Muñoz-Rodríguez et al. (2022) used an outdated generic sampling for Malpighiaceae based on Davis and Anderson (2010). Since then, several phylogenetic and taxonomic studies have been published, including the synonymy of several genera and the publication of two new genera for Malpighiaceae (Davis et al. 2020; Almeida and van den Berg 2021). In order to accurately test the phylogenetic placement of Keraunea brasiliensis within Malpighiaceae, we downloaded sequences of the markers matK, rbcL, and ITS stored on GenBank for all 77 currently accepted genera of Malpighiaceae according to POWO (2023) and the sequences of K. brasiliensis (Passos et al. 5263) presented as supplementary material by Muñoz-Rodríguez et al. (2022). Sequences of Elatine L. and Bergia L. (Elatinaceae), the sister group of Malpighiaceae (Davis and Chase 2004; Cai et al. 2016), were also used to root our analysis. Additionally, we performed a secondary analysis to test the placement of K. brasiliensis within the genus Mascagnia (Bertero ex DC.) Bertero, including 18 accepted species (out of 58) with available sequences on GenBank, four species of Amorimia W.R.Anderson as the outgroup, and Ectopopterys W.R.Anderson to root the analysis.

Datasets were compiled, for each marker, using Geneious v.4.8 (Kearse et al. 2012) and aligned using Muscle v.1.0 (Edgar 2004), with subsequent adjustments in the preliminary matrices by visual inspection. Separate and combined analyses of plastid, nuclear, and plastid + nuclear regions were performed using Bayesian Inference (BI) and Maximum Likelihood (ML) criteria for phylogenetic reconstruction. Both model-based methods were conducted with a mixed model (GTR+G+I) and unlinked parameters selected using jModelTest 2 (Darriba et al. 2012), and the analyses were done with MrBayes v.3.1.2 (Ronquist and Huelsenbeck 2003) and raxmlGUI 2.0 (Edler et al. 2021). For the BI, the Markov Chain Monte Carlo (MCMC) was run using two simultaneous independent runs with four chains each (one cold and three heated), saving one tree every 1,000 generations for ten million generations. We excluded 20% of the retained trees as “burnin” and checked for a stationary phase of the likelihood, checking for ESS values higher than 200 for all parameters on Tracer v.1.6 (Rambaut et al. 2014). The clades’ posterior probabilities (PP) were based on the 50% majority rule consensus, using the stored trees, and calculated with MrBayes v.3.1.2 (Ronquist and Huelsenbeck 2003). Bootstrap values are shown above the branches, while the posterior probabilities values are shown below. All datasets and consensus trees are available as supplementary files (https://doi.org/10.6084/m9.figshare.21961550.v2).

Images of the holotype and all isotypes of Keraunea brasiliensis (Passos et al. 5263) were searched in online specimen databases, such as GBIF (https://www.gbif.org), JSTOR (https://plants.jstor.org), Jabot (http://jabot.jbrj.gov.br), Reflora (https://reflora.jbrj.gov.br), and speciesLink (https://specieslink.net). The Kew isotype (K000979156) was consulted in person at the Kew herbarium (Rafael Almeida, Ana Rita Simões, and Martin Cheek), and the Brazilian duplicates (ALCB, CEPEC, HRCB, HUEFS, and SPF; acronyms according to Thiers 2023) were accessed by Rosangela Simão-Bianchini and local collaborators among the staff of the above-cited herbaria. At Kew herbarium, morphological details were photographed using a Leica S9i stereomicroscope with a coupled digital camera.

The topologies of the nuclear, plastid, and combined phylogenetic trees were found to be highly congruent. Hence, we have chosen to discuss the results in light of the combined analysis instead of the individual datasets (for additional information, see supplementary files). “Keraunea brasiliensis” (Passos et al. 5263) was recovered with high support as nested within the genus Mascagnia in the Malpighioid clade by both the individual and combined analyses, using BI and ML inference criteria (Fig. 2). The low support for most relationships within the Tetrapteroid clade is interpreted as the result of missing data between all three molecular datasets. Based on our secondary analysis, “K. brasiliensis” was recovered with high support as sister to Mascagnia cordifolia (A.Juss.) Griseb (Fig. 3). Thus, we compared the DNA sequences of K. brasiliensis to M. cordifolia in the individual matK and rbcL alignments and observed that the DNA sequences of the two species for these two genetic markers had the exact same nucleotide composition (see supplementary files). The rbcL sequence of “K. brasiliensis” was also shorter (600 bp) than that of M. cordifolia and the remaining Malpighiaceae (1,400 bp). Hence, our molecular phylogenetic results support the view that the sequence of the specimen Passos et al. 5263 (K) from Muñoz-Rodríguez et al. (2022) belongs to Mascagnia, very probably representing the species M. cordifolia, given the identical genetic sequences for the barcoding markers matK and rbcL.

Figure 2. 

Consensus tree of the combined analysis based on the markers matK, rbcL, and ITS showing the phylogenetic placement of “Keraunea brasiliensis” (Passos et al. 5263) (highlighted in red) within Malpighiaceae, making the family non-monophyletic. Elatinaceae (highlighted in light grey) represents the outgroup and the root of this analysis. Bootstrap values from the ML are shown above the branches, and posterior probabilities from the BI are shown below the branches. The tree on the left is presented for branch length visualisation. Photographs of Elatine gratioloides A.Cunn. by Melissa Hutchison, Stigmaphyllon angustilobum A.Juss. by Rafael F. de Almeida, and Keraunea spp. by Geovane S. Siqueira.

Figure 3. 

Consensus tree of the combined analysis based on the markers matK and rbcL showing the phylogenetic placement of “Keraunea brasiliensis” (Passos et al. 5263) (highlighted in red) within Mascagnia. Amorimia W.R.Anderson (highlighted in light grey) represents one of the outgroups, and Ectopopterys W.R.Anderson is the root of this analysis. Bootstrap values from the ML are shown above the branches, and posterior probabilities from the BI are shown below the branches. The tree on the left is shown for branch length visualisation. Photographs of Amorimia ssp. by Fabián Michelangeli, Mascagnia cordifolia by Marco O.O. Pellegrini, and Mascagnia australis C.E.Anderson by Climbie F. Hall.

In light of the confirmation of the molecular evidence by Muñoz-Rodríguez et al. (2022), we moved on to questioning if there could have been an issue with the source of the sequence, either by laboratory contamination or problems with the source of the samples. The first hypothesis of laboratorial contamination was discarded a priori since Muñoz-Rodríguez et al. (2022) themselves stated that their sample of K. brasiliensis was re-sequenced by different laboratories, which resulted in identical sequences. This led us to investigate next if there could have been an issue with the source of the sample. Hence, the solution to this taxonomic conundrum would rely on the analysis of the herbarium sheets sampled by Muñoz-Rodríguez et al. (2022).

Nonetheless, these authors do not explicitly state which duplicate of Passos et al. 5263 was sequenced (from the six available specimens: ALCB, CEPEC, HRCB, HUEFS, K, or SPF), although it is mentioned by the authors that only the K herbarium was visited in person. However, the K isotype (K000979156) was not annotated to indicate that it had been sampled for DNA studies by the authors. In fact, this isotype only had a DNA sample slip dated from 2019 for unpublished molecular phylogenetic studies led by Kew’s in-house researcher Dr Tim Utteridge. Thus, we also looked at the possibility of the sequences of K. brasiliensis generated by Muñoz-Rodríguez et al. (2022) being from one of the Brazilian isotypes (i.e. ALCB037775, CEPEC00077827, HRCB38156, and HUEFS0028681) or the holotype (i.e. Passos et al. 5263, SPF). The curators of the abovementioned herbaria were contacted, and it was confirmed that leaf material for DNA sequencing had not been sent from these herbaria to Muñoz-Rodríguez or collaborators. Additionally, the database of the Brazilian Federal Government authority that regulates and authorises the use of any genetic material from Brazilian biological diversity in scientific studies (SisGen - Sistema Nacional de Gestão do Patrimônio Genético e do Conhecimento Tradicional Associado, https://sisgen.gov.br) was checked for records of authorisation having been granted to sequence the DNA of K. brasiliensis from any of the Brazilian herbaria in which all of the type specimens of K. brasiliensis were deposited. This search retrieved no results, further supporting that none of these materials had been sampled by foreign researchers or sequenced outside Brazil. Hence, we conclude that the sampled specimen would have, indeed, been the K sheet.

During the examination of the K isotype of K. brasiliensis (Passos et al. 5263), it was found that the plant was entirely glued to the sheet, and some detached leaves and fruits were stored in a paper capsule on the left lower side of the sheet (Fig. 4). These fragments were likely the source of the sample collected by Muñoz-Rodríguez and collaborators, considering the difficulties in collecting leaf material from the glued specimen. Yet, we observed that one of the leaves inside the fragment capsule did not match the general morphology of the remaining leaves of K. brasiliensis (Fig. 4). Inspection under a hand lens and stereomicroscope revealed that this “distinct” leaf presented V-shaped, 1-celled malpighiaceous hairs instead of the unbranched, multicellular hairs with bulbous bases, of the remaining leaves of K. brasiliensis (Fig. 4). On this discovery, Rosangela Simão-Bianchini and several Brazilian herbaria collaborators also checked the holotype and isotype sheets. Particular attention was given to the content of fragment capsules to check for possible foreign leaf material mixed up with true leaves of Keraunea. This investigation confirmed that none of the duplicates of Passos et al. 5263 and their fragment capsules in Brazilian herbaria contained leaf fragments with V-shaped, 1-celled malpighiaceous hairs. Thus, we could conclude that this sample mixture only pertained to the K isotype, further supporting our assumption that Muñoz-Rodríguez et al. (2022) did, indeed, sample the K specimen and not one of the Brazilian specimens.

Figure 4. 

Photograph of the isotype of Keraunea brasiliensis (Passos et al. 5263) deposited at RBG Kew’s herbarium showing the open fragment capsule storing leaves of K. brasiliensis (lower photographic detail showing unbranched hairs with bulbous base) and a single leaf of Mascagnia cordifolia (upper photographic detail showing 2-branched hairs). Photograph by Rafael Felipe de Almeida.

Thus, our critical re-analysis of the molecular and morphological data used by Muñoz-Rodríguez et al. (2022) allowed us to conclude that the phylogenetic placement of K. brasiliensis in Malpighiaceae can be explained by the contamination of the fragment capsule in the K isotype (Passos et al. 5263, K000979156) by Malpighiaceae leaves and their fragments. It is possible that this contamination was accidentally introduced during fieldwork or during the processing of the specimens for herbarium incorporation when leaf material, most likely of Mascagnia cordifolia, was added to the capsule by mistake.

It is not uncommon for molecular phylogenetic studies to demonstrate that genera, and every so often species, are not monophyletic. In fact, the main contribution of molecular phylogenetic studies to plant systematics has been to challenge traditional classification systems based on morphological dogmas. One of the most famous cases is the dicotyledon/monocotyledon traditional division of flowering plants, which was challenged and reorganised into an entirely different system of subdivisions of flowering plants once the first molecular phylogenetic studies demonstrated the non-monophyletic nature of the dicotyledons (Chase et al. 1993). Nonetheless, to the best of our knowledge, Keraunea represents an unprecedented case in which the type species of a genus was presumed non-monophyletic. Different samples of an isotype and a paratype were placed in distinct clades of Pentapetalae (i.e. superasterids and superrosids) without any morphological support.

When Muñoz-Rodríguez et al. (2022) first observed the non-monophyly of K. brasiliensis, these authors re-sequenced the tissue samples in a different laboratory to further investigate their unusual results. In fact, this is only the first step towards troubleshooting phylogenetic incongruencies not corroborated by morphological evidence. The second and most important step is going back to the original sampled specimen to check for potential taxonomic misidentifications and/or contamination by foreign plant material. Unfortunately, skipping this second step seems to be common practice in plant phylogenetic studies, with several Sanger DNA sequences published on GenBank for different groups of seed plants showing misidentification problems, as discussed by Smith and Brown (2018). These authors had to exclude several shorter gene regions (i.e. DNA barcode regions such as matK and rbcL) from their seed plant phylogenetic study due to misidentification or contamination of samples, which had led to the incorrect placements of several of the analysed groups. Thus, even if only a single sequence constitutes a misidentification or contamination in a multiple loci analysis, this is enough to drive the incorrect placement of the taxon. Consequently, when conducting large-sized phylogenetic analyses, small percentages of bad data can dramatically inhibit accurate phylogenetic estimates (Smith and Brown 2018) and lead to incorrect and often disruptive taxonomic placements, such as in the case of Keraunea.

The case of the Keraunea brasiliensis isotype being phylogenetically misplaced in Malpighiaceae could have been avoided if Muñoz-Rodríguez et al. (2022) had noticed some macro-morphological differences in the leaves (Table 1) or the presence of V-shaped, 1-celled malpighiaceous hairs (Table 1, Fig. 4). Alternatively, in the face of the odd molecular results, these authors should have critically re-analysed the sampled herbarium specimens while looking for the potential source of the mistake. It is almost impossible for most families of flowering plants to be reliably identified on vegetative characters alone, especially using only leaf fragments. Malpighiaceae represents one of the few exceptions and maybe one of the most well-known. One of the most significant synapomorphies for this family, the indumentum of their vegetative structures, is always made of 1-celled hairs with two branches that can be T-, Y-, or V-shaped, united by a well-developed or inconspicuous base (i.e. foot) (Almeida and Morais 2022).

This unique hair morphology was first described in Malpighiaceae and has since been referred to as malpighiaceous hairs in taxonomic literature (e.g. Almeida and Morais 2022). However, malpighiaceous hairs are not exclusive to Malpighiaceae, being also found in 23 unrelated families of flowering plants (i.e. Acanthaceae, Aizoaceae, Asteraceae, Boraginaceae, Brassicaceae, Burseraceae, Cannabaceae, Capparaceae, Combretaceae, Convolvulaceae, Cornaceae, Connaraceae, Ebenaceae, Escalloniaceae, Lythraceae, Myrtaceae, Sapindaceae, Sapotaceae, Thymelaeaceae, Verbenaceae, Vitaceae, Vochysiaceae, and Zygophyllaceae; Rao and Sarma 1992). Within Convolvulaceae, 2-branched hairs (i.e. malpighiaceous hairs) are found in Ipomoea L. (Wood et al. 2016), Evolvulus L. (Silva and Simão-Bianchini 2014), Cordisepalum Verdc. (Staples 2006), Dinetus Buch.-Ham. ex D.Don (Staples 2006), Duperreya Gaudich. (Staples 2006), Poranopsis Roberty (Staples 2006), Tridynamia Gagnep. (Staples 2006), Stylisma Raf. (Myint 1966), Jacquemontia Choisy (Patel 2021), Erycibe Roxb. (Kochaiphat et al. 2021), and Neuropeltis Wall. (Breteler 2010).

Keraunea was proposed by Cheek and Simão-Bianchini (2013) to be close to Neuropeltis and Neuropeltopsis, two unusual Palaeotropical genera of Convolvulaceae with very enlarged bracts supporting the fruit and partly fused to the pedicel. Therefore, the enlarged membranous bracts of Keraunea greatly resemble the membranous to subcoriaceous bracts present in Neuropeltis and Neuropeltopsis, being one of the first morphological characters that suggested its original placement in Convolvulaceae. Nonetheless, molecular phylogenetic evidence did not corroborate the close relationship between Keraunea and Neuropeltis (Muñoz-Rodríguez et al. 2022; Simões et al. 2022). Although sampling of Neuropeltis was limited to only one African species, N. acuminata (P.Beauv.) Benth., and the genus Neuropeltopsis had not been sampled in either study.

Since Neuropeltis shows malpighiaceous hairs in several species (Breteler 2010), one could argue that this type of hair is common in several genera of Convolvulaceae, being the main reason justifying the results of Muñoz-Rodríguez et al. (2022). Nonetheless, some differences distinguish the hairs of Convolvulaceae from those of Malpighiaceae. In Convolvulaceae, the 2-branched hairs are always 2-celled and T-shaped, while in Malpighiaceae, the 2-branched hairs are always 1-celled and T-, Y-, or V-shaped, such as in the mixed material from Passos et al. 5263 (K000979156; Table 1; Fig. 4). Therefore, based solely on morphology, it can be confidently established that the mixed material from the isotype at Kew is of a Malpighiaceae species, most probably a specimen of Mascagnia cordifolia, which always shows V-shaped hairs such as those presented in Fig. 4. Additionally, the hairs present in all leaves of the known specimens of Keraunea brasiliensis are always unbranched, multicelled, and bulbous at the base, as described by Cheek and Simão-Bianchini (2013) and confirmed by the present authors in the re-examination of the specimens (Table 1; Fig. 4). Thus, we here refute the hypothesis of Keraunea brasiliensis – or Keraunea pro parte – belonging to Malpighiaceae.

Meanwhile, the family placement of Keraunea has been the subject of two subsequent morphological and phylogenetic studies. A few days after the online publication of the preprint of this study (Almeida et al. 2023), another preprint was published solving the phylogenetic placement of Keraunea in Ehretiaceae (Boraginales; Cheek et al. 2023). These authors sampled four DNA markers (ITS, matK, rbcL, and trnL-F) for K. brasiliensis, K. capixaba, and a third newly proposed species (= K. confusa Moonlight & D.B.O.S.Cardoso), for all currently accepted genera of Ehretiaceae, and for a few genera of Cordiaceae and Heliotropiaceae as outgroups (Cheek et al. 2023). A few days after the publication of Cheek et al. (2023)’s preprint, Moonlight and Cardoso (2023) effectively published an independent study also solving the phylogenetic placement of Keraunea within Ehretiaceae (Boraginales) alongside a taxonomic revision for this genus including three new species (i.e. Keraunea bullata Moonlight & D.B.O.S.Cardoso, K. confusa Moonlight & D.B.O.S.Cardoso, and K. velutina Moonlight & D.B.O.S.Cardoso). Both studies excluded the contaminated sequences of Passos et al. 5363 from their molecular sampling and recovered similar trees, corroborating the correct phylogenetic placement of Keraunea in Ehretiaceae by independent analyses of comparable molecular datasets. The main difference between these studies is that Cheek et al. (2023) sampled three species of Keraunea (including a non-contaminated sequence of K. brasiliensis from Passos et al. 5363), while Moonlight and Cardoso (2023) sampled a single species of Keraunea (i.e. K. confusa, described from Lombardi 1819). Hence, at present, Keraunea comprises five accepted species endemic to the Atlantic Rainforest and Caatinga biomes in the States of Bahia, Espírito Santo, Minas Gerais, and Rio de Janeiro, Brazil (Cheek et al. 2023; Moonlight and Cardoso 2023). Since several nomenclatural types of Keraunea (i.e. isotypes) were stored as undetermined specimens within Nyctaginaceae or Solanaceae in several herbaria, future taxonomic novelties may arise after a careful re-analysis of the herbaria from eastern Brazil or additional field collections are conducted.

The misidentification of herbarium sheets sampled in phylogenetic studies highlights the need to adequately address how molecular phylogenetic studies should proceed with contaminated and/or misidentified sequences for large-sized phylogenetic analyses. Hence, taxonomic expertise is fundamental to ensure the correct taxonomic identification of DNA samples in phylogenetic studies. On a superficial analysis of the first 60 open-access papers on plant phylogenomics retrieved for 2022 from the Google Scholar database (https://scholar.google.com, accessed 10 Jan. 2023), we gathered that only 22% of these studies included a taxonomic expert on the group (i.e. someone who has already published floras or taxonomic revisions in the analysed groups) in the authorship of the paper. Additionally, 78% of these studies do not specify the taxonomic criteria used in the study or mention the taxonomic specialist who would have confirmed the identification of the sampled specimens (see supplementary files). This scenario is worrisome since it might be a reflection of the view of biological collections by some plant molecular systematists as immutable DNA archives. This misconception of the dynamic nature of plant systematics and the need to constantly revise the determinations of the consulted specimens in the light of new evidence can be easily tackled by association and collaboration with a taxonomic expert(s) in the study group.

While large phylogenomic projects, such as RBG Kew Tree of Life (Baker et al. 2022), have developed thorough quality control pipelines to test the correct phylogenomic placement of the sequenced samples at the family rank, determinations below the family rank need to be treated with the same care and scrutiny. Furthermore, it is essential to maintain good taxonomic practice for collecting DNA samples, as relying on bioinformatic pipelines alone it is not always possible to effectively point out human errors introduced in the analyses. Good knowledge of plant taxonomy and morphology is a fundamental basis for minimising mistakes in the early stages of the molecular phylogenetic processes. To mitigate the dangers of cross-contamination in herbarium specimens or misidentification/contamination that underlie molecular systematic studies, we here suggest two sets of golden rules for herbarium sampling: