Taxon sampling

The present study included ten of the 11 species of Disocactus according to the classification proposed by Hunt & al. (2006) and incorporated a comprehensive sampling of ten terminals for five subspecies of D. speciosus (Cav.) Barthlott and two for D. ackermannii (Haw.) Ralf Bauer, D. biformis (Lindl.) Lindl., D. macranthus (Alexander) Kimnach & Hutchinson and D. phyllanthoides (DC.) Barthlott. Representatives from Epiphyllum, Hylocereus, Pseudorhipsalis, Selenicereus and Weberocereus were included to test the monophyly of the genus because they were recovered as part of a single clade in previous phylogenies (Nyffeler 2002). Outgroups were sampled based on previous phylogenetic analyses of the family (Nyffeler 2002; Arias & al. 2005; Hernández-Hernández & al. 2011) and included representative species of Acanthocereus Britton & Rose (Lodé 2013), Lemaireocereus Britton & Rose and Pereskia. This study involved 315 new sequences; their voucher information and GenBank accession numbers are listed in the Appendix.

Amplification and sequencing

Phylogenetic reconstructions for Cactaceae have mainly employed chloroplast molecular markers. Nuclear markers suitable for cacti have yet to be developed because the existing nuclear markers provide a lower number of informative sites compared with chloroplast markers, and they do not exhibit polymorphism as has been shown for ITS in other groups of cacti, e.g. Lophocereus (A. Berger) Britton & Rose (Hartmann & al. 2001), Mammillaria (Harpke & Peterson 2006), Hylocereus and Selenicereus (Plume & al. 2013). A combination of different chloroplast markers is necessary to increase the resolution of the phylogeny (Korotkova & al. 2010). Within the diverse chloroplast markers, it has been proposed that introns outperform spacers and coding regions in terms of phylogenetic utility (introns > spacer sequences > coding regions) (Korotkova & al. 2011). Six molecular markers were used: two introns (rpl16 and trnL-F), two spacers (psbA-trnH and trnQ-rps16) and two coding regions (matK and ycf1).

Collected samples were dried with silica gel and stored at -20 °C. Total DNA was extracted using the EZ-10 Spin Column Plant Genomic DNA Minipreps kit (Bio Basic Inc., Ontario, Canada) according to the manufacturer's protocol, but the incubation time in tissue lysis buffer was modified to 120 minutes. Total DNA was stored at -20 °C. Specific regions were amplified using a standard polymerase chain reaction (PCR) protocol in a total volume of 25 µL containing 2.5 µL of 10× buffer, 0.5 µL of dNTPs (10 mM), 1 µL of BSA, 0.3 µL of each primer (10 mM), 0.75 µL of MgCl₂, 0.125 µL of Taq, 19.125 µL of DEPC-treated water and 0.75 µL of DNA. The matK exon was amplified using the primers matK_3F_kim and matK_1R_ kim (CBOL Plant Working Group, 2009) by initially denaturing the DNA for 2 min at 94 °C, followed by 30 cycles of 30 s at 94 °C, 40 s at 48 °C and 40 s at 72 °C, with a final extension of 7 min at 72 °C. The rpl16 intron was amplified using the primers rpl161F and rpl163R (Hernández-Hernández & al. 2011) by initially denaturing the DNA for 5 min at 94 °C, followed by 26 cycles of 1 min at 94 °C, 50 s at 55 °C and 2 min at 72 °C, with a final extension of 4 min at 72 °C. The trnL-trnF region was amplified using the primers “C” and “F” designed by Taberlet & al. (1991). The PCR programme employed for this region consisted of 2 min at 94 °C, 29 cycles of 30 s at 94 °C, 30 s at 52 °C and 1 min at 72 °C, and a final extension of 7 min at 72 °C. The psbA-trnH intergenic spacer was amplified using the primers psbA (Sang & al. 1997) and trnH (Tate & Simpson 2003), with initial denaturation of the DNA for 2 min at 94 °C, followed by 29 cycles of 30 s at 94 °C, 30 s at 52 °C and 1 min at 72 °C, and a final extension of 7 min at 72 °C. The trnQ-rps16 intergenic spacer was amplified using the primers trnQ2 (Korotkova & al. 2010) and rps16×1 (Shaw & al. 2007), with initial denaturation of the DNA for 2 min at 95 °C, followed by 35 cycles of 30 s at 95 °C, 1 min at 55 °C and 1 min at 72 °C, and a final extension of 10 min at 72 °C. The ycf1 region was amplified using the primers ycf1-4182F and ycf1-5248R (Franck & al. 2012), with initial denaturation of the DNA for 3 min at 94 °C, followed for 35 cycles of 1 min at 94 °C, 1 min at 50 °C and 2 min at 72 °C, and a final extension of 5 min at 72 °C. The unpurified PCR products were sequenced at the University of Washington (UW-High Throughput Genomics Center,  http://www.htseq.org/). All regions were sequenced using amplification primers. Sequence assembly and editing was carried out using Sequencher 4.8 (GeneCode).

Phylogenetic analysis

The sequences were aligned using MAFFT (Katoh & al. 2002), and the alignments were then manually inspected. In all of the alignments, the extremes were eliminated due to the presence of ambiguities. Insertions and deletions (indel) were coded following the simple indel coding method suggested by Simmons & Ochoterena (2000) using the Indel Coder option of SeqState v. 1.40 (Müller 2005). Maximum parsimony (MP) phylogenetic analyses were carried out for each of the markers to demonstrate their congruence (trees with similar topologies) and concatenate them in a single matrix and to identify markers that maximize the resolution of the tree. MP phylogenetic analysis with the concatenated marker matrix without hotspots (highly variable regions) was carried out using PAUP (Swofford, 2002). A heuristic search with 15 000 replicates was conducted; the tree bisection and reconnection (TBR) branch-swapping algorithm was employed, and Pereskia lychnidiflora DC. was designated as an outgroup. A second MP phylogenetic analysis was performed with the concatenated marker matrix and coded indels and without hotspots using the above parameters. For branch support, 10 000 jackknife (JK) replicates and 10 000 bootstrap (BS) replicates were employed. Bayesian inference (BI) analysis was carried out using MrBayes 3.2 (Huelsenbeck & Ronquist 2001). The data were partitioned, and each partition used GTR+I+G as a substitution model, pursuant to the results obtained with jModeltest2 (Darriba & al. 2012). Two independent runs were conducted with four chains each, and six million generations were run, with the chains sampled every 1000 generations. Chain convergence was observed using Tracer (Rambaut & Drummond 2009). The first 20 % of trees were discarded as the burn-in. The remaining trees were then used to build a 50 % majority rule consensus tree. The alignment is available at  http://purl.org/phylo/treebase/phylows/study/TB2:S18804.

Table 3.

Phylogenetic analysis characteristics and numerical results for each sequence data set and the combined data set.

The contribution of the the markers and coded indels

The region matK provided the lowest number of informative sites. This region was also less variable in other groups of Cactaceae, such as in the genus Pfeiffera (Korotkova & al. 2010), the tribe Rhipsalideae (Korotkova & al. 2011) and the tribe Cacteae (Vázquez-Sánchez & al. 2013). Two sites supported the relationships at the genus level (214 for the genus Pseudorhipsalis, 706 for the genus Disocactus; Table 4), which shows that matK may, to a lesser degree, contribute information to resolve relationships at the genus level and not only at the family level. Conversely, markers with high evolutionary rates, such as the introns rpl16 and trnL-F and the spacer regions psbA-trnH and trnQ-rps16 (Borsch & Quandt 2009; Korotkova & al. 2011), provided 31 % of informative sites and allowed the establishment of relationships at primarily the genus and species levels in the genera Epiphyllum and Pseudorhipsalis. The region ycf1 contributed 49 % of informative sites. This region is also highly variable in other groups of cacti, such as Opuntia (L.) Mill. (Majure & al. 2012), Harrisia Britton (Franck & al. 2013) and Copiapoa Britton & Rose (Larridon & al. 2015). The region ycf1 produced the tree with the highest resolution, supporting the relationships at the genus and species levels; however, it did not individually resolve the relationships between close species. Moreover, the coded indels provided 13 % of the PI sites. The incorporation of coded indels in phylogenetic reconstructions is a frequent practice in Cactaceae (e.g. Nyffeler 2002; Butterworth & Wallace 2004; Korotkova & al. 2010, 2011; Calvente & al. 2011a, b; Franck & al. 2013; Vázquez-Sánchez & al. 2013; Bonatelli & al. 2014). Furthermore, their inclusion improved the resolution and support values for the tree compared with the tree that did not include the coded indels (data not shown), as previously suggested (Simmons & al. 2001).

Genus Disocactus s.str.

Based on our results, Disocactus consists of 13 known species, including Epiphyllum anguliger, E. crenatum and E. lepidocarpum (Fig. 2, 3). None of the three subgenera recognized by Hunt & al. (2006) (Table 2) was recovered; however, three subclades were recognized with some support and three sister species relationships were observed. Subclade 1D includes the relationship between D. eichlamii (Weing.) Britton & Rose and D. quezaltecus (Standl. & Steyerm.) Kimnach, with strong support (99 % JK, 100 % BS MP; 1 PP BI) and corroborates the hypothesis suggested by Kimnach (1959, 1993) of the closeness of these two species based on exhibiting flowers with straight tepals and exerted stamens and styles. Moreover, D. speciosus was not recovered as a monophyletic species. Disocactus speciosus subsp. bierianus Ralf Bauer, subsp. cinnabarinus (Eichlam ex Weing.) Ralf Bauer and subsp. speciosus form a group with moderate support (77 % JK, 89 % BS MP; 1 PP BI; Fig. 2, 3), and a poor resolution is observed within this group, which we consider to be a reflection of the morphological issues presented by the species. Our results partially support the proposal of Bauer (2003), as only three of the five subspecies were recovered in the clade mentioned above. The variability of the stems, spines and flowers shown by these three subspecies is high, and they are only distinguished by subtle morphological differences (Bauer, 2007). Their shared morphological characteristics include stems with rigid spines, large red or orange flowers (12–17 cm) and a pericarpel and receptacle tube with rigid spines. They are broadly distributed in W Mexico (Durango to Oaxaca) and C Guatemala (the Sierra de los Cuchumatanes), growing in Pinus forests and tropical montane cloud forest (Fig. 4). Disocactus speciosus subsp. aurantiacus (Kimnach) Ralf Bauer was not recovered as part of the group that includes the three subspecies mentioned above in either analysis (MP or BI) (Fig. 2, 3, subclade 1D). This position in the topologies is congruent with the morphological differences exhibited by D. speciosus subsp. aurantiacus, which is characterized by flexible and pendulous stems as well as longer hairs, instead of spines, on the stems and flowers. This taxon is distributed in the tropical montane cloud forests of El Salvador, Honduras and Nicaragua (Fig. 4). Bauer (2003) included it in D. speciosus because it also exhibits ribbed stems and large flowers (12.5–15.5 cm). Although our results do not completely clarify the relationship of this taxon within subclade 1D (Fig. 2, 3), the molecular autapomorphies revealed in this study (substitution in ycf1, one indel in psbA-trnH, and rpl16; Table 4) in addition to the morphological characteristics listed above indicate that it should be recognized as a separate species, as was originally proposed (Kimnach 1974; Barthlott 1991a). Hunt & al. (2006) classified D. speciosus and D. ackermannii as part of D. subg. Ackermannia because they both develop large red or orange flowers (8–20 cm). However, we indicate herein that the recognized groups of D. speciosus are independently related, but never to D. ackermannii (Fig. 2, 3, subclades 1D, 3D).

Fig. 4.

Geographical distribution of Disocactus taxa based on georeferenced specimen records.

In subclade 2D, Disocactus biformis was recovered as a sister taxon to D. nelsonii (Britton & Rose) Linding. with strong support (100 % JK, 100 % BS MP; 1 PP BI). These species share morphological characteristics including expanded external tepals, a scaly pericarpel and hairs measuring less than 0.5 cm, which allowed Kimnach (1993) to suggest the closeness of the species corroborated in this study. Disocactus biformis, D. eichlamii, D. nelsonii and D. quezaltecus are native to Guatemala, Honduras and Mexico (Chiapas) and are exclusive to tropical montane cloud forests (Fig. 4). Although these four species are included in the section or subgenus Disocactus (Table 2) because they share phylloclades, flowers, and fruits without spines, our results reveal that they do not form a clade. Five indels (two in trnL-F and three in psbA-trnH) plus five substitutions (one in matK, one in psbA-trnH, trnQ-rps16, rpl16 and ycf1) maintain the two groups in independent subclades (Table 4; Fig. 2, 3).

Subclade 3D supports the relationship among Disocactus ackermannii, D. ×kimnachii G. D. Rowley and D. phyllanthoides (97 % JK, 86 % BS MP; 1 PP BI; Fig. 2, 3), in contrast to the proposal made by Hunt & al. (2006). These authors believed D. ackermannii to be more closely related to D. macdougallii (Alexander) Barthlott (not included in this analysis) and D. speciosus, which together comprise D. subg. Ackermannia, whereas D. phyllanthoides was maintained in Nopalxochia (Table 2). However, we identified a relationship that was more consistent with the taxonomic treatment by Bravo-Hollis (1978): with the exception of D. ×kimnachii (a taxon not recognized as a species by Hunt & al. 2006), the two remaining species and D. macdougallii comprise the genus Nopalxochia. Disocactus ackermannii and D. phyllanthoides are native to tropical montane cloud forests in E and S Mexico (Veracruz to Chiapas; Fig. 4), whereas D. ×kimnachii is found only in Costa Rica. The initial hypothesis about the origin of D. ×kimnachii proposed that it is a hybrid between a member of Heliocereus or Nopalxochia and Epiphyllum crenatum (Kimnach 1984), whereas a subsequent hypothesis suggested that the putative parents include Epiphyllum crenatum subsp. kimnachii (Bravo ex Kimnach) U. Guzmán and a member of the group “Heliocereus” (Bauer 2003). Under this proposal and according to our results, D. phyllanthoides may be one of the parents. The three taxa are characterized by stems that branch close to the base, a pericarpel and receptacle tube with colourful scales that increase in size towards the perianth, and areoles with trichomes and flexible spines.

Three terminals of Disocactus speciosus aff. subsp. cinnabarinus from Chiapas (Mexico) were recovered with strong support (96 % JK, 99 % BS, MP; 1 PP BI; Fig. 2, 3). Neither the MP nor BI analysis recovered this taxon as part of D. speciosus, where the terminal from D. speciosus subsp. cinnabarinus from Chimaltenango (Guatemala) is nested, close to the type locality (Weingart 1910). The terminals from Chiapas (Arias 1666, Cruz 18, Cruz 22) were initially identified as D. speciosus subsp. cinnabarinus due to their shared floral morphology (Bravo-Hollis 1978; Bauer 2003), and it was even indicated that the observed differences were due to variations between populations (Bauer 2007). However, the adult stems of the Chiapas samples exhibit a consistent difference in the presence of phylloclades. This morphological difference was observed by Standley & Steyermark (1944) and was used as a basis for describing Heliocereus heterodoxus Standl. & Steyerm. for the region limited by the Tacana Volcano between Mexico (Chiapas) and Guatemala (San Marcos). We identified two indels in the markers psbA-trnH and ycf1 (Table 4) that are exclusive to these three terminals. Based on these molecular characteristics and the morphological difference mentioned above, we believe that the three terminals do not correspond to D. speciosus subsp. cinnabarinus; therefore, we propose that it be recombined as D. speciosus subsp. heterodoxus (Standl. & Steyerm.) M. Á. Cruz & S. Arias.

Epiphyllum anguliger, E. crenatum and E. lepidocarpum were recovered as part of Disocactus s.str. with strong support (Fig. 2, 3). This relationship is supported by seven shared substitutions, which we recognize as the molecular synapomorphies of the clade Disocactus (Table 4). The results do not allow discrimination of the closest relationships between these three species within Disocactus because the two topologies (MP and BI) are incongruent. However, we may recognize a possible relationship between E. crenatum and subclade 3D, which includes Nopalxochia sensu Bravo-Hollis (1978). This hypothesis is based on the shared presence of scales on the pericarpel and the receptacle tube, which are larger towards the perianth, as well as flexible spines. The generation of hybrids has also been reported for both species (D. ackermannii & D. phyllanthoides) (Anderson 2005). In addition, the distribution of E. crenatum complements the species of subclade 3D in México (Veracruz, Oaxaca and Chiapas), although E. crenatum extends towards Guatemala and Honduras. Although the relationship between E. anguliger and E. lepidocarpum is uncertain, the flexible spines observed on the stems and flowers with two sets of stamens are morphological characters shared with Disocactus (Bauer 2009). Epiphyllum anguliger is distributed in W Mexico (Guerrero, Jalisco, Nayarit and Oaxaca), whereas E. lepidocarpum is native to Costa Rica (Fig. 4).

Disocactus and sister genera

The topology of both phylogenies (MP and BI) is highly congruent: the tribe Hylocereeae consists of three groups (Fig. 2, 3). The subtribe Epiphyllinae, as redefined here, includes the genera Disocactus (Clade D), Epiphyllum (Clade E) and Pseudorhipsalis (Clade P), whereas the subtribe Hylocereinae consists of Aporocactus (Clade A), Hylocereus, Selenicereus and Weberocereus (HSW). Acanthocereus represents the third group. This result is partially consistent with the subtribes proposed by Buxbaum ( 1958) and the clades recovered in the phylogenetic analyses of Peniocereus by Arias & al. (2005). The main difference between the two subtribes is that phylloclades are more common in Epiphyllinae, whereas ribs are more common in Hylocereinae. In addition, both phylogenies indicate that Disocactus is more closely related to Epiphyllum and Pseudorhipsalis.

Within the subtribe Epiphyllinae, Disocactus is never recovered as a sister genus to Pseudorhipsalis. In the MP reconstruction, Disocactus is a sister to Epiphyllum and Pseudorhipsalis, whereas in the BI reconstruction, Pseudorhipsalis is sister to Disocactus and Epiphyllum. This classification is consistent with Bauer's argument (2002), which indicates that the distribution, habitat and certain morphological characteristics differ between Pseudorhipsalis and Disocactus. Pseudorhipsalis is distributed from Mexico to Bolivia, but its centre of diversity is found in Costa Rica and Panama, where it grows in low-elevation zones (below 650 m) and hot, humid regions. The primary stem of Pseudorhipsalis is easily distinguished from the secondary stem due to its cylindrical shape; its areoles are close-set, and its flowers are very small (0.7–2.5 cm), funnel-shaped or rotate and are characterized by a short flowering period. The flowers are not oriented towards the light, and the pollen is small (45–55 µm) and polycolpate. In contrast, Disocactus is distributed from NC Mexico to Nicaragua, where it grows in elevated (above 1200 m) tropical montane cloud forests. The primary stem of Disocactus may only be distinguished from the secondary stem in certain species; the distance between the areoles is greater; and the flowers may be small (5.5–6 cm) to large (16 cm) and funnel-shaped. The flowers spend several days in anthesis and are mainly oriented towards the light, which results in a curve in the receptacle tube when they grow on pendulous stems; the pollen is large (70–100 µm) and tricolpate. These morphological differences favour the hypothesis suggested in the BI analysis, in which Disocactus is the sister genus to Epiphyllum, which is distributed from Mexico to Argentina, although its greatest diversity is found between S Mexico and Costa Rica in low zones or at high elevation (0–2300 m) in hot and temperate regions (Bauer 2003). These two genera share phylloclades with a cylindrical base, flowers larger than 2.5 cm, and a scaly pericarpel and receptacle tube. Moreover, our results allow us to exclude Aporocactus from Disocactus and recognize it as an independent genus. Nevertheless, its relationship with the subtribe Hylocereinae merits its own study because only the BI phylogenetic analysis showed high support.

Based on the data from cpDNA sequences, this study provided significant evidence indicating that certain floral morphological characteristics that are used to delimit Disocactus have arisen more than once in Hylocereeae. The diurnal, funnel-shaped and intensely coloured (reds and others) flowers are not exclusive to Disocactus (Fig. 1 ). Based on these data, we hypothesize a convergent origin for the nocturnal, long, funnel-shaped flowers with fight colouration (white) found in Disocactus and Epiphyllum. Therefore, supplementary studies are required to clarify the relationship between Disocactus and its sister genera, as well as its possible origin within the subtribe Epiphyllinae (as it is recognized here). A comparative morphological study of vegetative (e.g. shape and succulence of cotyledons in seedlings; stem shape during the juvenile stage) and reproductive structures (e.g. number of sets of stamens; seed size and micro-ornamentation) is needed to explore the observed synapomorphies and their transformation within the subtribe. The origin of the epiphytic growth within Hylocereeae also needs to be explored based on climbing shape, including the change in stem growth from indeterminate to determinate. A top priority is to combine our molecular data with a set of morphological attributes in a total evidence analysis to corroborate the phylogenetic relationships obtained here and recover the currently unresolved terminal nodes. We also aim to understand the evolution of the highlighted characteristics in subtribe Epiphyllinae.