating these two species is the great variation within each (Fig. 2). Individuals within W. macrantha may bear flowers that have no spots, a pattern of green spots, or dense purple spots at the base of corolla lobes. Similarly, individuals within W. correana may bear spotted or unspotted flowers. In the phylogeny, accessions of these two species form clades with either W. meiantha or W. maculata. Even within these subclades, accessions of these two species do not form monophyletic groups. Further study will be required to determine whether these phylogenetic relationships reveal cryptic species within one or more of these taxa or whether this is a case of incomplete lineage sorting in a recently diverged group.

Finally, the phylogeny supports the existence of two new species. The accessions denoted as W. cf. solanacea were collected in Costa Rica and resemble W. solanacea except for their 5-merous flowers (vs. typically 4-merous flowers in W. solanacea), more woody habit, and more ovate leaves with cordate bases. Pollination studies have shown these plants to be self-compatible, in contrast to most W. solanacea accessions that have been tested (Bohs 2000; Stone and Jenkins 2008). Bohs (2015), in the treatment of Witheringia for Costa Rica, noted that these geographically and morphologically distinctive populations may represent a distinct species. The molecular results (Fig. 1) support this view; the two accessions sampled of this form do not cluster with other W. solanacea collections from Costa Rica, Colombia, and Bolivia.

The accession W. cf. asterotricha was collected in southeastern Colombia in Dept. Caquetá near Florencia. Morphologically, it is most similar to W. asterotricha, a species known only from Central America. Although this could indicate a geographical disjunction, the Witheringia species of Colombia are poorly known, and the identification of this collection was regarded as provisional. This collection does not cluster with W. asterotricha accessions from Central America in the combined tree (Fig. 1) or in any of the individual gene trees (Figs. S1–S3) and apparently represents an undescribed species.

Character Evolution—The Witheringia phylogeny can shed light on the evolution of several morphological characters. There has evidently been a transition from woody to herbaceous habit; early diverging taxa such as W. stellata, W. mexicana, B. knappiae, B. nelsonii, W. cf. solanacea, and B. stramoniifolius are usually small trees or subshrubs (> 2 m in height) with somewhat woody trunks, whereas the remainder of the Witheringia clade consists of herbaceous species that are typically less than 2 m tall. Several of these species (B. knappiae, B. stramoniifolius, and maybe B. nelsonii) can have leaf blades with lobed margins, in contrast to the entire leaf margins seen in most species of the Witheringia clade. Interestingly, the monotypic physaloid genus Schraderanthus Averett can be a woody shrub with lobed leaf blades. An ITS sequence of Schraderanthus viscosus was included in the phylogenies of Deanna et al. (2019) and it emerged in a clade with species of Witheringia and Brachistus. Unfortunately, we were not able to include Schraderanthus on our trees. Data available in GenBank include an ITS sequence the same length as ours (AY665848), but ITS is not useful in resolving that part of the tree (Fig. S1). Waxy is a more useful locus, but the waxy sequence in GenBank (AY665927) contains only 325 bp of nucleotides, spanning exon and intron 9. Obtaining additional genomic data for Schraderanthus should be a priority for resolving relationships in the Witheringia clade and in the broader context of the tribe Physalideae.

Fig. 2.

Variation within and among Costa Rican accessions of W. correana (A–C) and W. macrantha (D–F). Both species have flowers and fruits borne on peduncles (B, E), and dense stiff hairs throughout. Flowers may be spotted or unspotted in plants of either species. Flower size is variable in both taxa, but flowers are consistently smaller in W. correana. Scale bars = 1 cm. A, B. Stone & Flores 1543, Limon Province; C. Stone & Flores 1545, Guanacaste Conservation Area; D, E. Photo voucher only but cf. Bohs 2512, Monteverde; F. Stone & Flores 1548, Las Tablas. Photos by J. Flores.

Calyx lobing and accrescence have evolved repeatedly within this group. The presence of toothed or lobed rather than truncate calyces has previously been used to delineate Brachistus as a separate section within Witheringia (Hunziker 1969) or as a separate genus (D′Arcy et al. 1981). The emergence of multiple Brachistus lineages within Witheringia indicates that calyx lobing is more evolutionarily labile than previously thought, presumably arising at least twice within the group. Calyx accrescence also displays evolutionary lability. Witheringia clade species with accrescent calyces include B. stramoniifolius, W. correana, and W. macrantha, and B. nelsonii has accrescent-inflated calyces that completely envelop the ripe fruit. In an analysis of character-state evolution across the Physalideae, Deanna et al. (2019) inferred that accrescent calyces have arisen three times within the Witheringia/Brachistus clade. This estimate is consistent with our more fully sampled phylogeny.

Biogeography—Witheringia occupies a pivotal biogeographic position in the evolution of the physaloid genera, near the juncture where the ancestors of the large genus Physalis are hypothesized to have dispersed from the Andean region to current-day Mexico and thereby the North American continent (Zamora-Tavares et al. 2016). This scenario is supported by our finding that the lineages represented by the two narrow Mexican endemics, W. stellata and W. mexicana, diverged at the base of the Witheringia clade. Early diverging lineages also include B. stramoniifolius, B. nelsonii, B. knappiae, and W. cf. solanacea. Brachistus stramoniifolius ranges from Mexico to Panama, with the highest density of collections in southern Mexico, Guatemala, and El Salvador. Brachistus nelsonii occurs primarily in Mexico, Belize, and Guatemala. Brachistus knappiae, in contrast, is a narrow endemic that has been found only in the Talamanca mountains near the border of Costa Rica and Panama, and W. cf. solanacea appears to be a Costa Rican endemic. Thus, the clade may have originated in Mexico and subsequently dispersed southward over a landscape characterized by mountainous and geologically dynamic terrain (Coates and Obando 1996). However, recent field and herbarium work in South America, particularly Colombia, has unearthed numerous accessions that may represent undescribed species. More accurate species delimitation, thorough inter- and intraspecific sampling, and a more completely resolved phylogeny is needed for a more complete picture of the evolutionary relationships of the Witheringia clade.

In sum, the phylogeny presented here provides a systematic foundation for a taxonomic revision of Witheringia and its relatives. It also indicates avenues for future study of character evolution and the biogeographic context of diversification in the Witheringia clade.

Transfer of Brachistus to Witheringia—The phylogenetic results presented here show that four currently recognized species of Brachistus are nested within the genus Witheringia, which was published earlier and has priority. Three of these already have published names within Witheringia; one new combination is required to transfer Brachistus knappiae to Witheringia. Athenaea affinis C.V. Morton, the basionym of Witheringia affinis (C.V.Morton) Hunz. is lectotypified here. In the following section the accepted name in Witheringia is listed first, followed by the basionym (if necessary), and any homotypic synonyms.

A list is also provided for Brachistus names now assigned to other genera under Excluded Names.