Volume 108, Issue 2 p. 184-199
Free Access

Quaternary diversification of a columnar cactus in the driest place on earth

Felix F. Merklinger

Corresponding Author

Felix F. Merklinger

Nees Institute for Biodiversity of Plants, University of Bonn, Germany

Sukkulenten-Sammlung Zürich / Grün Stadt Zürich, Zürich, Switzerland

Author for correspondence (e-mail: [email protected])

Contribution: Conceptualization (equal), Data curation (equal), Formal analysis (equal), ​Investigation (equal), Methodology (equal), Visualization (equal), Writing - original draft (equal), Writing - review & editing (equal)

Search for more papers by this author
Tim Böhnert

Tim Böhnert

Nees Institute for Biodiversity of Plants, University of Bonn, Germany

Contribution: Conceptualization (equal), Visualization (equal), Writing - original draft (equal)

Search for more papers by this author
Mónica Arakaki

Mónica Arakaki

Museo de Historia Natural, Universidad Nacional Mayor de San Marcos, Lima, Peru

Contribution: Data curation (equal), ​Investigation (equal), Writing - original draft (equal)

Search for more papers by this author
Maximilian Weigend

Maximilian Weigend

Nees Institute for Biodiversity of Plants, University of Bonn, Germany

Contribution: Project administration (equal), Supervision (equal), Writing - review & editing (equal)

Search for more papers by this author
Dietmar Quandt

Dietmar Quandt

Nees Institute for Biodiversity of Plants, University of Bonn, Germany

Leibniz Institute of Plant Genetics and Crop Plant Research (IPK), Gatersleben, Germany

Contribution: Project administration (equal), Resources (equal), Supervision (equal), Writing - review & editing (equal)

Search for more papers by this author
Federico Luebert

Federico Luebert

Nees Institute for Biodiversity of Plants, University of Bonn, Germany

Departamento de Silvicultura y Conservación de la Naturaleza, Universidad de Chile, Santiago, Chile

Contribution: Conceptualization (equal), Data curation (equal), Formal analysis (equal), ​Investigation (equal), Methodology (equal), Project administration (equal), Supervision (equal), Writing - original draft (equal), Writing - review & editing (equal)

Search for more papers by this author
First published: 12 February 2021
Citations: 14



The cactus family (Cactaceae) is a speciose lineage with an almost entirely New World distribution. The genus Eulychnia with eight currently recognized species is endemic to the Atacama and Peruvian Deserts. Here we investigated the phylogeny of this group based on a complete taxon sampling to elucidate species delimitation and biogeographic history of the genus.


A family-wide Bayesian molecular clock dating based on plastid sequence data was conducted to estimate the age of Eulychnia and its divergence from its sister genus Austrocactus. A second data set obtained from genotyping by sequencing (GBS) was analyzed, using the family-wide age estimate as a secondary calibration to date the GBS phylogeny using a penalized likelihood approach. Ancestral ranges were inferred employing the dispersal extinction cladogenesis approach.


Our GBS phylogeny of Eulychnia was fully resolved with high support values nearly throughout the phylogeny. The split from Austrocactus occurred in the late Miocene, and Eulychnia diversified during the early Quaternary. Three lineages were retrieved: Eulychnia ritteri from Peru is sister to all Chilean species, which in turn fall into two sister clades of three and four species, respectively. Diversification in the Chilean clades started in the early Pleistocene. Eulychnia likely originated at the coastal range of its distribution and colonized inland locations several times.


Diversification of Eulychnia during the Pleistocene coincides with long periods of hyperaridity alternated with pluvial phases. Hyperaridity caused habitat fragmentation, ultimately leading to speciation and resulting in the current allopatric distribution of taxa.

The cactus family (Cactaceae) is a prominent and species-rich lineage of succulent plants with an almost entirely New World distribution (Barthlott and Hunt, 1993). The family is thought to have originated shortly after the Eocene–Oligocene global drop in CO2 levels and had a subsequent diversification during the Miocene in parallel to the expansion of arid environments (Arakaki et al., 2011). Cacti have colonized a broad range of semi- to hyperarid habitats from Patagonia to southern Canada. In southern South America, the Bolivian and northern Argentinian South Central Andes and the Atacama and Peruvian Deserts are considered as centers of diversity (Barthlott et al., 2015). The predominantly Chilean genus Eulychnia Phil. has a checkered taxonomic history—the most recent taxonomic account of the genus was produced by Ritter (1980, 1981), but many of the taxa he described are currently considered synonyms (Hunt, 2016), and the genus is in urgent need of revision. Eulychnia is a columnar cactus, usually attaining a shrubby, small tree-like or sometimes decumbent habit. Its species are most easily identified by their broadly campanulate flowers and a pericarpel often covered in dense wool or even spines (Fig. 1). The distribution of the genus is largely coastal, in particular in the northern half of its range, where the taxa are confined to the fog zone of the coastal cordillera. Taxa in the southern half of the distribution range occur in a more Mediterranean-like environment and are sometimes found further inland. Overall, the majority of taxa are found in the transition from arid to hyperarid regions of northern Chile (Luebert and Pliscoff, 2017). Only a single species of Eulychnia is known from southern Peru, with its populations separated from its Chilean relatives by ca. 1000 km (Fig. 2). In contrast, Austrocactus Britton & Rose, the sister genus of Eulychnia (Hernández-Hernández et al., 2011, 2014), is found mainly east of the Andes in Patagonia, with only a few records from the Andes of central Chile (Sarnes and Sarnes, 2012). The geographical center of distribution of Eulychnia is found in the coastal cordillera of the Atacama Desert, and the genus is one of 17 genera of Cactaceae in this region (Lembcke and Weisser, 1979; Hunt, 2016).

Details are in the caption following the image
Morphological diversity within Eulychnia. (A) E. castanea, decumbent habit. (B) Flower of E. castanea with typical spiny pericarpel. (C) E. acida near the type locality in Illapel. (D) Flower of E. acida with the typical pericarpel lacking a conspicuous indumentum. Flower color varies from white to light pink. (E) E. acida var. elata, massive form of E. acida at the type locality Estancia Castilla. (F) Flower of E. acida var. elata lacking the conspicuous indument on the pericarpel. (G) E. acida var. procumbens, decumbent habit. (H) Flower of E. acida var. procumbens with short, dark wool on the pericarpel. (I) Typical shrubby habit of E. breviflora. (J) Typical, densely woolly pericarpel of E. breviflora.
Details are in the caption following the image
Distribution of Eulychnia in Chile and Peru. Note the restriction of taxa to the coast north of Copiapó. Most populations of E. iquiquensis between Antofagasta and Iquique have disappeared, also most likely those in Arica, increasing the distance to the disjunct taxon E. ritteri in Peru to approximately 1000 km. Distribution of taxa based on herbarium vouchers (see Appendix S4).

The Atacama Desert of northern Chile is generally considered as one of the driest places on Earth (Dunai et al., 2005), with a modern hyperarid core receiving less than 10 mm of precipitation per year (Houston and Hartley, 2003). This extreme aridity is caused by (1) the desert’s position at the subtropical high-pressure belt, which has been stable since the late Jurassic (Hartley et al., 2005), (2) the cold Humboldt current along the Pacific coast restricting moisture uptake by onshore winds, and (3) the Andes preventing the entry of moist Amazonian air-masses (Rundel et al., 1991). While the precise timing of the onset of aridity in the Atacama Desert remains a matter of debate, recent studies of climate evolution and aridity in this region support the notion that aridity has not developed uniformly and, as such, generalizations for the entire Atacama Desert should be avoided (e.g., Sillitoe and McKee, 1996; Hartley and Chong, 2002; Dunai et al., 2005; Latorre et al., 2006; Rech et al., 2006; Evenstar et al., 2017). However, the stable position of the South American continent for the last 150 Myr (Hartley et al., 2005) in combination with the establishment of the Peru–Chile Current system (PCC) approximately 50 Ma (Cristini et al., 2012) has led to the generally accepted conclusion that the Atacama Desert is an ancient desert, with a hyperarid core since at least the Miocene (Dunai et al., 2005) or even earlier (Hartley et al., 2005). This long-lasting aridity was, however, repeatedly interrupted by wetter (though still semiarid) phases largely coinciding with globally warmer periods as shown by evidence obtained from 14C-dated vegetation fragments from rodent middens (Betancourt, 2000; Latorre et al., 2006), radiocarbon dates of fossil vegetation from the hyperarid core (Nester et al., 2007; Gayo et al., 2012), cosmogenic nuclide exposure dating (Ritter, Binnie, et al., 2018; Ritter, Stuart, et al., 2018), and palaeoclimatic reconstructions based on a drill core (Ritter et al., 2019). Present-day Atacama vegetation is restricted to the coastal cordillera that benefits from occasional winter precipitation and the influence of coastal fog (Rundel et al., 1991; Schulz et al., 2011) and the Andean foothills, receiving summer rain. These vegetation zones are separated by the hyperarid core, an area virtually devoid of plant life (Rundel et al., 1991; Luebert and Pliscoff, 2017). In spite of the overall hyperarid conditions, the northern coastal Atacama harbors a surprisingly high number of vascular plants (~550 species), with a high percentage of endemism of >60% (Dillon and Hoffmann, 1997). Regular advection fog that meets the west-facing slopes of the coastal cordillera facilitates the establishment of fog oases (lomas), in which the majority of species occur (Rundel et al., 1991). Although few molecular phylogenetic studies have investigated the origin, divergence times, and causes for diversification of Atacama Desert lineages, individual studies have demonstrated a correlation between Andean uplift, the onset of hyperaridity in the Atacama, and diversification in parallel to aridification (Gengler-Nowak, 2002; Luebert and Wen, 2008; Dillon et al., 2009; Heibl and Renner, 2012; Böhnert et al., 2019). Despite these studies, our knowledge of the evolution of the Atacama flora and, in particular, associated abiotic drivers of plant diversification in this extremely arid climate remains fragmentary.

In a recent evolutionary study based on 13 samples of Eulychnia, including seven accepted species, Larridon et al. (2018) obtained a phylogeny that defined two well-supported clades, with the sampled taxa falling into a northern and southern group corresponding to the main morphological characters used to distinguish between Eulychnia species. Unfortunately, support within these clades was weak. Here, we generated a comprehensive dated phylogeny of Eulychnia to investigate its biogeographic history and diversification. Our data set includes sequences from plants collected at the type localities for all previously published names, some of which are placed in synonymy of currently eight accepted species (Philippi, 1860; Cullmann, 1958; Ritter, 1964, 1980; Eggli et al., 1995; Hoxey and Klaassen, 2011; Guerrero and Walter, 2019). We attempted to resolve the evolutionary and biogeographic history of Eulychnia, specifically to estimate (1) the divergence time between Eulychnia and Austrocactus, (2) correlate this divergence to the onset of aridity of the Atacama Desert, (3) elucidate the processes of diversification within Eulychnia, and (4) infer its ancestral ranges. The analysis of data obtained from genotyping by sequencing is a relatively new approach toward the study of the cactus family and also to the study of the Atacama Desert flora in general.


Taxon sampling

A complete taxon sampling of the genus Eulychnia was gathered for the present study, based on our own collections realized between 2016 and 2019 in Chile and Peru. The type localities of all but one previously published taxon were visited and sampled (Philippi, 1860, 1864; Cullmann, 1958; Ritter, 1964, 1980, 1981; Eggli et al., 1995; Hoxey and Klaassen, 2011; Guerrero and Walter, 2019). The only population not sampled by us in the field was that of E. aricensis F.Ritter from Cerro Camaraca South of Arica. This population is likely extinct, and an attempt to re-collect material of this locality was not successful, and cultivated material from this locality introduced by Ritter was used instead. Vouchers were deposited in the herbaria BONN, EIF, K, and ULS (Table 1; Appendix S4). Collections from the type localities (locotypes) are here labeled with the oldest available name at the species or variety level. This decision does not reflect a taxonomic judgement. An additional subset of several species of Austrocactus was obtained from the living collection of E. & N. Sarnes (Eschweiler, Germany) grown from seed material. Previous studies recovered Austrocactus as the sister genus of Eulychnia (Hernández-Hernández et al., 2011, 2014), and we included these samples to obtain age estimates for the split between these two genera. A further sample of Corryocactus Britton & Rose was included as an outgroup since this genus was retrieved as sister to the PHB clade (Pachycereeae, Hylocereeae and three genera formerly included in Browningieae: Castellanosia, Neoraimondia and Armatocereus; Hernández-Hernández et al., 2011). A two-step analysis was conducted using (1) a three-plastid-marker alignment of the whole Cactaceae based on the work of Hernández-Hernández et al. (2014) to obtain age estimates for the split between Eulychnia and Austrocactus and (2) a data set of Eulychnia, Austrocactus, and two Corryocactus samples obtained from genotyping by sequencing (GBS) to produce a fully resolved phylogeny based on genome-wide single-nucleotide-polymorphism (SNP) data. A comprehensive list of all taxa used for this study, including GenBank accession and voucher information, are provided in Appendix S4.

TABLE 1. Eulychnia samples used for phylogenetic analyses. Taxa with an asterisk indicate type localities. For these locotype collections, the oldest available name at the species or variety level was used.
Taxon Sample no. Country Locality trnK-matK trnL-trnF rpl16 GBS
Austrocactus bertinii Britton & Rose ED3490 Argentina Sierra Grande x x x x
A. coxii (K.Schum.) Backeb. ED3492 Argentina Nahuel Huapi x x x x
A. sp. ED3491 Argentina Rio Senguer x x x
A. spiniflorus (Phil.) F.Ritter ED3494 Argentina Farellones x x x x
A. subandinus E.Sarnes & N.Sarnes ED3493 Argentina Los Molles x x x x
Corryocactus brevistylus Britton & Rose ED5158 Chile Chusmiza x
C. erectus (Backeb.) F.Ritter ED4540 Peru Huancarpay x
Eulychnia acida Phil. ED5135 Chile Tongoy x
E. acida Phil. ED5138 Chile Quebrada Los Choros x
E. acida Phil. ED5137 Chile Cuesta Buenos Aires x
E. acida var. elata F.Ritter ED5143 Chile Canto del Agua x
E. acida var. elata F.Ritter ED5152 Chile Caleta Pajonales x
E. acida var. elata F.Ritter ED5153 Chile Nantoco x
E. acida var. elata* F.Ritter ED3114 Chile Estancia Castilla x x x x
E. acida var. procumbens F.Ritter ED5139 Chile Llano Choros x
E. acida var. procumbens F.Ritter ED4234 Chile Llano Choros x x x x
E. acida var. procumbens F.Ritter ED5140 Chile Domeyko-Freirina x
E. acida var. procumbens* F.Ritter ED3113 Chile Freirina x x x x
E. acida* Phil. ED3110 Chile Illapel x x x x
E. aricensis* F.Ritter ED4541 Chile Arica x x x x
E. barquitensis* F.Ritter ED3117 Chile Barquito x x x x
E. breviflora* Phil. ED3111 Chile Coquimbo x x x x
E. breviflora Phil. ED5144 Chile Quebrada Copiapó x
E. breviflora Phil. ED5150 Chile Chungungo x
E. breviflora Phil. ED5141 Chile Tres Playitas x
E. breviflora Phil. ED5142 Chile Carrizal Bajo x
E. breviflora var. tenuis F.Ritter ED5145 Chile El Morro x
E. breviflora var. tenuis* F.Ritter ED3116 Chile Caldera x x x x
E. castanea Phil. ED5134 Chile Totoralillo x
E. castanea Phil. ED4667 Chile El Toro x
E. castanea Phil. ED5136 Chile Tongoy x
E. castanea* Phil. ED3109 Chile Los Molles x x x x
E. iquiquensis (K.Schum.) Britton & Rose ED3122 Chile Punta Gruesa x x x x
E. iquiquensis (K.Schum.) Britton & Rose ED5149 Chile Alto Chipana x
E. iquiquensis var. pullilana* F.Ritter ED3120 Chile El Cobre x x x x
E. morromorenoensis* F.Ritter ED3121 Chile Cerro Moreno x x x x
E. ritteri Cullm. ED4430 Peru Lomas de Atiquipa x
E. ritteri* Cullm. ED3489 Peru Lomas de Atiquipa x x x x
E. saint-pieana F.Ritter ED5146 Chile El Caleuche x
E. cf. saint-pieana F.RItter ED5147 Chile La Madera x
E. cf. saint-pieana F.Ritter ED5148 Chile Las Tórtolas x
E. saint-pieana* F.Ritter ED3118 Chile Chañaral x x x x
E. spec. 1 ED3115 Chile Piedra Colgada x x x x
E. spec. 2 ED3112 Chile La Serena x x x x
E. taltalensis (F.Ritter) Hoxey ED5151 Chile Paposo x
E. taltalensis* (F.Ritter) Hoxey ED3119 Chile Taltal x x x x
E. vallenarensis* P.C.Guerrero & Helmut Walter ED4802 Chile Panamericana km 645 x x x x


  • GBS, genotyping by sequencing

Our definitions of the terms such as arid and hyperarid follow those used for the spatial distribution of ombrotypes by Luebert and Pliscoff (2017).

Molecular methods

For both data sets, genomic DNA was extracted from silica-dried stem tissue using the NucleoSpin Plant II kit (Macherey-Nagel, Düren, Germany) and the manufacturer’s protocol but with an increased incubation time of 90 min and an increased volume of lysis and binding buffers to reduce viscosity of the lysate. Following the work of Hernández-Hernández et al. (2014), three plastid DNA regions (trnK-matK region, trnL-trnF region, rpl16 group I intron) were amplified using the primer combination and PCR cycling conditions given in Appendix S5a–d. To obtain high-quality reads, we purified the PCR products using gel extraction and the NucleoSpin Gel and PCR Clean-up kit (Macherey-Nagel, Düren, Germany) according to the manufacturer’s protocol. Sequencing was performed at a sequencing service on a 3730XL DNA Analyzer (Applied Biosciences, Waltham, MA, USA), and DNA sequences were manually edited using PhyDe v. 0.9971 (Müller et al., 2005).

Because higher quality and purity are needed for DNA used for high throughput sequencing, extracted DNA was electrophoresed on 1% agarose gels using Lonza GelStar Nucleic Acid Gel Stain (100x; Lonza Bioscience, Basel, Switzerland), and a sample of 20 ng linear, double-stranded Lambda DNA (New England Biolabs, N3011S; Ipswich, MA, United States) was added for testing the quantity and quality of the genomic DNA. Qubit Fluorometer (Life Technologies, Carlsbad, CA, United States) measurements were taken to assign a numerical value to the agarose-gel readings. Samples were standardized to 20 ng/µL, and aliquots of 15 µL per sample were used for library preparation and sequencing.

Library preparation and genotyping by sequencing protocols followed those of Merklinger et al. (2020). Genomic DNA (200 ng) was digested with restriction enzymes PstI-HF (New England Biolabs, R3140S) and MspI (New England Biolabs, R0106S), followed by size selection, individual barcoding, and single-end sequencing on the Illumina HiSeq 2500 (Illumina, San Diego, CA, USA).

Barcoded reads were de-multiplexed using the CASAVA pipeline v. 1.8 (Illumina). The obtained raw sequence reads (0.6–3 million per individual) were adapter- and quality-trimmed with a phred score of >25 using CUTADAPT v. 1.12 (Martin, 2011). Reads shorter than 50 bp after adapter removal were discarded. Sequence reads for the GBS Illumina runs were deposited in the European Nucleotide Archive under study accession PRJEB39114.

A de novo assembly of the GBS data of initially 47 taxa belonging to Eulychnia, Austrocactus, and the outgroup Corryocactus was carried out using ipyrad v. 0.9.17 (Eaton and Overcast, 2020). Different output files were generated with a minimal number of samples per locus set to 8, 12, 16, and 20, respectively, the maximum cluster depth within samples set to 0.9, and the ploidy level set to diploid. For all other parameters, the default settings of ipyrad were used.

Analytical methods

(1) The complete taxon sampling of Hernández-Hernández et al. (2014) was downloaded from GenBank and complemented with our original sequence data obtained from Sanger sequencing for 23 individuals of Eulychnia and Austrocactus. The final data set was manually aligned using PhyDe v. 0.9971 (Müller et al., 2005). Five hairpin-associated inversions of 109 bp were detected in the whole alignment (two in trnK-matK, two in trnL-trnF and one in rpl16) and reverse-complemented and aligned for analysis following Quandt et al. (2003). Additionally, 18 hotspots (Borsch et al., 2003) with a total of 228 bp were detected in all three DNA regions and excluded from analysis. Positions and sizes are documented in the supplementary files available on the CRC1211-database (see data availability). Tree topology was tested for congruence with those presented by Hernández-Hernández et al. (2014) applying maximum likelihood (ML) analyses using RAxML v. 8.2.9 (Stamatakis, 2014). The GTRCAT substitution model was specified, and the analyses were run with 1000 rapid bootstrap replicates, treating each gene region as a single partition. The final tree was visualized using the python package toytree (Eaton, 2019).

(2) RAxML was used to infer maximum likelihood trees based on concatenated supermatrices of each of the most-inclusive data sets (min12) obtained from the GBS assembly in ipyrad. The GTR+Γ substitution model was used with 20 tree searches and 1000 bootstrap replicates to calculate node support. A species tree based on SVDquartets (Chifman and Kubatko, 2014) under multispecies coalescence was estimated using TETRAD as implemented in ipyrad with 1000 bootstrap replicates.

STRUCTURE v. 2.3.4 (Pritchard et al., 2000), also implemented in ipyrad was used to cluster individuals into K distinct populations based on the min12 SNP data set (30,100 sites after filtering), and using the imap dictionary to group individuals into populations. The minmap command was specified at 0.5, requiring that 50% of samples have data in each group. Multiple values for K (2–6) were tested, and 20 replicates were run per test. Each replicate was run for 500,000 Markov chain Monte Carlo (MCMC) steps. A burn-in of 100,000 was applied. Results were visualized and exported using toyplot v. 0.18.0 (Shead, 2014).

Molecular clock dating

(1) According to the work of Hernández-Hernández et al. (2014), Bayesian relaxed clock analyses using BEAST 2.5.1 (Bouckaert et al., 2014) were conducted for the whole Cactaceae plastid data set to estimate divergence times of the separation of Eulychnia and Austrocactus. BEAUTI 2.5 (Bouckaert et al., 2014) was used to set up an XML file. The partitions of the three chloroplast DNA (cpDNA) regions were unlinked with respect to site model, but linked with respect to clock and tree models. Further, we specified a birth–death model as tree prior and applied a relaxed lognormal clock with an estimated clock rate (Drummond et al., 2006; Gernhard, 2008). Since there are no fossil records within Cactaceae, we used a secondary calibration following Hernández-Hernández et al. (2014) with a uniform prior distribution and a lower value of 22.71 and an upper value of 42.43 for the Cactaceae crown node. The MCMC was run for 75 million generations, sampling every 7500 generations. The log file was checked in Tracer v. 1.71 (Rambaut et al., 2018), and TreeAnnotator produced a maximum clade credibility tree (MCCT) using mean heights, a burn-in of 10% and a posterior probability limit of 0.95. Finally, R packages ape v. 5.0 (Paradis and Schliep, 2019), phyloch v. 1.5-5 (Heibl, 2008 onward), strap v. 1.4 (Bell and Lloyd, 2015), and geoscale v. 2.0 (Bell, 2015) were used in R v. 3.5.1 (R Core Team, 2018) to plot and annotate the dated tree. Phylogenetic and dating analyses were conducted on the CIPRES Gateway (Miller et al., 2010).

(2) Penalized likelihood analysis for GBS data: To estimate divergence times within Eulychnia, we first tested the molecular clock hypothesis of the data set using the R package treedater v. 0.3.0 (Volz, 2019). We used the GBS-based ML phylogeny of Eulychnia and Austrocactus and the penalized likelihood (PL) approach (Sanderson, 2002) as implemented in the R package ape. We conducted a cross validation (Sanderson, 2002; Paradis, 2012) to determine the best value of the smoothing parameter lambda and used that value to obtain divergence time estimates with the PL method. Two outgroup samples of Corryocactus were removed from the analysis. We calibrated the stem node of Eulychnia and Austrocactus and the crown nodes of Eulychnia and Austrocactus, respectively, setting three calibration points as minimum ages corresponding to the median ages obtained from our BEAST analysis (see above).

Ancestral area reconstruction

To reconstruct the range evolution of Eulychnia, we used the dispersal extinction cladogenesis (DEC) approach (Ree and Smith, 2008) as implemented in the R package BioGeoBEARS v. 1.1.1 (Matzke, 2013). The DEC+j model was not employed due to reported statistical problems (Ree and Sanmartín, 2018). Two analyses were run with differing assumptions in the number of maximum areas, which were set to two and three, without dispersal constraints over time, as their effect on the results is negligible (Chacón and Renner, 2014). We used the time-calibrated ML tree based on the GBS data set and removed most of the 44 accessions, to be left with 14 samples, one per well-supported clade, each representing an operational taxonomic unit (OTU), which was designated based on our locotype collections using the oldest available name at the species or variety levels. The analysis was conducted for these taxa, including one sample of Austrocactus to assess the directionality of this mostly trans-Andean distribution, and a lineage through time (LTT) plot was generated with the R package ape. Every sample in the phylogeny was assigned to one of six areas based on the distribution patterns of Eulychnia in Chile and Peru and Austrocactus in Patagonia. These geographic units corresponded to (A) Patagonia, (B) southern Atacama: inland, (C) southern Atacama: coastal, (D) northern Atacama: inland, (E) northern Atacama: coastal, and (F) Peru. The assignment of OTUs to geographical areas was based on information obtained from our own collections and from herbarium vouchers revised at the herbaria of the Zürich Succulent Collection (ZSS), the Natural History Museum in Santiago de Chile (SGO), and the Naturalis Biodiversity Center in Leiden (L). Additional voucher images were consulted online at the herbaria in Kew (K), Halle (HAL) and Concepción (CONC). Our first division into geographic areas was based on the distribution of Austrocactus in Patagonia and of Eulychnia in Chile and Peru. Secondly, we separated “North” and “South”, with the border between this separation set at the Huasco River (Coquimbo region), reflecting the transition from arid to hyperarid climate of the Atacama Desert, coinciding with the N–S limit of several OTUs. Further, we differentiated between “coastal” and “inland” because some taxa such as E. castanea and E. breviflora have a strictly coastal distribution, while others occur farther inland. Lastly, we defined “Peru” as its own geographical area reflecting the disjunct distribution of E. ritteri. The geographical distribution of each taxon for the two analyses is documented in Appendix S6.


cpDNA phylogeny and divergence times in Cactaceae

The final alignment of the three cpDNA regions encompassed 248 taxa and had a length of 6065 bp after exclusion of 18 hotspots with a length of 228 bp and reverse-complementing five hairpin-associated inversions. The BEAST ML analyses of whole Cactaceae (Fig. 3; Appendices S1, S2) confirmed the phylogenetic relationships of Eulychnia within the core Cactoideae and as sister to Austrocactus with high branch support (BS = 100%, PP = 1). Within Eulychnia, two well-supported clades were retrieved (BS = 100%, PP = 1). The first clade includes E. breviflora and the other Chilean taxa, E. iquiquensis and E. taltalensis (BS = 97%, PP = 1). The Peruvian taxon E. ritteri is sister to this E. breviflora clade with high support (BS = 82%, PP = 0.97). A second clade was retrieved for the E. acida group including E. castanea, E. acida var. procumbens, and E. vallenarensis, with E. castanea as sister to the other taxa in this clade, and with E. acida (type locality sample ED3110) as sister to E. acida var. procumbens, E. vallenarensis, and E. acida var. elata (BS = 83%, PP = 1). While the branch support for the major clades separating morphologically distinct groups is high, the support values within the respective clades are low (BS < 70%, PP < 0.95).

Details are in the caption following the image
Section of the BEAST-dated plastid phylogeny based on the sampling of Hernández-Hernández et al. (2014) and including our additional samples, showing the relationships and age estimates within Eulychnia and Austrocactus. Numbers above nodes represent posterior probabilities, those below nodes represent bootstrap values. The full phylogeny is available in Appendix S1.

The last common ancestor of Eulychnia and Austrocactus was dated to the Miocene–Pliocene transition at 6.70 Ma (95% HPD: 3.28–10.51). The crown node age for Eulychnia was retrieved for the early Pleistocene (2.19 Ma, 95% HPD: 0.88–3.84). The split between Peruvian E. ritteri and the Chilean E. breviflora clade is dated to the middle to upper Pleistocene (1.49 Ma, 95% HPD: 0.55–2.72). Diversification within the two Chilean Eulychnia clades was dated to the middle Pleistocene for both the E. breviflora clade (0.68 Ma, 95% HPD: 0.23–1.29) and the E. acida clade (0.82 Ma, 95% HPD: 0.22–1.62).

Genotyping-by-sequencing phylogeny and biogeography

The ML analysis based on SNP data for Eulychnia and Austrocactus including Corryocactus as an outgroup retrieved a tree with the split between Eulychnia and Austrocactus fully supported (Fig. 4). Eulychnia itself is resolved with Peruvian E. ritteri sister to all other (Chilean) taxa. The Chilean clade is further resolved into two sister clades, clade A and clade B with full support (BS = 100%). Clade A includes E. breviflora, E. iquiquensis, E. taltalensis, E. breviflora var. tenuis, E. spec. 1 and E. saint-pieana/barquitensis. Clade B includes E. acida, E. acida var. elata, E. acida var. procumbens, E. vallenarensis, and E. castanea. Taxa in both clades roughly follow a geographical order from South to North. The support for clade A was significant (BS = 97%), with most of the subclades showing full support (BS = 100%). Clade B received maximum statistical support of consistently BS = 100% with the exception of the E. spec. 2, E. castanea, and E. acida subclade (BS = 61%) and the E. vallenarensis and E. acida var. elata subclade (BS = 66%).

Details are in the caption following the image
(A) Maximum likelihood phylogeny and (B) TETRAD species tree based on the GBS. Numbers at nodes represent bootstrap values.

The tree obtained from the TETRAD analysis is congruent in the major clades, but with some differences in the relationships between sister taxa. Eulychnia is again monophyletic and retrieved as sister to Austrocactus. Eulychnia ritteri is sister to both Chilean clades, also fully supported E. acida var. procumbens is sister to E. vallenarensis in the northern E. acida clade, while in the RAxML analysis, E. vallenarensis is sister to E. acida var. elata. Overall, statistical support in the TETRAD phylogeny is lower than in the RAxML tree.

The STRUCTURE analysis (Fig. 5) revealed three distinct genetic groups for the various values of K = 2–6, with these groups generally corresponding to the three major clades retrieved in the RAxML analysis. There is a clear separation of taxa belonging to the E. acida and E. breviflora groups at K = 2, with E. ritteri genetically predominantly part of the E. breviflora group. At K = 3, E. ritteri has started to separate into its own genetic group. At K = 4–6, E. ritteri is distinct, as are the E. acida and the E. breviflora groups. Noteworthy is the position of E. acida from Cuesta Buenos Aires and E. castanea from Tongoy, both of which retain a mixed genetic signal from the E. acida and E. breviflora groups, respectively.

Details are in the caption following the image
STRUCTURE analysis showing the clustering of individuals into K = 2–6 distinct populations based on the min12 SNP data set.

The test of the molecular clock hypothesis failed to reject the strict clock (coefficient of variation of rates ≈ 0). Accordingly, the cross validation resulted in a best smoothing parameter λ = 100. The penalized-likelihood dating of the RAxML tree (Fig. 6) provided congruent results to the dated plastid phylogeny in terms of divergence times: the split between Eulychnia and Austrocactus is retrieved for the Miocene–Pliocene transition at 6.70 Ma. For the crown node of Eulychnia, an age of 2.20 Ma is inferred, separating the Peruvian taxon E. ritteri from the Chilean congeners in the early Pleistocene. The crown node age of the Chilean taxa is dated to the early Pleistocene at 2.16 Ma when they separated into the E. acida and E. breviflora clades. The crown node age for the E. breviflora clade was retrieved for the early Pleistocene at 2.11 Ma, when it further separates into a southern and a northern clade, with the northern clade of E. breviflora and relatives separating into four subclades at 2.03 Ma, still during the early Pleistocene, with these subclades in geographical order from north to south. The crown node age for E. acida and relatives was dated to 2.07 Ma in the early Pleistocene, with the samples again following a south to north order in two subclades separating at 1.97 Ma; the southern one included E. castanea and E. acida, and the northern subclade included an E. acida sample as sister to all other taxa northward, including E. acida var. procumbens, E. vallenarensis, and E. acida var. elata.

Details are in the caption following the image
Penalized likelihood phylogeny and ancestral area reconstruction based on the genotyping by sequencing data set and tested for two different settings, maximum areas = 2 and maximum areas = 3.

Ancestral area reconstruction

The two analyses with two and three maximum areas resulted in a likelihood difference of 0.93 (LnL = −32.35 and LnL = −31.42, respectively; Fig. 6). The ancestral area reconstruction with maximum areas of two indicates a separation of an ancestral taxon of Austrocactus and Eulychnia between Patagonia and the south coastal Atacama during the Miocene–Pliocene transition, though no highly likely ancestral area was recovered. During the Pliocene, the coastal Atacama range of Eulychnia separated from Patagonia. Peru was colonized in the early Pleistocene from the northern coastal Atacama. With ancestral taxa widely distributed along the Atacama coast from south to north, ancestors of the E. acida clade apparently disappeared over most of the range, surviving only at the southern coast. The origin of the E. acida clade is the southern coastal Atacama, with two independent inland colonizations in the south (E. acida and E. acida var. procumbens) and one inland colonization in the north (E. acida var. elata). The origin of the E. breviflora clade in the early Pleistocene was estimated at the coast, with most extant members restricted to the northern coastal Atacama, but with one inland colonization in the northern Atacama (E. spec. 1).

In contrast, the maximum area setting of three indicates a higher probability of a coastal-Peruvian origin of Eulychnia. Vicariance between Peru and Chile led to the divergence of E. ritteri and a colonization along the Chilean coast by an ancestral Eulychnia during the early Pleistocene. From here, the processes are congruent with those of the first analysis.


Our results from the plastid phylogeny, the GBS RAxML phylogeny and the TETRAD analysis provide a largely congruent tree topology. In all three analyses, two clades of Chilean Eulychnia were retrieved, one designated as Eulychnia breviflora clade and a second clade including E. acida and relatives. These two groups are supported by general morphology, with individuals of the E. breviflora group sharing a densely woolly pericarpel and a higher number of softer spines, and those of the E. acida group having a nonhairy (but sometimes spiny) pericarpel and fewer but coarser spines (Fig. 1). Pericarpel indumentum has been used as the most reliable diagnostic character in Eulychnia since Britton and Rose (1920). The fact that several species appear nonmonophyletic may be explained by several reasons. First, there is a difference between species and names. The names we have applied to different populations are those available in the literature, but they do not necessarily reflect real taxa. That is likely the case of the infraspecific taxa in E. breviflora and E. acida. In these cases, we may need to recognize that there is cryptic diversity—morphological characters that have not yet been assessed, and that current diagnostic characters are perhaps insufficient to separate these infraspecific taxa as different species. Alternatively, using broader species concepts in Eulychnia would also eliminate the problem of nonmonophyly. Further, hybridization between taxa has been suggested between, e.g., E. acida and E. castanea (Eggli and Leuenberger, 1998), pointing to gene flow between these taxa and supporting our idea of incomplete speciation. An updated taxonomic revision of the genus should address these issues.

The position of E. ritteri, which in the plastid phylogeny is sister to the E. breviflora group, but in the GBS RAxML tree and TETRAD analysis is sister to all Chilean Eulychnia is interesting. This taxon was in the past considered a subspecies of E. iquiquensis (E. breviflora clade; Hunt, 2006), or alternatively, as a species in its own right (Hunt, 2016; Larridon et al., 2018). The incongruence between the chloroplast phylogeny and the GBS and TETRAD phylogenies may be due to several reasons. Firstly, it may be due to the different phylogenetic signals from the different data sets. While the chloroplast regions are of maternal inheritance, the loci obtained from the genotyping by sequencing data are distributed across the whole genome (Davey and Blaxter, 2010). The clades with polyphyletic relationships in our chloroplast phylogeny were moderately to highly supported (e.g., E. ritteri as sister to the E. breviflora clade; E. morromorenoensis as sister clade to all other northern taxa), which points to both recent and ancestral interspecific hybridization/introgression as likely reasons. While hybridization/introgression is difficult to be distinguished from incomplete lineage sorting (Joly et al., 2009), introgression among Eulychnia taxa is likely to occur, even with the current scenario of allopatric distribution of taxa and hybrid zones have been observed at the periphery of various taxa (e.g., Ritter, 1980; Eggli and Leuenberger, 1998; Hoxey and Klaassen, 2011). Individuals that did not form a clade in the chloroplast phylogeny formed clades with other conspecific members in the GBS tree (with the exception of two individuals that had been collected as of hybrid origin due to morphological features, such as E. castanea ED5136 and E. acida ED5137, two samples which also showed great admixture in our STRUCTURE analysis). Chloroplast captures stemming from introgression events are common in plants (Rieseberg and Soltis, 1991; Acosta and Premoli, 2010), and the incongruence we see in our chloroplast phylogeny may also be due to the heterospecific origin of the chloroplast genome. In the GBS tree, the two Chilean clades further separate into a southern and northern clade, comprising taxa that fall into a clear south–north geographical area, with the exception of a sample of E. castanea from Tongoy, which was retrieved as sister to a sample of E. breviflora from Coquimbo, a few kilometers north of Tongoy. On the basis of the results of the STRUCTURE analysis and its close geographic position to the distribution range of E. breviflora, we interpret this sample as the result of introgression between E. castanea and E. breviflora. The retrieved split into two major clades of Eulychnia into a southern and northern clade corroborates the results of Larridon et al. (2018), and we favor the hypothesis of a climatic transition into a hyperarid environment as an explanation. Further studies on Chilean Cactaceae (Guerrero et al., 2019) also found such a north–south division, although unfortunately, the authors provided neither age estimates nor an explanatory hypothesis. The transition between the two clades seems to be the region around Copiapó. Here the southern species’ distribution ends and the distribution of the northern species begins, with the exception of E. breviflora along the coast. The coast was perhaps better connected in the past, possibly through the distribution of the guanaco, which may explain the continuous distribution of E. breviflora there. It is also possible that refugia existed where, due to more humid conditions, taxa persisted even during arid phases. Based on populations from these refugia, an expansion southward from northern populations and a parallel expansion northward from southern populations with subsequent colonizations inland (as shown by our ancestral area reconstruction) seems plausible. Similar expansion events have been reported for, e.g., Nolana (Ossa et al., 2013, 2017). A scenario of isolated refugia could be understood as a more extreme form of the present-day isolated lomas separated by hyperarid habitat (Rundel et al., 1991), which reconnect to become more continuous during climatically favorable periods, and the diversification processes of species under these circumstances seems compatible with the extremely short time periods of diversification that we obtained from our study. The study by Larridon et al. (2018) was based on six chloroplast markers, and E. ritteri was retrieved as sister to the E. breviflora clade, albeit with moderate support. By including almost twice as many samples in the plastid phylogeny and expanding our methods to include a GBS phylogeny with more than three times as many samples than in the previous study by Larridon et al. (2018), we were able to study the diversification processes within Eulychnia in much more detail and also expand more precisely on the age estimates previously provided (Hernández-Hernández et al., 2014).

Evolution of Eulychnia at the Miocene–Pliocene boundary

Our results identify a last common ancestor of Eulychnia and Austrocactus at the Miocene–Pliocene boundary, with a minimum crown node age of 6.70 Myr. Previous studies provided age estimates for the crown node of Eulychnia and Austrocactus at 4.9 Myr and a stem age of divergence at 9.17 Myr (Hernández-Hernández et al., 2014). Although our estimates overlap and lie within the error rates of this study, the discrepancy in ages obtained may be traced back to our increased taxon sampling. While Hernández-Hernández et al. (2014) only included a single species of Austrocactus and two species of Eulychnia, our taxon sampling for Eulychnia is the most comprehensive to date, and we also included several samples of Austrocactus. It is well known that increasing the number of species tends to increase the age of the clades (Linder and Rieseberg, 2004). A further source of these differences may lie in the alignment, which we conducted anew and excluded hotspots (see Methods) that may have impacted the substitution regime. In any case, our age estimates agree with previous studies, postulating the origin of the majority of species-rich clades in the Cactaceae as well as the emergence of other succulent lineages in North America and Africa at the Miocene–Pliocene boundary, ca. 10–5 Ma (Arakaki et al., 2011; Hernández-Hernández et al., 2014). This simultaneous global emergence of succulent lineages suggests a common trigger, most likely a shift in global climate. While the evolution of novel pollination syndromes or local adaptation to edaphic factors may play an important role in driving speciation of succulent lineages (Ellis et al., 2006; Good-Avila et al., 2006; Kellner et al., 2011), the underlying processes that have been used to explain the genetic differentiation and thus species diversification of these lineages are linked to shifts in climate (Trejo et al., 2016; Scheinvar et al., 2017), and as such, biotic and abiotic processes are linked with each other. The Miocene–Pliocene boundary is characterized by gradual cooling and an expansion of the ice sheets in West Antarctica (Alpers and Brimhall, 1988; Zachos, 2001). The Antarctic ice-sheet expansion led to the cooling of the Peru–Chile Current system (PCC) and deep Pacific waters and took place in parallel with a global cooling trend (Lamb and Davis, 2003), resulting in the establishment of (semi-) arid conditions in Africa (Horn et al., 2014; Klak et al., 2004) and North America (Moore and Jansen, 2006), and an overall expansion of grasslands dominated by C4 photosynthetic species (Edwards et al., 2010). According to Lamb and Davis (2003), the cooling of the PCC correlates with the most rapid phase of central Andean uplift and the Plio-Pleistocene epoch, when the Andes for the first time exceeded an average elevation of ~3 km (Gregory-Wodzicki, 2000). The cooling of the PCC and the now increasing orographic barrier to moist air from the Amazon basin imposed by the Andes would have caused an additional shift from long-term aridity (Dunai et al., 2005) to hyperaridity. The split of the two genera Eulychnia and Austrocactus coincides with this transition and can possibly be attributed to changes in vegetation triggered by this increased aridity. Since the distribution of these genera is separated by the current spread of the Mediterranean sclerophyllous woodlands of Central Chile (Luebert and Pliscoff, 2017), this split might be related to the emergence of these woodlands. Paleoecological studies have suggested that the sclerophyllous vegetation of central Chile may have originated as a consequence of the development of the South American arid diagonal during the late Miocene resulting from the major phase of Andean uplift (Hinojosa and Villagrán, 1997; Armesto et al., 2007), roughly coinciding with the split between Austrocactus and Eulychnia. We must emphasize here that although in the present study, we followed the definition of Luebert and Pliscoff (2017) for hyperaridity, there are discrepancies in estimates among the scientific community pertaining to the different levels of aridity in the past and the present. So far, there is no consistent definition of hyperaridity or consensus on the timing of onset of hyperaridity (see also the review by Garreaud et al. [2010] on this matter).

Quaternary diversification of Eulychnia in the Atacama Desert

Our biogeographic analysis finds a high probability for a coastal origin of Eulychnia with a coastal range evolution and several independent inland colonizations (Eulychnia spec. 1 and Eulychnia acida, E. acida var. elata, and E. acida var. procumbens). Ritter (1980) proposed a southern coastal Atacama origin of Eulychnia based on the assumption that E. castanea should be regarded as the most-basal representative of the genus. Although Ritter does not elaborate on this statement, the current distribution of E. castanea at the southernmost edge of Eulychnia (Los Molles, 32°S) is geographically the closest to the northernmost locality of Austrocactus (Farellones, 33°S) in Chile. A possible explanation of the divergence of E. castanea and E. acida, the only two taxa at the southern end of Eulychnia distribution, may be found in peripatric speciation processes (Losos and Glor, 2003; Eduardo Palma et al., 2005), with E. castanea forming divergent populations at the southwestern edge of the geographical range of E. acida. This process could also be used to interpret the speciation of other taxa farther north, and a process that, again, might ultimately be driven by climatic variables.

Our biogeographical analysis is unable to confidently identify a southern or northern coastal origin. Similar studies in the cactus genus Copiapoa (Larridon et al., 2015) proposed a northern origin with a subsequent southward range evolution. However, considering the current disjunct distribution of several non-Andean desert cacti that are restricted to either Peru or Chile (Rundel et al., 1991; Pinto and Luebert, 2009), generalizing statements on geographic origins should best be avoided until more studies have become available.

The Peruvian E. ritteri is clearly retrieved as an isolated, early lineage of Eulychnia and cannot go back to recent long-distance dispersal. However, an early-dispersal event cannot be ruled out, in particular considering the large, juicy fruits of some Eulychnia taxa, which may in the past have been dispersed by now extinct herbivore mammals (Cares et al., 2018). The guanaco (Lama guanicoe), for example, once widespread from northern Peru to Tierra del Fuego but now distributed in less than 30% of its range at the time of arrival of Europeans to South America (Marin et al., 2013 and references therein), may have acted as an important disperser of Eulychnia seeds. While animal dispersal may explain the geographical extension of Eulychnia as a whole, it does not shed light into the current geographical differentiation within Eulychnia because the past distribution of the guanaco may have remained continuous in the Atacama Desert throughout the Quaternary (Politis et al., 2011; González et al., 2013; Marin et al., 2013). Rather, the present-day isolated populations of guanaco in Chile and adjacent countries appear to be a human-induced phenomenon based on the introduction of livestock, hunting, habitat loss, and altering climatic requirements (Baldi et al., 2001; Castillo et al., 2018). Under a scenario of climate-induced vegetation shifts in the past, populations of Eulychnia would have experienced isolation and secondary contact: Eulychnia originated in the Pliocene and diversified in a very short time span around 2 Ma (Figs. 3, 6; Appendix S3) into an already arid environment. The Pliocene–Pleistocene experienced several pluvial phases that were interrupted by stages of marked aridity (Jordan et al., 2014; Ritter et al., 2018), particularly between approximately 2.65 to 1.27 Ma, which agrees well with our crown node age for Eulychnia dated to 2.2 Ma. Pluvial phases might have enabled Eulychnia populations to expand, forming extensive meta-populations ranging into Peru, perhaps similar to the wide-ranging E. acida populations of the southern Atacama we see today (Merklinger, 2018). The onset of hyperarid conditions during the early Pleistocene would have led to habitat fragmentation, promoted speciation, and possibly caused the disjunction of the Peruvian E. ritteri from the Chilean congeners. Despite a reported floristic break along the coast between northern Chile and southern Peru (Ruhm et al., 2020), several studies have reported a similar pattern of disjunct species distribution in other plant groups, such as Malesherbia, Nolana, and Cristaria (Gengler-Nowak, 2002; Dillon et al., 2009; Böhnert et al., 2019). Although we favor here the contraction–expansion model of populations during climatic oscillations as the ultimate driver for speciation, we do not dismiss the importance of newly emerging pollination syndromes or local adaptation to edaphic factors as other important reasons for Eulychnia diversification. These aspects could be explored in further studies based on the analyses of population-level data. Pollination syndromes in cacti are considered not stable (Nyffeler and Eggli, 2010), and other Chilean cacti have been shown to fit into this concept (Guerrero et al., 2019); however, Eulychnia flowers are functionally conserved across taxa and throughout the range of the genus. Larridon et al. (2018) provided interesting insights into the possible causes of the current allopatric distribution patterns of the different Eulychnia species, based on climatic gradients and precipitation regimes linked with geographic changes and associated bioclimates, ombrotypes, and vegetation zones (Larridon et al., 2018 and references therein). These ideas are congruent with our hypothesis, that ultimately, climatic oscillations shaped the habitat occupied by and the diversification of Eulychnia. In the study by Larridon et al. (2018) and in our own results presented here, taxa clustered into groups that largely correspond to geographical areas. For example, E. breviflora var. tenuis and E. saint-pieana clustered together, rather than with E. breviflora (to which they were originally thought to be most closely related), and these two taxa occur only between the Río Copiapó and Chañaral. Farther north, E. taltalensis clustered with E. iquiquensis. The former is distributed only between Taltal and El Cobre, an area that may well have acted as a refugium during arid phases, an idea that is supported also by other taxa present only here, such as Tillandsia tragophoba (Bromeliaceae; Zizka et al., 2009) and for other coastal areas of the northern Atacama Desert (Ossa et al., 2013). The second clade obtained by Larridon et al. (2018) including E. acida and relatives also shows a similar picture of allopatric distribution patterns in Eulychnia. Ritter (1980) already stated that the Huasco River forms a natural barrier between E. acida to the south and E. acida var. elata to the north. This observation is confirmed by the results of Larridon et al. (2018) that retrieved E. acida var. elata as sister to the southern E. acida samples. Further, E. acida var. procumbens is restricted to the Llano Choros and the Huasco Valley and may be the result of speciation in this area due to more favorable conditions, even during persistent aridity in the surrounding areas. Overall, these results are in line with our hypothesis, that climatic oscillations are ultimately responsible for driving speciation in Eulychnia, supported by repeated isolation of populations with subsequent expansions and reconnections to form new species assemblages.

It is likely that during predominantly arid conditions, Eulychnia was restricted to the more humid coastal cordillera, as shown by our ancestral area reconstruction, with bursts of range expansion during the later Pleistocene pluvial phases. Range expansions would have promoted secondary contact of populations not yet fully genetically isolated from each other and so explain the current distribution of Eulychnia and the potential hybridization that appears to take place in areas of contact between putative, morphologically different species. According to Ritter et al. (2018), there were very long periods of aridity during the Pleistocene, periods long enough to allow species to evolve (Levin, 2019). Considering also the numerous potential hybrids that have been observed (Ritter, 1980; Eggli and Leuenberger, 1998; F. F. Merklinger and F. Luebert, personal observations), also hybrid speciation (Levin, 2019) must be considered a relevant process, even for a potentially long-lived species with long generation times as is the case in Eulychnia. The mosaic of local climates and variable precipitation regime we see today is in line with our hypothesized oscillation of aridity particularly during the Pleistocene as the main driver for Eulychnia diversification, and it also explains the present-day distribution of taxa in southern Peru and northern Chile.


Our particular thanks goes to Elisabeth and Norbert Sarnes, Andrew Gdaniec, and Jörg Schneider for contributing plant material for this study. We gratefully acknowledge Claudia Schütte (Nees Institute, Bonn) and Susanne König and Axel Himmelbach (IPK, Gatersleben) for guidance and support with laboratory work. Thanks to Julius Jeiter for support with the compilation of figures, Maria Anna Vasile for discussing the BioGeoBEARS R script, and two anonymous reviewers whose comments helped to improve the manuscript. This work was funded by the German Research Foundation (DFG) – Project 268236062 – SFB 1211 (http://sfb1211.uni-koeln.de/). Research and collection permits in Peru were granted by SERFOR, RDG N 280-2019-MINAGRI-SERFOR-DGGSPFFS.


    F.F.M., T.B., M.A., D.Q., and F.L. undertook fieldwork. F.F.M. generated the sequence data. F.F.M., T.B., and F.L. designed the study and did the analyses. All authors contributed to writing and critical revision of the manuscript draft.


    The authors declare that all research was conducted independently of any commercial or financial relationships that could be interpreted as a conflict of interest.

    Data Availability

    Raw sequence reads for the GBS Illumina runs were deposited in the European Nucleotide Archive under study accession PRJEB39114.