Volume 234, Issue 5 p. 1863-1875
Full paper
Open Access

Plant migration under long-lasting hyperaridity – phylogenomics unravels recent biogeographic history in one of the oldest deserts on Earth

Tim Böhnert

Corresponding Author

Tim Böhnert

Nees Institute for Biodiversity of Plants, University of Bonn, 53115 Bonn, Germany

Author for correspondence:

Tim Böhnert

Email: [email protected]

Search for more papers by this author
Federico Luebert

Federico Luebert

Nees Institute for Biodiversity of Plants, University of Bonn, 53115 Bonn, Germany

Facultad de Ciencias Agronómicas and Departamento de Silvicultura y Conservación de la Naturaleza, Universidad de Chile, 8820000 Santiago, Chile

Search for more papers by this author
Felix F. Merklinger

Felix F. Merklinger

Nees Institute for Biodiversity of Plants, University of Bonn, 53115 Bonn, Germany

Sukkulenten-Sammlung Zürich/Grün Stadt Zürich, 8002 Zürich, Switzerland

Search for more papers by this author
Dörte Harpke

Dörte Harpke

Leibniz Institute of Plant Genetics and Crop Plant Research (IPK), 06466 Gatersleben, Germany

Search for more papers by this author
Alexandra Stoll

Alexandra Stoll

Centro de Estudios Avanzados en Zonas Áridas Ceaza, 1720256 La Serena, Chile

Instituto de Investigación Multidisciplinar en Ciencia y Tecnología, Universidad de la Serena, 1720170 La Serena, Chile

Search for more papers by this author
Julio V. Schneider

Julio V. Schneider

Botany and Molecular Evolution and Entomology III, Senckenberg Research Institute and Natural History Museum, Frankfurt, 60325 Germany

Search for more papers by this author
Frank R. Blattner

Frank R. Blattner

Leibniz Institute of Plant Genetics and Crop Plant Research (IPK), 06466 Gatersleben, Germany

Search for more papers by this author
Dietmar Quandt

Dietmar Quandt

Nees Institute for Biodiversity of Plants, University of Bonn, 53115 Bonn, Germany

Leibniz Institute of Plant Genetics and Crop Plant Research (IPK), 06466 Gatersleben, Germany

Search for more papers by this author
Maximilian Weigend

Maximilian Weigend

Nees Institute for Biodiversity of Plants, University of Bonn, 53115 Bonn, Germany

Search for more papers by this author
First published: 11 March 2022
Citations: 5

Summary

  • The post-Miocene climatic histories of arid environments have been identified as key drivers of dispersal and diversification. Here, we investigate how climatic history correlates with the historical biogeography of the Atacama Desert genus Cristaria (Malvaceae).
  • We analyze phylogenetic relationships and historical biogeography by using next-generation sequencing (NGS), molecular clock dating, Dispersal Extinction Cladogenesis and Bayesian sampling approaches. We employ a novel way to identify biogeographically meaningful regions as well as a rarely utilized program permitting the use of dozens of ancestral areas.
  • Partial incongruence between the established taxonomy and our phylogenetic data argue for a complex historical biogeography with repeated introgression and incomplete lineage sorting. Cristaria originated in the central southern part of the Atacama Desert, from there the genus colonized other areas from the late Miocene onwards. The more recently diverged lineages appear to have colonized different habitats in the Atacama Desert during pluvial phases of the Pliocene and early Pleistocene.
  • We show that NGS combined with near-comprehensive sampling can provide an unprecedented degree of phylogenetic resolution and help to correlate the historical biogeography of plant communities with cycles of arid and pluvial phases.

Introduction

Hyperarid deserts are among the most challenging environments for plant life, requiring a range of adaptations and strategies for survival. In these marginal habitats global climate variability may drastically impact plant life (Berdugo et al., 2020; Maestre et al., 2021; Shi et al., 2021). Pluvial and arid cycles and the associated expansion and contraction of populations have been proposed as important stimuli for plant evolution (Stebbins, 1952; Carrasco-Puga et al., 2021). We argue that climate-mediated population dynamics lead to complex biogeographic histories that can be traced with modern sequencing techniques and molecular tools (Andrews et al., 2016). At a global level, the spread and diversification of drought-adapted lineages in the late Miocene has been linked to global cooling and concomitant aridification (Zachos et al., 2001; Edwards et al., 2010). Numerous studies on the historical biogeography of particular taxa on the global or continental levels have been published in the recent past (e.g. Wu et al., 2018; Čalasan et al., 2022), but there have been few finer-grained analyses of hyperarid ecosystems at the regional or landscape levels (Larridon et al., 2015; Merklinger et al., 2021), not least because the requisite tools have only recently become available. Here we investigate the historical biogeography of the mallow Cristaria Cav. in the Pacific Desert of South America based on a comprehensive genus-wide sampling and genome-wide single nucleotide polymorphism (SNP) data.

The Chilean Atacama Desert (18°S–30°S) on the west coast of South America is one of the driest places on earth with arid to hyperarid conditions prevalent since the early Miocene or the Oligocene (Dunai et al., 2005). It is widely accepted that there has been a trend of increasing post-Miocene aridity in the Atacama Desert, interrupted by pluvial periods since the early Pliocene (Fedorov et al., 2006; Jordan et al., 2014; Baker & Fritz, 2015; Ritter et al., 2018a), corresponding, e.g. to the interglacial periods over the last 250 kyr (Ritter et al., 2019). Vegetation cover is highly patchy and unevenly distributed in the Atacama Desert. At the southern limit (30°S, near La Serena; see Fig. 1c) vegetation cover is found from the coast to the high Andes and is characterized by a mixture of floristic elements belonging to the Mediterranean floristic region of central Chile as well as elements from the desert vegetation of the Atacama Desert. Further north plant life is increasingly restricted to a narrow coastal strip on the one hand and the Andean foothills between 1000 and 3000 m above sea level (asl) on the other hand. The coastal and Andean vegetation belts are widely separated by the hyperarid desert core (‘Pampa’) and are highly divergent in floristic composition (Luebert & Pliscoff, 2017; Ruhm et al., 2020). Rainfall cycles in the Atacama Desert are strongly influenced by the El Niño Southern Oscillation (ENSO) phenomenon leading to irregular mass flowering events (Vidiella et al., 1999). The coastal vegetation – often referred to as lomas (e.g. Rundel et al., 1991) – is highly patchy and restricted to so-called fog-oases on the coastal range. These lomas are characterized by a unique and species-rich flora with a high degree of endemism (Schulz et al., 2011). A total of approximately 550 species from 225 genera of vascular plants have been reported from the Atacama Desert lomas (Dillon & Hoffmann, 1997), with species richness decreasing from south to north. The vegetation of the Andean foothills is influenced by summer rain and forms a more continuous band mostly dominated by annuals. The coastal lomas formations show a clear floristic break at the Chile-Peru border, while the vegetation of the Andean foothills shows distinct floristic connectivity between Chile and Peru (Ruhm et al., 2020).

Details are in the caption following the image
(a) Distribution of the genus Cristaria in South America (black dots). (b) RAxML tree based on genome wide single nucleotide polymorphism (SNP) data, 70% allowed missing data and Lecanophora as outgroup. Node values represent bootstrap support; only values ≥ 50 are shown. Specimens identified as Cristaria aspera are shown in red. Samples marked with TT are topotypes, asterisk (TT*) indicates a topotype of a nomenclatural synonym. ED and W numbers shown in bold are laboratory identifiers (IDs), which act as unified IDs, helping to connect specimen information with Supporting Information Figs S1, S2; Table S1. Major clades representing accepted species or taxa with considerable support are labeled from 1 to 6d. Each of them is marked in different colors, which correspond to the color coding in Figs 2, 4. (c) Map section of South America showing northern Chile and the Atacama Desert (green = desert vegetation; beige = absolute desert; after Luebert & Pliscoff, 2017) as well as major cities mentioned in the text.

The Atacama flora comprises some taxa that seem to have undergone little or no in situ diversification since entering the region, such as Zygophyllaceae with five native species all having colonized the Atacama Desert separately since the early Miocene (Böhnert et al., 2020). Other genera seem to have undergone considerable diversification in the region and can be considered as species-rich (≥ 10 spp.; Dillon & Hoffmann, 1997). Well known examples are Nolana L.f. (Solanaceae; Tu et al., 2008), Heliotropium L. (Heliotropiaceae; Luebert & Wen, 2008), Oxalis L. (Oxalidaceae; Heibl & Renner, 2012), Copiapoa Britton & Rose (Cactaceae; Larridon et al., 2015), Atriplex L. (Amaranthaceae; Brignone et al., 2019) and Cristaria (Malvaceae; Böhnert et al., 2019). In some cases, diversity is the result of in situ diversification only (e.g. Nolana), in other cases it is also enhanced by multiple colonization events (e.g. Atriplex, Brignone et al., 2019; Cryptantha Lehm. ex G. Don, Boraginaceae, Guilliams et al., 2017; Oxalis, Heibl & Renner, 2012). Processes responsible for the diversification in these groups are poorly understood. Repeated shifts in climatic conditions throughout the Pleistocene might have caused populations to contract during dry phases, causing fragmentation, followed by expansion and secondary contact in pluvial phases. This results in complex biogeographic patterns and puzzling character mosaics as exemplified by the genus Eulychnia Phil. (Cactaceae; Merklinger et al., 2021). Metapopulation dynamics (Hanski, 1998) may facilitate introgression as shown in the genus Palaua Cav. (Malvaceae; Schneider et al., 2011) or result in incomplete lineage sorting (Avise et al., 1987). Dispersal within the Atacama Desert appears to be an important process for establishing new populations (e.g. Tillandsia L., Bromeliaceae; Merklinger et al., 2020) and subsequent divergence (e.g. Nolana; Dillon et al., 2009). Böhnert et al. (2019) were, however, unable to demonstrate that dispersal played a significant role in the diversification of Cristaria in the Atacama Desert.

Muñoz-Schick (1995) recognized 19 species in the genus Cristaria in Chile, a single species is recognized for Peru (Cristaria multifida Cav.; Schneider, 2013). Cristaria is largely restricted to the Atacama Desert, while only two species range into higher elevations of the Andes above 3000 m (Cristaria andicola Gay and Cristaria dissecta Hook. & Arn.; Muñoz-Schick, 1995) and a single species is found on the Desventuradas Islands off the coast of Chile. Cristaria is a widespread and common representative of the Atacama flora with a large number of morphologically weakly differentiated species currently recognized and is therefore ideally suited for a detailed study of historical biogeography and diversification in this extreme ecosystem. Cristaria and its sister genus Lecanophora Speg. were identified as comprising the earliest branching clade of the tribe Malveae (Malvoideae; Böhnert et al., 2019). Böhnert et al. (2019) showed that Cristaria diverged from its sister genus Lecanophora c. 20 Ma and its subsequent radiation appears to coincide with the onset of hyperaridity in the Atacama Desert. The multi-marker plastid phylogeny, however, could only provide limited insights into the phylogenetic relationships within Cristaria (Böhnert et al., 2019). Lack of resolution is a common phenomenon in phylogenetic reconstructions for species-rich, and – more importantly – young clades of the Atacama Desert (Dillon et al., 2009; Luebert et al., 2011; Heibl & Renner, 2012; Larridon et al., 2015). To overcome these limitations, next-generation sequencing (NGS) techniques are increasingly being applied to obtain deeper insights into phylogenetic relationships and speciation processes (Lemmon & Lemmon, 2013; McCormack et al., 2013; Bravo et al., 2019). Merklinger et al. (2021), for instance, successfully used genotyping-by-sequencing (GBS) to unravel the evolutionary and biogeographic history of the Cactaceae genus Eulychnia in the Atacama Desert.

In the present study we therefore aim at resolving the spatiotemporal diversification patterns of Cristaria in the Atacama Desert with a comprehensive approach using GBS, molecular clock dating and two complementary methods to reconstruct the biogeographic history of the genus. We hypothesize that current diversity patterns and ranges of Cristaria within the Atacama Desert are the product of repeated range expansions during pluvial phases after the Miocene–Pliocene transition alternating with genetic isolation during hyperarid phases. If there is a causal link between isolation and divergence we expect the molecular clock estimates of the major biogeographic events (e.g. dispersal or range expansion) to coincide with these humid phases.

Materials and Methods

Taxon sampling

Muñoz-Schick (1995) accepted 19 species of Cristaria for Chile and was unable to confirm the validity of Cristaria concinna Phil. We were able to recollect this species and can confirm its identity as a distinct morphospecies. We therefore accept 20 morphospecies for Chile, plus a single species from Peru (Schneider, 2013). Our taxon sampling comprises 19 of the 21 species of Cristaria here accepted, since we were unable to obtain samples of Cristaria insularis Phil. and Cristaria cordato-rotundifolia Gay. Compared to Böhnert et al. (2019), the sampling could be significantly expanded to 128 samples with up to 42 individuals per species. Two species of Lecanophora were used as outgroup. The sampling is mainly based on our own collecting trips between 2016 and 2019, supplemented with samples from the DNA Bank of the Senckenberg Research Institute (Frankfurt am Main, Germany). Herbarium vouchers are held at herbaria in Chile (ULS), Peru (USM, HUSA) and Germany (BONN, FR). Voucher information is provided in the Supporting Information (Table S1). Morphological identification of specimens and names assigned to the specimens follow Muñoz-Schick (1995) and Schneider (2013).

DNA extraction, library preparation and sequencing

Genomic DNA from silica-dried leaf samples was obtained following the modified extraction protocol of Böhnert et al. (2019) using the NucleoSpin Plant II kit (Macherey-Nagel, Düren, Germany) with an increased amount of lysis buffer (600 instead of 400 µl). DNA quality was checked on 1% agarose gels using Lonza GelStar Nucleic Acid Gel Stain (100×) with 20 ng of double stranded Lambda DNA (N3011S; New England Biolabs, Ipswich, MA, USA). DNA quantification was done with a Qubit 2.0 Fluorometer (Life Technologies). Samples were standardized to 20 ng DNA µl−1 and 15 µl were used for library preparation and sequencing. We followed the library preparation protocol for GBS by Elshire et al. (2011) with the modifications suggested by Wendler et al. (2014); Merklinger et al. (2020). Genomic DNAs were digested using the restriction enzymes PstI-HF® (New England Biolabs, R3140S; recognition site: CTGCA′G) and methylation-sensitive MspI (R0106S; recognition site: C′CGG; New England Biolabs). Subsequently, DNA fragments were size-selected with a SYBR gold-stained electrophoresis gel to 200–600 bp. Individual samples were barcoded and sequenced with 100 bp single-end reads on an Illumina HiSeq 2500 (Illumina Inc., San Diego, CA, USA) at the Genome Center of the Leibniz Institute of Plant Genetics and Crop Plant Research (IPK Gatersleben, Germany). Raw sequences are available online from the European Nucleotide Archive under ID: PRJEB48173. We sequenced 128 samples in total, of which 11 had to be excluded due to low read numbers or sequencing failure, including Cristaria fuentesiana I.M. Johnst., thereby reducing the number of species successfully analyzed to 18.

Assembly and phylogenetic analyses of GBS data

After de-multiplexing the reads (Casava pipeline 1.8; Illumina Inc.), raw sequences were adapter and quality (−q 25, −m 50) trimmed using Cutadapt v.1.16 (Martin, 2011). We checked for read quality and the presence of remaining adapter sequences using FastQC (Andrews et al., 2012). Quality-trimmed and checked GBS data were processed using Ipyrad v.0.9.5 (Eaton & Overcast, 2020). The parameter specifying the datatype was set to ddRAD as our approach uses two restriction enzymes in contrast to the original GBS approach (Elshire et al., 2011). The similarity threshold was set to 90% and all other parameters were left at their default values. Missing data can have significant impact on tree topology and support (Huang & Knowles, 2016; Tripp et al., 2017). Here, 70% missing data were allowed for the actual number of 117 Cristaria ingroup samples but excluding the outgroup samples. For phylogenetic reconstruction we employed a maximum likelihood approach using RAxML v.8.2.10 (Stamatakis, 2014) with the GTRGAMMA substitution model and 500 rapid bootstrap replicates followed by a full search for the best tree. Assembly as well as phylogenetic analyses were conducted within the Ipyrad-analysis toolkit.

Molecular dating and ancestral area reconstruction

A penalized likelihood approach (Sanderson, 2002) implemented in the R (v.3.5.1; R Core Team, 2018) package ape 5.3 (Paradis & Schliep, 2019) was used for molecular dating based on the GBS tree. Stem and crown node ages of Cristaria were set to 20.7 and 7.3 Ma, respectively, based on the family wide fossil-calibrated phylogeny in Böhnert et al. (2019). By applying a cross-validation criterion the best fitting smoothing parameter was identified as lambda = 1.0 (Supporting Information Fig. S1). For the first ancestral area reconstruction, the number of tips from the initial time calibrated tree was reduced to one tip per clade and outgroups were excluded. BioGeoBears v.1.1.1 (Matzke, 2013) was used for the ancestral area reconstruction, testing the three implemented biogeographic models: Dispersal Extinction Cladogenesis (DEC; Ree & Smith, 2008), Dispersal-vicariance-like (DIVA; Ronquist, 1997) and BAYAREALIKE (Landis et al., 2013). All three models were also run with the jump parameter (+J), which incorporates founder events into the model. Despite the ongoing debate (see Ree & Sanmartín, 2018; Matzke, 2021) model selection was based on the Akaike information criterion. The DEC+J model performed best and therefore was used in this study (Table S2). Biogeographic areas were defined based on species occurrences. Therefore, specimens cited in Muñoz-Schick (1995) were georeferenced – together with the collections made for this study – and assigned to one of the 11 clades as defined based on the RAxML tree (see ‘Results’ section; for full list of specimens, see Table S3). Based on this dataset, all point occurrences of each clade were plotted on individual maps. Six distinct geographical ranges were identified and used for the ancestral area reconstruction (see Figs 2,3): southern Peru (area A), the northern part of the Atacama Desert including the northern Chilean Andes (area B), the southern Atacama Desert above 2000 m asl (area C), the southern Atacama Pampa (area D) and the two coastal areas (area E: southern Atacama Coast; area F: northern Atacama Coast). R-scripts are available on Zenodo.org (see the ‘Data availability’ section).

These six areas were used for reconstructing historical biogeography of Cristaria. However, in order to understand the historical biogeography of Cristaria at a finer scale a higher number of areas is required. The number of areas that can be incorporated in BioGeoBears is, however, limited to nine (Lawing & Matzke, 2014). We therefore used the program BayArea (Landis et al., 2013). BayArea, as implemented in the BioGeoBears package (BAYAREALIKE; not used here, see previous paragraph), is also limited to a maximum of nine areas, but the original BayArea method uses a Bayesian sampling approach and allows the user to assign a much larger number of areas. Furthermore, Landis et al. (2013) introduced an alternative approach where the study area is separated into equal sized grid cells. The occurrence records of every individual tip can then be intersected with those grid cells. This model thus permits a quite fine-grained resolution of biogeographic patterns. However, BayArea samples only anagenetic events along branches, while cladogenetic processes are limited to simply copying ancestral ranges (Lawing & Matzke, 2014) and therefore does not consider as many potential biogeographic processes as BioGeoBears.

In the present study, we divided the study area (known range of the genus Cristaria in northern Chile and southern Peru; see Fig. 2) into 50 km × 50 km grid cells and intersected the point occurrence data of every tip of the dated phylogeny of Cristaria resulting in 113 tips and 39 areas with occurrence data. Grid cells were coded as presence/absence data and numbered from 1 to 39 from north to south (see Fig. S2a). BayArea requires coordinates for every area as it calculates distance-dependent dispersal-rates (Landis et al., 2013). Here the centroids of every grid cell were used. The BayArea v.1.0.2 (Landis et al., 2013) analysis was set to run for 500 million MCMC generations with a 10% burn in and a sample frequency of 10 000. Chain convergence was checked in Tracer v.1.7 (Rambaut et al., 2018) and all ESS values were higher than 200. Input and parameter files are available on Zenodo.org (see the ‘Data availability’ section). Visualization of the results was carried out in the geographic information system (GIS) software ArcMap v.10.7 (ESRI Inc., Redlands, CA, USA).

Details are in the caption following the image
Distribution maps of all Cristaria clades. The maps (a–j) show the distribution of a specific clade in relation to its sister clade (shown in small gray dots). Each specimen was plotted with a 20 km buffer in the color of the respective clade in Fig. 1. Collection locations of specimens included in the current phylogenetic analyses are overlaid with black dots. The phylogenetic trees below each map are simplified versions of Fig. 1(b), indicating of which clade point occurrence records are shown (compare Fig. 1 for color code) in relation to the remaining clades (dark gray). Clades or parts of the trees which are shown in light gray, in turn, indicate that occurrence records of these clades are not shown in the corresponding map. The numbers in the bottom right corner of each map also indicate which clades are shown in relation to each other.

Results

GBS assembly and phylogenetic reconstruction

A total of 153M reads from 117 samples (mean = 1.3M/sample; SD = 987K) of the genus Cristaria and two outgroup taxa of the genus Lecanophora were available for phylogenetic reconstruction. The assembly covering all samples including outgroups resulted in 308K reads in total and 2546 reads per sample on average (SD = 710; Table S1). The RAxML analyses of the whole dataset resulted in a phylogenetic tree with a fully supported backbone and six highly supported clades with one to several subclades each comprising samples from one to four morphospecies (Fig. 1; clades labeled from 1 to 6d). The recorded distribution of each clade is presented in Fig. 2. Only Cristaria gracilis (clade 3, Fig. 1) and Cristaria molinae (clade 6c, Fig. 1) are clearly monophyletic, while six subclades comprise samples assigned to two or more morphospecies. Most species are retrieved as nonmonophyletic, with Cristaria aspera Gay (tip labels marked in red; Fig. 1) an extreme example: The seven samples included in the analyses are retrieved in four different clades.

Dating and ancestral area reconstruction using BioGeoBears

The ancestral area reconstruction is provided in Fig. 3, while the full dated GBS phylogeny is provided in Fig. S1. Crown group Cristaria apparently diverged c. 7 Ma (calibrated node), i.e. during the late Miocene, from a widely distributed last common ancestor of the southern Atacama Desert. Clades 2 and 3 both diverged during the late Miocene (6.5 Ma and 5.5 Ma, respectively) in the southern Atacama Pampa (area D). From this region, Cristaria most likely colonized the southern Atacama Coast (area E, clade 2) as well as the Mediterranean Andes and northern Chilean Andes (areas B and C, clade 3). Peru was colonized via the northern Chilean Andes during the early Pliocene (clade 4, Cristaria multifida clade). During the mid-Pliocene, two major lineages diverged from an ancestor that was distributed in the southern Atacama Pampa (area D): First, a lineage comprising the clades 5a, 5b and 5c with an ancestral in the southern Atacama Pampa and Coast (areas D and E). Clade 5a colonized the southern Atacama Coast (area E) in the late Pliocene and, while the two partially sympatric clades 5b and 5c diverged in the early Pleistocene. The second lineage encompasses the remaining four clades (clades 6a–6d), which are endemic to the northern Atacama Coast where their distributions partly overlap (Fig. 2h–j).

Details are in the caption following the image
Ancestral area reconstruction of Cristaria using BioGeoBears plotted on time calibrated RAxML tree which was reduced to one tip per clade. Tips of the tree represent the 11 clades from Fig. 1. Squares between tips and tip labels indicate the distribution assigned to each clade (areas A–F) with color codes corresponding to the areas indicated on the map section of the Atacama Desert and the legend. Pie charts at the nodes depict relative probabilities of ancestral areas as estimated from the Dispersal Extinction Cladogenesis (DEC) analysis with BioGeoBears. Letters next to the pie charts indicate the areas with highest relative probabilities. Green vertical bars represent pluvial phases in the Atacama Desert as dated by Jordan et al. (2014). The arrows in the map section represent the potential dispersal history of the genus.

Ancestral area reconstruction using BayArea

BayArea provides a largely congruent, but much more fine-grained picture of the biogeographic history compared to the BioGeoBears analysis (Fig. 4a–j). The ancestral range of Cristaria is also placed in the central Pampa of the southern Atacama Desert. The common ancestor of the first branching clade (clade 1) colonized both the Mediterranean south and the Andes during the Pliocene (Fig. 4a). The last common ancestor of clade 2 expanded south in the late Miocene/Pliocene (Fig. 4b). Two major northward dispersals are inferred for the Pleistocene (clade 3; Fig. 4c) and Pliocene (clade 4; Fig. 4d). The southern species complex (clade 5a) dispersed towards the southern Atacama coast in the late Pliocene-early Pleistocene (Fig. 4e). Two clades (clades 5b and 5c) remained nearly stationary throughout their biogeographic history (Fig. 4f). The current distribution of clades 6a–6d dates back to a single colonization event towards the coastal area of Taltal in the Pliocene and a subsequent split in two major clades. From there clade 6a expanded further north towards Antofagasta quite recently (Fig. 4g), while clade 6b migrated back southwards in the early Pleistocene (Fig. 4h). Clades 6c and 6d are widely sympatric between Copiapó in the south and Antofagasta in the north and showed similar colonization histories (Fig. 4i,j).

Details are in the caption following the image
Ancestral area reconstruction of Cristaria using BayArea. Biogeographic history is shown per clade on individual maps (a–j). Color coding of the individual clades follow Fig. 1. Large circles represent internal nodes, small dots indicate terminal tips (in case the inferred ancestral area of an internal node is the same as for the following terminal tip an internal node can also represent terminal nodes). Ancestral areas of backbone nodes are given in gray, if a gray dot is surrounded with the clade specific color this indicates that the same area is also an ancestral area of an internal node of the respective clade. Decimal values at circles shown in italics represent corresponding support values for a specific area (compare Supporting Information Fig. S1) and correspond with the nodes in the time calibrated tree in Fig. S2. Values along branches given in bold face represent respective time spans when the specific biogeographic event took place. Please refer to Fig. S2 for a time calibrated tree and subsequently Supporting Information Table S3 for ancestral areas and corresponding support values.

Discussion

Restriction site associated DNA sequencing (RADseq) techniques such as GBS (Elshire et al., 2011; Wendler et al., 2014) are increasingly used to resolve complex diversification processes (e.g. Andrews et al., 2016; Meier et al., 2017; Pérez-Escobar et al., 2017; Vargas et al., 2017; Ahmed et al., 2019). The present study confirms the usefulness of GBS-data to provide a highly resolved picture of the phylogeny and historical biogeography of a group of closely allied morphospecies, in our case of species-rich Cristaria, in the Atacama Desert. Species delimitation in Cristaria is notoriously difficult due to the morphological plasticity with few, if any, consistent diagnostic characters (e.g. Philippi, 1892). Generative characters are of limited use and leaf morphology is therefore extensively employed to distinguish species (Johnston, 1929; Muñoz-Schick, 1995). Our results cast doubt on some of the morphospecies currently accepted, since few of them are retrieved as monophyletic. This is particularly obvious for the morphospecies ‘Cristaria aspera’, which is retrieved in four clades. Here we propose that the high morphological plasticity within sub-clades indicates incomplete lineage sorting due to repeated introgression during Quaternary climate oscillations (Stebbins, 1952; Degnan & Rosenberg, 2009; Avise et al., 2016; Ritter et al., 2019).

The size of the data-set used here helped to resolve relationships in great detail and with high support, but excluded the use of commonly used Bayesian molecular clock dating methods. We therefore used penalized likelihood (Sanderson, 2002), an established maximum likelihood approach, which permits the analysis of large datasets derived from NGS (e.g. Niissalo et al., 2022). Penalized likelihood methods do not provide confidence intervals and therefore, together with the calibration scheme applied here, suggest a level of precision that is, of course, misleading in terms of its accuracy and therefore should be interpreted with caution. Like in all historical biogeography reconstructions, the dates here obtained are estimates for purposes of temporal correlation. However, our data provide a highly stratified historical biogeography of Cristaria based on two complementary approaches. We present an incremental approach to identify meaningful biogeographic units based on clade specific distribution data for the analysis in BioGeoBears (Matzke, 2013). Choosing appropriate units for ancestral area reconstructions is a key aspect in historical biogeography (Morrone & Crisci, 1995). We here inferred biogeographic units based on the distribution of each individual clade in relation to the remaining taxa. In order to get a more detailed understanding we additionally applied a fine-scale, grid-based approach as implemented in BayArea (Landis et al., 2013), which, to our knowledge, has rarely been used (e.g. Pečnerová et al., 2015). This approach provides analytical detail that to the best of our knowledge has not been reached in any other study due to the unlimited number of geographical units paired with comprehensive sampling of multiple individuals per species across the entire range (compare Fig. 2). We understand the estimated ancestral areas as geographical niches characterized by a set of climatic and ecological conditions rather than distinct present-day geographical ranges.

Ancestral range of Cristaria in the miocene

Böhnert et al. (2019) demonstrated that the divergence of Cristaria and its sister genus Lecanophora coincided with Andean uplift phases between 30 and 15 Ma (Scott et al., 2018). An in-depth discussion on the influence of Andean orogeny on plant life in the Atacama Desert is given by Böhnert et al. (2019). Similar timing for trans-Andean vicariance has been proposed for the origin of Heliotropium L. sect. Cochranea (Miers) Kuntze (Luebert & Wen, 2008) and Bulnesia Gay (Zygophyllaceae; Böhnert et al., 2020).

Climatic conditions in the region under study are influenced by wind circulations transporting moisture from the south during austral winter. These are reinforced during the irregularly occurring ENSO events (Houston, 2006). It has been suggested that during pluvial phases these westerlies lead to conditions similar to present-day ENSO events (i.e. permanent El Niño; Fedorov et al., 2010). These phases would have led to a northward displacement of vegetation zones, e.g. an expansion of the present-day Mediterranean zone of central Chile into the southern part of the present-day Atacama Desert (Luebert & Pliscoff, 2017), alternating with southward shifts during arid periods. The relic rain forest islands of the Fray Jorge National Park with a remarkable floristic similarity to the Valdivian temperate rainforests of southern Chile have been considered as such relics from a pluvial period (del-Val et al., 2006). The early-branching (late Miocene) and most widespread clades of Cristaria are found exclusively in the winter-rain region of north-central Chile, which argues for phylogenetic niche conservatism (Pyron et al., 2015), regarding their response to changing climatic conditions. In fact, several plant genera from the Atacama Desert show a surprisingly high number of early branching endemics in this region, e.g. Heliotropium sect. Cochranea (Luebert & Wen, 2008), Argylia D. Don. (Bignoniaceae; Gleisner & Riccardi, 1969; Glade-Vargas et al., 2021), Eulychnia (Merklinger et al., 2021) and several members of Brassicaceae (Toro-Núñez et al., 2015).

Climate cycles in the Atacama Desert coincide with biogeographic events

The onset of prevailing hyperarid conditions in the Atacama Desert is generally dated to the early Miocene, or even earlier (Dunai et al., 2005; Evenstar et al., 2017). Short pluvial phases interrupted hyperaridity throughout the Miocene, but also in the Pliocene and Pleistocene (Fedorov et al., 2010; Jordan et al., 2014; Evenstar et al., 2017; Ritter et al., 2018b). This climatic variability has been shown to have left a signature in the historical biogeography of the Atacama Desert (Heibl & Renner, 2012; Merklinger et al., 2021). Most of the early cladogenesis in Cristaria appears to have taken place in the southern Atacama Pampa at the end of an extensive hyperarid phase in the late Miocene and during the beginning of a pluvial period in the early Pliocene (Jordan et al., 2014): Four of the six major clades originated during this interval. These have current distributions in the southern Atacama Pampa (clades 1 and 3), Mediterranean Chile (clades 1, 2 and 5a), respectively the Andes (clade 1). One of these clades colonized Peru in the early Pliocene, the Cristaria core group subsequently expanded towards the Chilean coast and diverged into four subclades comprising the majority of the extant lineages by the mid Pliocene. Luebert & Wen (2008) linked this first pluvial phase in the early Pliocene and the subsequent arid phase (Jordan et al., 2014) to the diversification of a major clade of Heliotropium sect. Cochranea. Similar spatiotemporal patterns of diversification have been correlated with the climatic cycles throughout the Pliocene in several other species-rich genera of the Chilean and Peruvian coastal deserts (e.g. Malesherbia sect. Malesherbia Ruiz & Pav., Passifloraceae, Gengler-Nowak, 2002; Nolana, Tu et al., 2008; Tiquilia Pers., Ehretiaceae, Moore & Jansen, 2006).

Two early diverging clades (clades 3 and 4) expanded further north along the Andean foothills. Increasing climatic fluctuations during the Pliocene probably led to repeated expansions of early Cristaria into what is today the dry core of the Atacama Desert. Multiple pluvial phases leading to more humid conditions across the entire Atacama have been documented since the end of the Miocene (Sáez et al., 2012; Jordan et al., 2014; Ritter et al., 2018a). Cristaria showed several northward expansions with three species migrating further north along the Andean foothills, one of them reaching Peru (Muñoz-Schick, 1995; Schneider, 2013). The specimens of Cristaria dissecta from the northern Chilean Andean foothills are nested in Peruvian Cristaria multifida, representing southern outliers of that clade and thus a biogeographical and genetic link as previously proposed by Muñoz-Schick (1995). This underscores the floristic connectivity along the Andean foothills of Chile and Peru documented by Ruhm et al. (2020) and Gengler-Nowak (2002). However, the two ancestral area reconstructions applied in this study differ between the estimated biogeographic histories of Cristaria multifida: The BayArea model indicates long-distance dispersal to coastal southern Peru and a Pleistocene southward expansion into northern Chile. The DEC+J model suggests a northward dispersal along the Andean foothills. Both scenarios require a shift from winter into a summer rain ecology of the ancestor of Cristaria multifida, two major climatic regimes that, during arid phases, are separated by a pronounced climatic barrier called the Western South American Dry Diagonal (WSADD; Luebert, 2021).

The Atacama coast was colonized twice separately. Cristaria glaucophylla is found on the southern Atacama coast as well as the coast of Mediterranean Chile (clade 5a) mostly in sandy coastal habitats (Muñoz-Schick, 1995). The colonization of these environments correlates with a pluvial phase along the Pliocene–Pleistocene transition, followed by a southward expansion of this species. Some specimens morphologically resembling Cristaria glaucophylla and here identified as such are found in clade 2. Representatives of both clades (clades 2 and 5a) have reached the coastal area of the transition zone between the desert and central Chile (c. 30°S) at the same time, but possibly via different routes (Fig. 4b,e). Therefore, we argue that secondary contact and subsequent processes of introgression seem to be the best explanation for the conflicting positions of Cristaria glaucophylla, as has been previously argued for similar patterns in the genus Palaua (Schneider et al., 2011).

Approximately half of the morphospecies currently recognized belong to clade 6 and are distributed along the Atacama Desert coastline from 28.5°S to 22°S, representing the second colonization of the Atacama coast during the second half of the Pliocene. Four lineages diverged during the late Pliocene–Pleistocene transition during a pluvial phase between 3.6 and 4 Ma (Jordan et al., 2014). Up to 10 morphospecies have been recognized in these coastal clades (clades 6a–6d) in the past (Muñoz-Schick, 1995; Böhnert et al., 2019), but our phylogenetic data only retrieve four well-supported clades showing little agreement with accepted species limits. Repeated secondary contact and introgression are likely reasons for this conflict between morphology and phylogeny. The Taltal region plays a crucial role as a hub for the lineages of clade 6d (Fig. 4). This region is well known for its high species richness and presumably long-term habitat stability (Johnston, 1929; Dillon & Hoffmann, 1997; Schulz et al., 2011). We propose that an ancestral population of clade 6d was isolated in this climatically favorable niche around Taltal (see Fig. 1c) after the onset of hyperaridity in the core desert and subsequently dispersed and diversified north and south along the coast.

Conclusions

The data here presented provide a uniquely detailed picture of the phylogeny and historical biogeography of Cristaria, showing complex patterns of range expansions and migrations and a striking correlation between the Miocene onset of hyperaridity and subsequent cycles of mesic and pluvial phases during the Pliocene and Pleistocene and cladogenesis and diversification. The analyses underscore the utility of GBS for resolving complex biogeographic histories in recently diverged taxa. The spatio-temporal diversification patterns of Cristaria are congruent with several other species-rich genera of the Atacama Desert, showing an overall south–north diversification and a clear segregation between coastal and Andean lineages. The incremental approach for identifying biogeographic areas based on current ranges, but especially the use of the BayArea program paired with near-comprehensive sampling across the entire range are shown to represent powerful tools for unravelling a recently diverged species complex. Our study provides additional support for a growing body of evidence indicating a Miocene origin of many plant groups in the Atacama Desert. As previously argued, speciation in this ancient desert appears to be rather recent and took place largely in the Pliocene–Pleistocene. We argue that cycles of isolation and secondary contact likely explain the puzzling character mosaics observed in Cristaria, reflected in its controversial taxonomy and problems with morphological species delimitation. Future studies should be directed towards identifying the possible role of Quaternary hybridization and introgression in present-day phenotypic signatures. Current species concepts are only partially congruent with phylogenetic relationships and taxonomic adjustments are required.

Acknowledgements

The authors highly appreciate the help during the laboratory work by Claudia Schütte, Karola Maul and Nicole Schmandt as well as support and technical help with herbarium specimens by Thomas Joßberger. The authors would also like to thank the staff of the Senckenberg DNA Bank (Frankfurt am Main, Germany) for providing DNA samples. For fruitful discussion on paleoclimatic dating the authors highly appreciate the support of Dr Benedikt Ritter (University of Cologne, Germany). The authors gratefully acknowledge sequencing support by the Genome Center of the Leibniz Institute of Plant Genetics and Crop Plant Research (IPK Gatersleben, Germany). Plant material from Peru was collected under the Resolución de Dirección General no. 158-2019-MINAGRI-SERFOR-DGGSPFFS. This study was funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) – Projektnummer 268236062 – SFB 1211. Finally, the authors would like to thank all colleagues in the framework of the Atacama project: Earth – Evolution at the dry Limit (http://sfb1211.uni-koeln.de/).

    Author contributions

    TB, FL and MW designed the study. TB, FL, FFM, AS and JVS conducted fieldwork and contributed data. TB and JVS determined plant collections. FRB, DH and DQ supported sequencing. TB curated and analyzed the data with substantial support from FL. TB wrote the manuscript with edits from FL, FFM, AS, JVS and MW. All authors made edits and approved the final manuscript.

    Data availability

    Raw sequence reads for the GBS Illumina runs were deposited in the European Nucleotide Archive under study accession no. PRJEB48173. Additional data and results are available from the open science platform zenodo.org (https://doi.org/10.5281/zenodo.6328566).