The material cited in the Myrcianthes monograph of Grifo (1992) was the core of the database (42% of the records). Google Earth was used to georeference all collections that had detailed geographic data. This was complemented with specimens with original geographic coordinates downloaded from SpeciesLink (https://specieslink.net/). Herbarium records that recorded country centroid coordinates were eliminated or, if enough geographic information was provided, refined to exact localities. The database was also complemented with data from Tropicos (http://www.tropicos.org) for Central America. All records were exported into BRAHMS v.7.9.14 (https://herbaria.plants.ox.ac.uk/bol/brahms), which was used for data curation. Duplicate records, those unidentified to species level without images, or those whose identification were considered unreliable, were excluded. The BRAHMS tool “look for similar/identical records” (using different combinations of collector names, collection numbers, and day, month and year of collection) was used to eliminate or combine duplicates. Reliable identifications were those made by specialist taxonomists or by generalist taxonomists with excellent knowledge of local floras. Altitudes were obtained from herbarium records; altitudes given in feet were converted to meters. Where an altitudinal range was given on the specimen, the lower limit was used to calculate mean altitude per species, therefore mean altitudes probably underestimate the true values (Suppl. material 1, Table S1). The final dataset is available from the corresponding author. Preliminary maps were made in BRAHMS v.7.9.14. in Google Earth and visually checked against geographic distributions reported by Grifo (1992) to detect possible incongruities. Final maps for publication were generated in QGIS v.3.40.3-Bratislava (QGIS 2024) and R v.4.1.1, using packages sf v.1.0–20 (Pebesma 2018), raster v.3.6–32 (Hijmans 2025), sp v.1.4–5 (Bivand et al. 2013), and phytools v.2.4–4 (Revell 2012).

To compare latitudinal and altitudinal distributions of Myrcianthes with other New World genera of Myrtaceae, we first checked distributions on Plants of the World Online (POWO 2024, https://powo.science.kew.org/) (that uses TDWG regions) for all New World Myrtaceae genera. This revealed that only Eugenia, Myrcia DC., and Psidium L. had similar wide distributions. These genera are all absent from Chile where Myrcianthes occurs but roughly match it in extending from Mexico and Florida in the north to Uruguay and Argentina in the south. The next step was to establish extreme latitudinal and altitudinal records in country and regional floras (i.e. for Argentina: Legrand 1962; Rotman 1995, 2005; for Uruguay: Legrand 1943, 1968; for the USA, Florida: Landrum 2022). Online records, particularly SpeciesLink, the NY and MO herbaria, and the map searches and altitudinal searches in CoTRAM – Cooperative Taxonomic Resource for American Myrtaceae (https://cotram.org/) were also consulted. For Psidium, the first author’s personal database of ca 6,000 records was consulted (Proença et al. 2022). We applied the term narrow-endemic to species with a known distributional range < 1° latitude × longitude, a common minimum geographic unit adopted in biogeographic studies (Vieira et al. 2008; Diniz-Filho et al. 2009).

We sampled 123 species, of which 11 were Myrcianthes species (ca one third of the species in the genus; Suppl. material 2, Table S2). To encompass geographic variability, we included two to three vouchers of the following species: M. cisplatensis (Cambess.) O.Berg, M. ferreyrae (McVaugh) McVaugh, M. fragrans (Sw.) McVaugh, and M. pungens (O.Berg) D.Legrand (Suppl. material 2, Table S2). To encompass phylogenetic variability within Myrtaceae, representatives from all sections of Eugenia, sister genus of Myrcianthes and the only other genus of subtribe Eugeniinae (Mazine et al. 2018), were included (19 species), as well as representatives of nine other Myrteae subtribes (44 species in 34 genera). To encompass phylogenetic variability outside of tribe Myrteae and attain better sampling for dating the lineages, we included representatives of other Myrtaceae tribes (26 species in 20 genera) and of seven other Myrtalean families (23 species in 15 genera).

A total of 403 sequences of the internal and external transcribed spacer (ITS and ETS, respectively) of the ribosomal nuclear region and two plastid markers (matK and psbA-trnH) were already available. Ten new sequences were generated for this study; the remainder were downloaded from GenBank (Suppl. material 2, Table S2). All molecular data were combined (3650 bp) into one matrix with two partitions, the first partition representing the nuclear sequences (1768 bp), while the second representing the plastid sequences (1882 bp). Mantel correlations (in which a value of 1 would indicate identical trees with identical bootstrap values; Legendre and Legendre 1998) were recently calculated for these four markers compared to a nine-marker phylogenetic tree (all trees 0% missing data) for 53 species of tribe Myrteae (NMWG 2024). This 53-species tree was totally resolved, and all bootstrap values (henceforward BS; Soltis and Soltis 2003) were > 75. The correlation values for these four markers were 0.97, 0.97, 0.96, and 0.91, respectively, suggesting that they are highly informative (at least under the condition of 0% missing data).

To construct the phylogenetic tree, we applied Maximum Likelihood (ML) analysis to a matrix with 11 Myrcianthes species (ca 1/3 of the genus); the missing data was 25% for Myrcianthes. ML analysis was performed with RAxML v.8.2.12 (Stamatakis 2014) using the rapid bootstrap algorithm with 1000 replicates, combined with a search of the best-scoring ML tree under default parameters via the CIPRES platform (Miller et al. 2010). Concatenation of sequences from different accessions was done when the morphology and geographic locality among vouchers showed a reasonable match; geographically distant samples were not concatenated.

The matrix was used to explore temporal divergence among species. To calibrate the tree, we used the Bayesian inference approach implemented in BEAST2 (Drummond et al. 2012) via the CIPRES platform (Miller et al. 2010) using two partitions and the best nucleotide substitution model estimated separately for each partition. The GTR + G + I model was found to be best for both partitions, according to the Akaike information criterion (AICc) calculated in jModelTest v.2.1.4 (Darriba et al. 2012). An uncorrelated relaxed molecular clock, assuming a lognormal distribution of rates and a Yule model, was applied.

Four runs of 200,000,000 generations were performed, sampling one tree every 1000^th generation. Parameter convergence was confirmed by examining their posterior distributions in TRACER v.1.7.2 (Rambaut et al. 2018). The MCMC sampling was considered sufficient when effective sampling size (ESS) of each parameter was > 184. A maximum clade credibility tree with median branch lengths and a 95% highest posterior density interval on nodes was built using TreeAnnotator v.2.6.0 (Drummond et al. 2012), based on the remaining set of trees after burn-in (for each run a burn-in of 25% was applied).

Species names and circumscriptions for Myrcianthes follows Grifo (1992) and the continuously updated World Checklist of Vascular Plants (Govaerts et al. 2021a). Three names that are listed by Govaerts et al. (2021b) as accepted have not been included in this study: Myrcianthes bradeana Mattos is a synonym of Eugenia capparidifolia DC. (Mazine and Faria 2022); M. esnardiana (Urb. & Ekman) Alain was excluded from Myrcianthes by Grifo (1992); and M. storkii (Standl.) McVaugh was treated as a synonym of M. rhopaloides by Grifo (1992).

Morphological data was gathered from the taxonomic descriptions in Grifo (1992) for all species that were described before 1992. For more recently described species, morphological data was gathered from original descriptions (Sobral et al. 2012; Parra-O. and Bohórquez-Osorio 2016; Kawasaki et al. 2019; Proença et al. 2023), i.e. respectively M. riparia Sobral, Grippa & T.B.Guim., M. roncesvallensis Parra-Os. & Bohórq.-Os., M. rubra B.Holst & M.L.Kawas, and M. cruciata M.Ibrahim & Proença. Characters discussed in the text are those considered of taxonomic importance in Neotropical Myrtaceae (Landrum and Kawasaki 1997; Lucas et al. 2019).

Wood anatomy data was gathered from the literature. The species Myrcianthes cruciata M.Ibrahim & Proença was published together with the characterization of its wood anatomy, and an overview of the wood anatomy of the genus Myrcianthes (Rebollar et al. 1993; Ramírez-Martínez et al. 2017; Proença et al. 2023). For Eugenia, wood anatomy data was taken from the personal database of wood anatomical characters of Myrtaceae tribe Myrteae compiled by one of us (J.S.-O.) based on the following literature: Santos and Marchiori (2011), Santos et al. (2014, 2015), Schmid and Baas (1984), Sonsin et al. (2014), and data available on the InsideWood (2004–) website, that aims to be a comprehensive database of anatomical characters of extant and fossil wood following the IAWA traits and classification (IAWA Committee 1989). When images of slides were available on InsideWood (2004–), these were also examined. When choosing fossil woods for calibration, we focused on characters that are considered of taxonomic importance (type of porosity, vessel grouping, type of perforation plate, size and position of intervessel pits, vessel diameter, vessel density, borders of vessel-ray pits, type of axial parenchyma and ray width; Baas et al. 2000). For the comparison between Myrcianthes and Eugenia only characters that varied between these genera, or between species of these two genera, are shown; characters that did not vary (the majority) are not shown.

The geographic and altitudinal database is stored in BRAHMS and includes 1,228 georeferenced records of the 36 currently accepted species of the genus and is available from the corresponding author. A list of extreme latitudinal and altitudinal records of Myrcianthes is available for each species in Suppl. material 1, Table S1 and for other genera in Suppl. material 3, Table S7. Myrcianthes is distributed from 29°02’N (Florida, USA) to 34°47’S (Maldonado, Uruguay; total latitudinal range 63°48’) and from near sea level to ca 3729 m a.s.l. (Oruro, Bolivia) (Fig. 2). Its closest rivals are its sister genera, the megadiverse Eugenia (> 1000 species), Myrcia (ca 800 species), and Psidium (ca 100 species). Eugenia occurs from ca 28°14’N (Florida, USA) to 32°57’S (Vergara, Uruguay) or possibly 34°36’S (Reserva Ecológica Costanera Sur, Buenos Aires, Argentina); total latitudinal range 61°12’ (or possibly 62°50’) and from sea level to ca 2900 m a.s.l. (Imbabura, Ecuador) (Suppl. material 3, Table S7). Myrcia occurs from 25°43’N (Florida, USA) to 31°16’S in Salto, Uruguay; total latitudinal range ca 57°) and from sea level to possibly 3100 m a.s.l. (2950–3100; Pasco, Peru). Psidium occurs from ca 27°02’N (Chihuahua, Mexico) to 37°50’S (Buenos Aires, Argentina; total latitudinal range 64°52’) and from sea level to 2840 m a.s.l (Ecuador, Pichincha). Thus, of the three comparable genera, Psidium surpasses Myrcianthes in latitudinal amplitude by ca 1°04’ (less than 2% of the total range of Myrcianthes) but has an altitudinal range that is 75% of that of Myrcianthes. Eugenia and Myrcia have both narrower latitudinal and altitudinal distributions than Myrcianthes. Therefore, Myrcianthes is confirmed to have a broader latitudinal/altitudinal distribution than any other Neotropical genus of Myrtaceae (Suppl. material 3, Table S5).

Plotting the latitudinal and altitudinal averages (with interval bars for minimum and maximum values of altitude) allows a rough overview of the geographic/altitudinal ecospace (G/A ecospace) of each species in the Americas (Fig. 2). This enabled us to conclude that all species in the more easterly, lowland clade have their midpoint between 100 and 1200 m and their maximum altitude below 1400 m, with the exception of M. fragrans, which characterizes this clade as essentially lowland. All species in the more westerly (mostly Andean) clade have midpoints between 700 and 3200 m, which characterizes this clade as essentially highland (Fig. 2).

Narrow-endemic species occur in two different areas of the G/A ecospace. The first suite of narrow endemics (four species) appears in the right-hand corner, at the more southerly latitudes and lower altitudes. The second suite of narrow-endemic species are scattered over the lower latitudes (7°N–13°S) and have their altitudinal midpoints above 2300 m, i.e. they are cloud forest species, with the possible exception of M. monteucalyptoides Proença & L.V.S.Jenn. (recorded altitude 1460 m) although the original authors questioned this altitude since the type locality, Tarma, Peru, is located at ca 3000 m and the nearest locality in the vicinity of Tarma with the altitude cited on the type specimen is 50 km distant (Proença et al. 2011).

Two-thirds of the 36 Myrcianthes species remain unsampled in the phylogeny and they are mainly from the Andes suggesting where future efforts should be focused, as well as the Guyanas. Myrcianthes emerged as a monophyletic genus with high bootstrap support in the ML tree (96 BS; Fig. 3; Suppl. material 4, Fig. S1) and maximum support in the Bayesian tree (1 PP; Suppl. material 4, Fig. S2). It was recovered as sister to Eugenia with maximum bootstrap support (100 BS) and maximum Bayesian support (1 PP). Myrcianthes coquimbensis (Barnéoud) Landrum & Grifo emerged as sister to the rest of the genus, which then splits into two clades (Figs 3, 4; Suppl. material 4, Fig. S3; Suppl. material 5, Fig. S4 for the non-collapsed tree). We have called these two clades the highland clade and lowland clade, respectively, based on the dominant altitudinal distribution in each clade (Fig. 2).

Divergence time estimates (Fig. 4) indicated that Myrcianthes diverged from Eugenia ca 32 mya. Myrcianthes coquimbensis is sister to the rest of the genus and the divergence is deep, dated at ca 24 mya. The two Myrcianthes clades have more or less similar ages. The highland clade is slightly older (ca 13.7 my old) than the lowland clade (ca 11.4 my old). However, the sampled species of the highland clade diversified later than those of the lowland clade species.

The model that conferred the best likelihood on our Myrcianthes dataset was BAYAREALIKE+J with an AICc of 67.36 (80% of probability), while the next closest model (DEC) had an AICc of 72.07, and all the rest were > 73.29 (Table 1). The BAYAREALIKE+J model is similar to the Bay Area model implemented in Landis et al. (2013) but BioGeoBEARS allows the inclusion of a founder event and jump speciation captured by the j parameter (Matzke 2013) that permits a daughter lineage to have a different area from the direct ancestor. The addition of j parameter improves the log likelihood of resulting inferences of ancestral areas in comparison to a model with only two free parameters (Table 1), showing jump speciation (i.e. dispersal between non-adjacent areas) as an important pattern in range variation of Myrcianthes. The Chacoan and Paraná dominion are highly likely to be the ancestral range for most Myrcianthes species (Fig. 5) including the ancestral node of each of the clades (highland and lowland). In the lowland clade, subsequent nodes show shifts into the South Brazilian dominion and into the Northwest of Latin America. These shifts are estimated to date from ~12–9 mya. In the highland clade, the first diversification occurred ~10 mya in the South Brazilian dominion, specifically in the Yungas province where we found two species (Myrcianthes callicoma McVaugh and M. osteomeloides (Rusby) McVaugh, Fig. 5) endemic to this province. The second diversification of the highland clade occurred ~6 mya; this node was recovered as being from Chacoan-Paraná-South American transition zone, with two of its four species emerging as endemic to the South American transition zone (Myrcianthes ferreyrae and M. roncesvallensis).

We examined the floral characters from the literature (Grifo 1992; Proença et al. 2011, 2023) but could not find any obvious lowland/highland split for type of inflorescence, number of sepals (4, 5, or 4–5), number of stamens, disk size (1.5–8 mm in diameter but most species 2–4 mm), or number of ovules per locule (5–30). Although no clear-cut lowland/highland dichotomy in floral morphology was found, we did detect a tendency for species with more austral distributions to have fewer ovules per locule (see Discussion).

Five species of Myrcianthes have been sampled from previous studies on wood anatomy (Schmid and Baas 1984; Rebollar et al. 1993; Richter and Dallwitz 2000; Santos et al. 2015; Ramírez-Martínez et al. 2017; Proença et al. 2023). Although most wood characters were uniform within the genus, four characters differed among Myrcianthes species and two differed between Myrcianthes and its sister genus Eugenia (Table 2). Two wood anatomy characters that are apparently conserved in Eugenia varied between species and clades of Myrcianthes: 1) helical thickenings (absent in Eugenia) are often found in Myrcianthes; 2) prismatic crystals in the axial parenchyma (universal in Eugenia) are either present or absent in Myrcianthes.

Myrcianthes coquimbensis (Fig. 1E), sister to all other Myrcianthes, is in the extreme right-hand corner of the G/A ecospace (high latitude, low altitude) and is a narrow-endemic species restricted to maritime scrub in Coquimbo, Chile. The G/A ecospace close to M. coquimbensis is occupied by six species (Fig. 2). Three of these are also narrow-endemic species and occur between 16° and 34°S. Two of these are of unknown phylogenetic affinities: M. riparia (restricted to one river basin in southern Brazil; Sobral et al. 2012) and M. pedersenii D.Legrand (restricted to fields in Misiones, Paraguay; Grifo 1992). The third, M. ferreyrae (restricted to “lomas”, which are “fog oases” of vegetation in the desert belt in Arequipa, Peru; Song et al. 2023) is a member of the older, highland clade (Figs 2, 6A). The other three species are relatively widespread (M. cisplatensis, Figs 2, 6B; M. pungens, Figs 2, 6A; M. gigantea D.Legrand, Figs 2, 6B) and were sampled for the phylogenetic reconstruction, belonging to both the lowland and highland clades (Fig. 7). Moreover, two of them are the only species of the genus to straddle the Chaco sensu Cabrera and Willink (1973) – not the Chacoan Dominion of Morrone et al. (2022) that covers the whole South American dry diagonal of Caatinga, Cerrado, and Chaco. The area now occupied by the Chaco has been a barrier to wet forest species expansion since the Miocene; first by marine transgressions and in the late Miocene by the uplift of the western hills that created a barrier to NE winds that led to extreme desertification (Pascual and Ortiz-Jaureguizar 1990). Although the possibility of later long-distance dispersal cannot be discarded, it seems more likely that these two species predated these events: M. cisplatensis (lowland clade), is the most southerly Myrcianthes species and sister to the rest of the lowland clade, and M. pungens (highland clade) is also the most southerly species of its clade. Myrcianthes gigantea (lowland clade) is the last of the six species within the G/A ecospace (near M. coquimbensis) and is restricted to southern Brazil and Uruguay but is distributed somewhat further north than M. cisplatensis (Fig. 6B).

The lowland clade has a broad latitudinal range (Fig. 6B), but a relatively narrow altitudinal range, while the highland clade has a somewhat narrower latitudinal range, but an altitudinal range that is approximately three times broader than the former (Fig. 2). This is congruent with the rising of the Andes providing opportunities for this clade to colonize higher altitudes, which were unavailable in eastern South America. A latitudinal gradient is also perceptible, i.e. species generally occur at low altitudes in the southern part of Myrcianthes’ range but reach higher altitudes as they approach the equator. Myrcianthes fragrans was found to be the species with the widest altitudinal range (Fig. 2). It occurs in northern South America, the Greater and Lesser Antilles, Mexico, and Florida. Myrcianthes fragrans overlaps both altitudinally (ca 1500–2000 m) and geographically (5°–8°N and 70°–72°W) with several putative (based on geographic distribution) highland clade species in Colombia and Venezuela (e.g. M. hallii (O.Berg) McVaugh, M. karsteniana (O.Berg) McVaugh, and M. leucoxyla (Ortega) McVaugh). Myrcianthes fragrans is exclusive to the Northern Andean zone and does not extend beyond the Amotape-Huancabamba North-South Biogegraphical barrier (Luebert and Weigend 2014 and references therein).

Our results for Myrcianthes show that the genus is monophyletic and confirm prior studies that it is sister to Eugenia (Lucas et al. 2007; Vasconcelos et al. 2017; Giaretta et al. 2022), with the mean date for the split estimated at ca 32.1 mya (middle Eocene). The date of the split is congruent with the mean age of 33.2 my estimated for the Eugeniinae lineage based on a very large sample of tribe Myrteae (NMWG 2024). The Eocene was marked by climatic stability (Morrone et al. 2022) and extensive dispersal of other plant clades since the beginning of this period has been postulated (e.g. the clusioid clade; Ruhfel et al. 2016).

We also confirm the finding of Retamales (2017) that the Chilean endemic M. coquimbensis is sister to the rest of the genus. Myrcianthes coquimbensis split from the main Myrcianthes lineage in the late Oligocene (ca 24.2 mya), but the main Myrcianthes lineage did not diversify before the middle of the Miocene (later than Eugenia). The main Myrcianthes lineage subsequently split into two major clades that we have called lowland and highland clades for simplicity, with the former more easterly and the latter mostly Andean (Figs 5–7). The clades overlap at the base of the tree in Uruguay and southern Brazil (Fig. 5), and probably meet again in northern South America (in the Cordillera Oriental, Colombia and in Parque Nacional El Ávila, Cundinamarca, Venezuela), although this scenario depends on the unproven assumption that the sympatric species, collected within a few kilometres of M. fragrans (M. hallii, M. karsteniana, and M. leucoxyla) belong to the highland clade. A similar pattern, of two separate clades, had been found in another ancient myrtaceous genus with a compatible distribution, Myrceugenia, and, as we had hypothesized, both genera apparently had their origin in Chile. The Western clade of Myrceugenia diversified mainly in Chile and western Argentina (14 species) but, in contrast to Myrcianthes, did not colonize the Andes to any significant degree (Murillo-A. et al. 2016). The Eastern clade of Myrceugenia, on the other hand, diversified much more in the east (38 species in Uruguay, eastern Argentina, and southeastern Brazil; Vieira et al. 2025) than did Myrcianthes. A single species, Myrceugenia linda F.C.S.Vieira & Proença, occurs in the Yungas forests of Bolivia (Proença et al. 2024). Although Myrceugenia is much more diverse in the east than in the west, the genus only extends to ca 13°S in the Bahia highlands, Brazil (SpeciesLink 2022). This is the opposite of what happened in Myrcianthes, in which, although there was little speciation in the east, there was a significant range expansion (due to M. fragrans). It is possible of course that speciation did occur, but the species have not survived.

Vasconcelos et al. (2017) suggested a quick northerly vertical expansion of tribe Myrteae into South America soon after its initial diversification; the only exception was Myrcianthes (based on M. fragrans) that required a long-distance dispersal event (LDDE) into the Caribbean to explain its current distribution. Here, we were able to show, with a larger sample of the genus (including north-eastern Brazilian M. cruciata that was not described until 2023), that the origin of Myrcianthes is also southern (Chacoan-Paraná dominion, Fig. 5) with a northerly expansion as in other Myrteae, corroborating the long-held theory that Myrtaceae diversified from south to north (Berry 1915; Vasconcelos et al. 2017). Barreda et al. (2021) found that Myrtaceae was an important component of the Patagonian palaeoflora from the Cenozoic to the early Miocene, decreasing in diversity during the middle Miocene (~16–12 mya) with global cooling, and becoming extinct in Patagonia by the late Miocene (~12–6 mya). Landrum (1981) postulated that during the late Oligocene (~25–23 mya), Myrceugenia grew across South America in a temperate/subtropical austral forest, and our study corroborates that theory and suggests a similar pattern for Myrcianthes. An interesting parallel was also found between the Myrcianthes lowland clade and Drymis Juss., a genus also belonging to an ancient southern Gondwanan family (Winteraceae). Similar to Myrcianthes, Drymis showed two vicariant clades, one endemic to Chile and the other in the Brazilian Atlantic Forest and the Northern Andes (Marquínez et al. 2009), i.e. both Drymis and the Myrcianthes lowland clade appear to have reached the northern Andes from Atlantic Forest ancestors.

The highland clade is represented in our study by six species (Figs 5, 6A) that occur in Argentina, Uruguay, southern and central Brazil, Bolivia, Paraguay, and the Andes at least to Colombia (based on the inclusion of M. roncesvallensis in the highland clade). The recently published NMWG (2024) tree with a “where we stand with the Myrteae phylogeny” approach looked at most of the same species of Myrcianthes (it did not include M. coquimbensis) but had significantly more missing data than our study, as nine molecular markers were used. Their Myrcianthes reconstructed phylogenetic tree was highly congruent to ours for the highland clade. They found a strongly supported (94 BS) highland clade composed of the same two subclades.

The highland clade crown is slightly older (ca 13.7 my old) than the lowland clade crown (ca 11.4 my old). Miocene diversification of the Myrcianthes highland clade in the Andes is congruent with increased uplift rates in the Eastern Cordillera, Cordillera Real, and Cordillera de Merida from the late Miocene onwards (van der Hammen 1988; Villamil 1999; Mora et al. 2008). The highland clade shows two well-supported subclades, one Central Andean and the other with a wider range. The Central Andean subclade, with two species, M. callicoma and M. osteomeloides from Argentina, Bolivia, and Peru, ranges from ca 26° to 13° S (crown ca 9.8 my old). A period starting at ca 13 mya was marked by rapid uplift of the Central Andes and several clades, e.g. in Peperomia Ruiz & Pav. (Piperaceae) diversified in the Central Andes during this period (Luebert and Weigend 2014). The wider-ranging clade, composed of M. ferreyrae, M. pseudomato (D.Legrand) McVaugh, M. pungens, and M. roncesvallensis from Argentina, Bolivia, Brazil, Paraguay, Peru, and Colombia ranges from ca 30°S to 4°N and diverged later (crown ca 6.6 my old). This would be congruent with the late Miocene climatic instability of the Central Andes, particularly the period between ca 7.9–6 mya (Uba et al. 2007). During this period, the uplift of the Eastern Cordillera in Colombia reached elevations high enough to produce an orographic barrier that intercepted moisture-laden winds and increased rainfall on the Eastern flanks (Mora et al. 2008), which might have favoured this essentially subtropical/montane genus. This division into two subclades could be an artefact of the low level of sampling in the highland clade or indicate that the highland clade of Myrcianthes underwent more than one radiation; further sampling is needed.

The lowland clade of Myrcianthes is represented by four species in our study (Figs 6B, 7) that occur in the Atlantic Forest, from Rio Grande do Sul to Ceará, but reaching northern South America, the Caribbean, southern Mexico, and Florida. Three species of the lowland clade (M. cisplatensis, M. cruciata, and M. gigantea) have relatively narrow distributions, but the fourth, M. fragrans, as mentioned above, is widely distributed (Fig. 6; Suppl. material 5, Fig. S6). Myrcianthes fragrans is the most northerly species of the lowland clade and of the genus. Conversely, M. cisplatensis is the most southerly species of the lowland clade and of the genus and is basal to the other three species of the lowland clade. The lowland clade was not recovered by NMWG (2024). They found that a clade formed by lowland species M. gigantea plus two accessions of M. fragrans was sister to the rest of the genus (35 BS), and another clade formed by lowland species M. cruciata plus M. cisplatensis (58 BS) was sister to the highland clade. We attribute this difference to the higher proportion of missing data in the NMWG tree as their bootstrap values were much lower than ours. In fact, NMWG (2024: 8) stresses that: “…given the low support of some relationships at both higher and lower taxonomic levels, it is important to emphasize that our supermatrix tree should not be considered as the final word in terms of phylogenetic relationships within Neotropical Myrtaceae. It is rather a first step towards evaluating sampling gaps…”.

There is a very broad study across the Myrtaceae of helical thickenings in the vessels (Schmid and Baas 1984). Their results have been re-interpreted in the light of the most recent Myrteae phylogeny (NMWG 2024) and currently recognized tribes and subtribes (Lucas et al. 2019; Wilson et al. 2022). Helical thickenings occur in tribe Xanthomyrteae (in Xanthomyrtus Diels) and in tribe Myrteae (in eight genera). In tribe Myrteae, helical thickenings occur in the two most basal subtribes, the Decasperminae (Gossia N.Snow & Guymer; Schmid and Baas 1984) and the Myrtinae (Myrtus L. and Mosiera Small; Schmid and Baas 1984; Fahn et al. 1986), and in three other subtribes, the Luminae (Luma A.Gray and Myrceugenia; Rancusi et al. 1987), the Eugeniinae (Myrcianthes; Schmid and Baas 1984; Ramírez-Martínez et al. 2017), and the Myrciinae (Myrcia sect. Myrcia; Santos and Marchiori 2011). The function and phylogenetic significance of helical thickenings (synonym: spiral thickenings) in the vessels has been the object of much debate (Baas 1973; van den Oever et al. 1981; Carlquist 1982). It was proposed by these authors that they are an adaptation to episodes of drought, ground freezing (physiological drought), or other sorts of water stress that can cause air bubbles (embolisms) to form in the vessels and block water movement. Helical thickenings in the vessels of the wood are considered a specialized anatomical feature, and do not appear in the fossil record until the Eocene, first reaching modern levels of incidence in Oligocene fossil floras (Wheeler and Baas 1991). Kohonen and Helland (2009) have convincingly demonstrated using physical models that helical thickenings in the vessels do increase vessel wettability (as predicted by Carlquist 1982) and lead to the formation of regular-shaped air bubbles that shrink much faster than the irregular bubbles that form in vessels that lack helical thickenings, thus restoring water flow more rapidly.

Most species of Myrcianthes sampled in the study by Schmid and Baas (1984) showed helical thickenings in the vessels. Myrcianthes cisplatensis (lowland clade, with helical thickenings) is the most southerly species of Myrcianthes (reaching 34.9°S; average altitude 282.5 m) and is a small tree (4–6 m; Grifo 1992). Myrcianthes gigantea (lowland clade, with helical thickenings) reaches similar latitudes (32.9°S; average altitude 878 m) and is a much taller tree (5–15 m; Grifo 1992). Myrcianthes cruciata (lowland clade, ranging from 3.9°S to 13.1°S; average altitude 305 m), is a tall northern Atlantic Forest tree (13–18 m; Proença et al. 2023) and lacks helical thickenings. Myrcianthes fragrans (lowland clade, most wide-ranging species of the genus) is (possibly) variable for this character: trees studied in Mexico showing helical thickenings (Ramírez-Martínez et al. 2017) but in other trees studied in Florida and Mexico no helical thickenings were reported (InsideWood 2004–; Rebollar et al. 1993). A review of Myrcianthes wood anatomy (Proença et al. 2023) suggested that the voucher specimen for the study of Ramírez-Martínez et al. (2017), that reported helical thickenings in M. fragrans, might have been misidentified, as its wood anatomy differs markedly from other species of the genus in vessel size, vessel density, and the reduced to apparently simple borders of the ray-vessel pits, besides the presence of helical thickenings. The third most southerly species of Myrcianthes (reaching 30.1°S; average altitude 77 m) is M. coquimbensis (sister species to the rest of the genus, lacking helical thickenings). Myrcianthes coquimbensis is a densely branched shrub (Fig. 1E) up to 1.5 m tall from coastal scrub vegetation (Landrum 1988; Landrum and Grifo 1988). Myrcianthes mato (Griseb.) McVaugh (clade unknown, with helical thickenings) reaches 27.5°S (average altitude 1904 m) and is a small tree (3–8 m; Grifo 1992). Under these climatic conditions and habits, it is possible that the retention of ancestral helical thickenings occurred in M. cisplatensis, M. mato, and M. gigantea, all trees from cool subtropical or montane forests, has been advantageous, but not in M. coquimbensis, a shrub growing in maritime rocky scrub at sea level nor in M. cruciata, a tree but adapted to lowland tropical rainforest. The difference in this character across Myrcianthes is perhaps not surprising, if the wide altitudinal and latitudinal range of the genus is considered, with many of its species restricted to specific altitudinal ranges (Grifo 1992; Worthy et al. 2019; this study). This scenario is congruent with the literature. Global studies of helical thickenings reported higher percentages in temperate, high latitude floras than in tropical floras (van der Graaff and Baas 1974) and higher percentages of helical thickenings in seasonal cool temperate or montane habitats (Wheeler and Baas 1991). In Brazil, a trend for helical thickenings to be prevalent at higher latitudes or in colder climates was also observed, although it was not statistically significant for the sampled species (Alves and Angyalossy-Alfonso 2000).