Genomic DNA was extracted from approximately 20 mg of dried leaf tissue using the Genomic Mini Plant Kit (A&A Biotechnology, Poland) following the manufacturer’s instructions. Polymerase chain reaction (PCR) amplifications were performed in a total volume of 25 µl containing 1× buffer, 2 mM MgCl₂, 0.2 mM dNTPs, 0.2 µM of each primer, 1.0 U of Color Perpetual Taq DNA polymerase (EURx Ltd. Poland), and DNA template (~30 ng). Amplified products were purified using the Wizard SV Gel and PCR Clean-UP System (Promega, USA).

The same primer pairs were used for both amplification and sequencing. Amplification of the ITS region (ITS1–5.8S–ITS2) was performed using primers 17SE and 26SE (Sun et al., 1994), while for the matK region, primers −19F (Wilson et al., 2000) and 1326R (Cuénoud et al., 2002) were used. PCR thermal cycling conditions began with an initial denaturation at 94°C for 4 min, followed by 30 cycles of denaturation at 94°C for 45 s, annealing at 52°C for 45 s, and an extension step at 72°C for either 1 min (ITS) or 2 min (matK) depending on the target region. Sequencing of the purified PCR products (~80 ng and 2.5 μM primer) was performed externally by Macrogen (Amsterdam, the Netherlands). Both DNA strands were sequenced to ensure accuracy in base calling. The sequences were edited using FinchTV v. 1.4.0 (Geospiza, Inc.), and the two complementary strands were assembled with AutoAssembler (ABI).

DNA sequences were aligned using MAFFT v.7 (Katoh and Standley, 2013). Separate alignments were prepared for the ITS region (850 bp) and matK gene (1,361 bp). For each dataset, the best-fit evolutionary model was estimated using IQ-TREE v.2. Due to the absence of significant topological conflicts, nuclear and plastid DNA data were combined into a single concatenated matrix (2,211 bp) for further analyses. Additional details about the datasets analyzed are presented in Table 1 .

Phylogenetic reconstruction was performed using the maximum likelihood (ML) method in IQ-TREE v.2 (Minh et al., 2020). Clade support was assessed through a nonparametric bootstrap analysis with 1,000 replicates in the same software. Additionally, divergence time estimation within the studied group was conducted using BEAST v.1.8.4, executed on the CIPRES Science Gateway server (Miller et al., 2010). The analysis was run in two independent replicates, each consisting of 40 million generations. An uncorrelated relaxed molecular clock (lognormal) and the Yule model of speciation were applied. Two calibration points, following Givnish et al. (2015), were used: 27.34 million years ago (Mya) for the Polystachyinae node and 21.21 Mya for the ingroup. The results from both independent runs were assessed in Tracer v.1.6 to ensure convergence and subsequently combined in LogCombiner v.1.8.4 where 25% of the initial generations from each run were discarded as burn-in. The final maximum clade credibility (MCC) tree was generated using TreeAnnotator v.1.8.4.

Our investigations confirm the previous phylogenetic framework for Vandeae and especially for angraecoid orchids (Simo-Droissart et al., 2018). Ambrella is embedded in the clade along with the so-called vitsyangraecoides (Farminhão et al., 2021), which includes a few taxa found mainly in Madagascar and also in eastern Africa. The phylogenetic position of Ambrella, along with Cryptopus Lindl. and Oeonia Lindl., is strongly supported. Interestingly, both of these genera are very different from Ambrella ( Figure 3 ). In both, we can observe elongated, ascending stems with loosely spaced leaves, at least several-flowered, elongated inflorescences, and wide-open flowers with a freely accessible gynostemium and spur entrance. The generative parts in representatives of both genera relate in structure to Angraecum Bory s.str., i.e., the rostellum is dome shaped, with a weakly marked middle lobe. In contrast, in Ambrella, Cryptopus, and Oeonia, the pollinia are attached to two tegulae, at the base of which separate viscidia are present. The genus Neobathiea Schltr. is then sister to them, followed by the genus Lemurella Schltr., with the East African Lemurangis humilis (Summerh.) Szlach., Mytnik & Grochocka, a representative of the Angraecum complex (=A. humile Summerh.), and finally Beclardia A. Rich., which is sister to all of them. Of the genera mentioned, elongated stems with significantly distant leaves can be observed in Neobathiea (e.g., N. spatulata H. Perrier), but already in N. perrieri (Schltr.) Schltr. or N. hirtula H. Perrier and Lemurella, the leaves have a similar form and distribution on the stem as in Ambrella. One could speculate that the primary feature in this evolutionary line, maintained in Ambrella, is the shortened stem and its relatively dense foliage. In Ambrella ancestors, there is a tendency to elongate the stem by making it ascending. This trait is undoubtedly an adaptation to habitat conditions. In this clade, a clear differentiation in the morphology of the floral parts is observed, while the structure of the generative parts is maintained. In all these genera, with the exception of Beclardia, the short, dome-shaped rostellum, with a weakly marked central patch, is present, and the pollinia are attached to two tegulae, with an independent viscidium on each. In B. macrostachya (Thouars) A. Rich., the rostellum is strongly elongated, and the equally elongated tegula is attached to a single viscidium. Both pollinia are attached to the apical, disc-shaped extended part of the tegula.

Such a resolution of phylogeny is consistent with previous studies (Simo-Droissart et al., 2018; Farminhão et al., 2021). Interestingly, the vitsyangraecoides group is not most closely related to the other Malagasy angraecoids and is sister to the Afroneotropical clade. Thus, Ambrella is not closely related to the Oeoniella and Sobennikoffia compared in the protologue. In both genera, the stem is elongated, as are the leaves, and the inflorescences are multiflowered. The lip in Oeoniella is tubular, with a relatively narrow spur entrance and access to the gynostemium. The structure of the gynostemium in the two genera is different. In Oeoniella, the lateral lobes of the rostellum are elongate, the median lobe is strongly reduced, and the pollinia are attached to two tegulae that connect to a single viscidium. In Sobennikoffia, the lateral lobes of the rostellum are strongly reduced, with both pollinia attached to a common tegula, to which two independent viscidia are attached on the other side.

Despite being related to other taxa than assumed, Ambrella is morphologically a completely separate taxon. Vegetative traits in all Angraecinae are confusing, and taxonomic distinctiveness cannot usually be recognized, but flower architecture resolves this problem. The inflorescence of Ambrella longituba, with one to three flowers, does not resemble the multifloral inflorescence of Cryptopus and Oeonia or uniflowered inflorescence in Neobathiea. Finally, the tubular labellum, with hairs inside and very short spur observed in Ambrella, is completely different from the spurless flowers of Cryptopus and Oeonia, which have a flat labellum, at most slightly tubular at the base, through fusion with the gynostemium. Moreover, the Neobathiea has a very long, downward-hanging spur, which is typical of many angraecoid orchids.

Ambrella longituba falls within a clade of taxa that separated in the Early Miocene (ca. 8.8–3.3 Mya), and diversification within it continued up to the Pliocene. The Miocene period is a time of significant changes in the vegetation of Africa (Couvreur et al., 2021). Due to the cooling of the climate in the middle Miocene, many forest formations experienced decline and division especially on the continent, but it could also be the reason for the lack of diversification or extinction in Madagascar (Buerki et al., 2013). The estimated divergence time for the most recent common ancestor of Ambrella and sister genera falls within the Late Miocene to Early Pliocene, approximately 8.83 to 3.31 Mya. This is a time of ongoing changes in the vegetation of Madagascar and the development and expansion of humid forests (Buerki et al., 2013), which provide the main habitat for Ambrella and closely related taxa, promoting their cladogenesis.

Representatives of the Angraecinae are fairly well known for their mechanisms of pollination by long-proboscis insects (Lepidoptera), which is related to the presence in many species of a long spur containing nectar (Micheneau et al., 2014; Netz and Renner, 2017). However, other groups have also been observed as effective pollinators in the subtribe such as Diptera and Hymenoptera (Singer and Cocucci, 1999) for Campylocentrum aromaticum Barb.Rodr. or even Orthoptera (Micheneau et al., 2010) for Hadrangis cadetii (Bosser) Szlach., Mytnik & Grochocka (=Angraecum cadetii Bosser).

Nothing is known, so far, about pollination in Ambrella longituba. According to the note of Perrier left under the protologue (1924), it may be associated with a dependence on a narrow group of potential pollinators. The insect that inserts its head into a small window at the base of the lip to pump nectar from the spur can only exit through a window covered by two pollen-connecting stipes. The narrowness of this window must allow pollination of this flower only by a few very exceptional insects. The role of the hairs inside the tubular labellum remains unclear. Such a feature is often an adaptation to zoogamy, and in many plants, such a closed pathway to the reproductive organs is evidence of trap flowers (Case and Bradford, 2009; Matallana-Puerto et al., 2024). The hairs may also be part of the attraction and food for pollinators (Davies et al., 2002).

Finally, self-pollination cannot be eliminated, which has been confirmed for other closely related taxa, such as Cryptopus elatus, Oeonia rosea, and Beclardia macrostachya (Thouars) A.Rich (Agnew, 1986; Jacquemyn et al., 2005). Only for Neobathiea grandidieriana was pollination by Lepidoptera observed (Nilsson et al., 1987), which is obviously related to the presence of a spur, which is also present in Ambrella, but certainly not in this form. Therefore, pollination in Ambrella remains an interesting question to investigate.