The importance of environmental conditions in maintaining lineage identity in Epithelantha (Cactaceae)

Abstract The use of environmental variables to explain the evolution of lineages has gained relevance in recent studies. Additionally, it has allowed the recognition of species by adding more characters to morphological and molecular information. This study focuses on identifying environmental and landscape variables that have acted as barriers that could have influenced the evolution of Epithelantha species and its close genera. Our results show that soil pH, isothermality, temperature seasonality, and annual precipitation have a significant phylogenetic signal for Epithelantha. Soil type and landforms are also relevant as ecological barriers that maintain the identity of Epithelantha species. The variables associated with the soil (pH) have influenced the evolution of Epithelantha and probably in other genera of Cactaceae. Additionally, Epithelantha is frequent in the piedmont and haplic kastanozems. Bioclimatic variables reinforce the recognition of E. micromeris, and E. cryptica as independent species. Therefore, ecology can be considered as a factor to explain the high level of endemism in Cactaceae.

Soil characteristics are factors that could also influence lineage diversification in plants (Anacker & Strauss, 2014). Soil plays an essential role in determining the distribution of plant communities and in consequence the distribution of endemic taxa (Bárcenas-Argüello et al., 2013;Burge, 2014). Del Castillo (1996) identified species as calcareous-tolerant and calcareous-intolerant, being calcareous soils the ones that harbor more endemism. Soil characteristics have been identified as a limiting factor in the distribution of Echinocactus platyacanthus (Trujillo, 1984), and Ariocarpus (Aguilar- Morales et al., 2011). This condition is not exclusive to Cacteae, as it has been reported in Cephalocereus (Tribe Echinocereeae, Bárcenas-Argüello et al., 2010), and Opuntia plus Cylindropuntia from the Chihuahuan desert (Subfamily Opuntioideae; Lebgue-Keleng et al., 2014).
To study the impact of environmental factors on speciation, we focused on the genus Epithelantha, which due to its moderate diversity and endemicity can serve as a model without the confounding factor of geographic distance. This genus is nested in a clade composed of Ariocarpus, Rapicactus, Strombocactus, Turbinicarpus, and Kadenicarpus (Aquino et al., 2019;Vázquez-Sánchez et al., 2019).
The six genera are mainly distributed in the ecoregion called the Chihuahuan desert, and neighboring regions in North and Northeastern Mexico (Figure 1a). Epithelantha is the subject of several taxonomic controversies. Aquino et al. (2019) recognized 10 species based on a combination of morphological and molecular characters (Figure 1b). However, E. micromeris and E. cryptica, can only be identified by morphological characters, thus are good candidates to test the relevance of environmental factors maintaining species identity (Rissler & Apodaca, 2007). Environmental information has been used to identify cryptic species of other plant groups such as Leucanea (Fabaceae, Govindarajulu et al., 2011) and

| Raw geographic and environmental data
In order to characterize the environmental preferences each taxon,

| Phylogenetic reconstruction and dating
We constructed a sequence data matrix using the following cpDNA regions: petL-psbE, psbA-trnH, trnL-F, and trnQ-rps16. The DNA extraction, amplification, and sequencing methods are described in Aquino et al. (2019), and GenBank accession numbers can be found in Table S1. The coding of indels and morphological characters can be found in Aquino et al. (2019). A calibrated Bayesian phylogenetic reconstruction was constructed using a representative of each species

| Environmental characterization
Categorical variables (soil type and landform) were analyzed using a contingency matrix, registering the frequency of observations of each species for each variable. Based on that, the expected obser- The bioclimatic variables that were used for ancestral state reconstruction were chosen by first evaluating their independence using Pearson's correlation coefficients (Table S2). The variables used were annual mean temperature, isothermality, temperature seasonality, mean temperature in the driest quarter, mean temperature of the coldest quarter, annual precipitation, and soil pH. In all cases, the mean was estimated and was used to reconstruct the ancestral states based on the calibrated phylogenetic tree (Table S3).
The reconstruction of ancestral states was performed with the R package phytools (Revell, 2012;R Development Core Team, 2018).
Ancestral state reconstruction of discrete variables (landform and soil type) was performed by first evaluating character evolution models using fitpolyMK implemented by phytools (Revell, 2012).
First, we reduced the number of character states by keeping only those states that were represented in 70% of the sample. For soil type we had kastanozems, lithosols and xerosols; and for landforms, the final character states were fold mountains, karstic formations, river systems, plains, and piedmonts. The data was considered unordered since at least one taxon had more than 2 character states.
We used the function fitpolyMK to test for two models: ER and transient, and for each dataset we kept the model with the lowest AIC. The best fitting model for soil type was ER-unordered and for landforms was transient-unordered. We used the state transition matrix generated by fitpolyMK to reconstruct the ancestral character states using make.simmap for 100 trees.
Turbinicarpus has a moderate correlation to rendzines (Pearson's The goodness of fit test applied to evaluate the association between species and soil type was significant (p = .00269, Figure 3b) The test between genera and landforms was also significant (p = 6.943 e−10 , Figure 3c). Epithelantha is commonly found in piedmonts (Pearson's r = 4.055) and has negative association with plains

| Phylogenetic signal and ancestral state reconstruction
The estimated Blomberg's K and Pagel's λ were significant for Bio3  F I G U R E 4 Ancestral character state reconstruction for four environmental variables whose phylogenetic signal was significant F I G U R E 5 Ancestral character state reconstruction for three environmental variables whose phylogenetic signal was not significant | 4527 AQUINO et Al. λ = 1.022611). The ancestral state reconstruction of variables can be seen in Figure 4, where the genus Epithelantha shows higher pH values (7.48-8-03) overall. The exception is E. ilariae with the lowest average (pH = 7.84), similar to Ariocarpus (pH = 7.81) and Rapicactus (pH = 7.84). The genera Kadenicarpus, Turbinicarpus, and Strombocactus are found in neutral soils (pH = 7.01-7.18) ( Figure 4b). Regarding isothermality (Bio3 = 49.31-51.56), Epithelantha has the lowest values, which means that temperature is more variable (Figure 4c), with the exception of E. potosina, which has an isothermality similar to the sister genera The ancestral state reconstruction model fitting for soil type showed a complex pattern due to the high levels of intraspecific polymorphism and the unordered nature of characters (Figure 6a, Figure S1a,b). Soil type in particular showed no phylogenetic signal, as all soil types were present in the genus Epithelantha and the outgroup. However, there were differences between sister species, like E. polycephala (lithosols) and E. pachyriza (xerosols) (Figure 6a).
The results for landforms were even more complex, due to the high number of character states (Figure 6b, Figure S1b). For ease of representation, we grouped the three more widely represented landforms (fold mountains, karstic formations, and river systems) into one single category used only for plotting purposes. That highlights the presence of piedmonts exclusively in Epithelantha, and plains in the outgroup (Figure 6b). Just as we observed for soil types, the sister species E. polycephala and E. pachyriza are found in different landforms (fold mountains and piedmonts respectively) and are not polymorphic ( Figure 6, Figure S1c).

F I G U R E 6
Ancestral character state reconstruction of soil types and landforms for Epithelantha and sister genera

| D ISCUSS I ON
Our results show that Epithelantha has a preference for calcareous soils, with higher pH and drier more seasonal climate than its closely related genera. Within the genus, several species showed a preference for different environmental conditions related to temperature for sister species E. micromeris, E. cryptica, and soil type and landforms for E. polycephala and E. pachyrhiza. These environmental differences between sister species suggest differential selection pressures and local adaptation, which could have driven the speciation process (Levin, 2004;McCormack et al., 2009).
The results of our phylogenetic analyses showed that Epithelantha diverged ca. 9 Mya in the Miocene, which coincides with the estimated date of diversification of Cactaceae, and the aridification and establishment of the vegetation of the Chihuahuan desert (from mid-Miocene to late Pliocene; Arakaki et al., 2011;Hernández-Hernández et al., 2014;Scheinvar et al., 2020;Van Devender, 1990).
During this time, the deserts of North America developed and there was an increase in species richness in the region (Wiens et al., 2013).
Most of the species within the genus are younger than 2.5 My, which coincides with the onset of the climatic oscillations of the Pleistocene which had a strong effect on the Chihuahuan Desert biota (Ezcurra et al., 2020;Gámez et al., 2017;Sánchez-Escalante et al., 2005;Scheinvar et al., 2020).

| Influence of soil and bioclimatic variables on the evolution of Epithelantha
Soil conditions are among the factors that determine the distribution of species in Cactaceae. Similar patterns have been observed in other groups, for example, the ability to grow in serpentine soils is considered to be plesiomorphic for Oxera (Lamiaceae) and Guioa  (Figure 3b), which are characterized by accumulation of carbonates and a reduced layer of humus, as well as in piedmonts (Fischer et al., 2008). The pH conditions where species are commonly found go from 7.84 to 8.04, in contrast with other genera which grow in more acidic soils (Figure 4a).
These later patterns suggest that the colonization of alkaline soils could have driven the differentiation of Epithelantha from its sister genera. Four species show a strong association with haplic kastanozems: E. bokei, E. pachyrhiza, E. pulchra and E. spinosior ( Figure 3b). Only E. bokei is broadly distributed in the Chihuahuan desert, while E. pulchra and E. spinosior are found at the foothills of pine-oak forests in the Sierra Madre Oriental, and E. pachyrhiza has the smallest distribution range of the genus (Figures 1b-3d).
Close association with a specific soil type has been reported for Cephalocereus totolapensis, found in a pH range of 5.5-6-9 (Bárcenas-Argüello et al., 2010), while C. parvispinus is found in gypsum-rich soils , where the pH is more alkaline, thus suggesting that the soil type may have a similar role in the diversification of Epithelantha (Figure 4a). However, each species is associated to different soil types, and together with morphological and molecular information, there is enough evidence to recognize two species (Aquino et al., 2019). The other species that deviates from the mean precipitation observed in the genus is E. ilariae (Figure 4d), which is found in areas with higher precipitation (544 mm), in the transition between Chihuahuan desert and the Tamaulipan thorn scrub (Figure 1b).

| Environmental variables to support species delimitation
While there were several bioclimatic variables that showed phylogenetic signal and had similar values on members of Epithelantha, the variables without phylogenetic signal can be more informative to detect ecological divergence between sister species. Mean annual temperature (Bio1) and mean temperature of the coldest quarter (Bio11) show this pattern ( Figure 5) Both variables show climatic divergence of sister species E. micromeris and E. cryptica (Figure 5a,c). Aquino et al. (2019) Figure 6, Figure S1). The reports of associations between cacti and landforms are scarce. Trujillo (1984) (Huggett, 2011). These landform dynamics have been important in the evolution of the Chihuahuan desert's flora, as small changes in moisture levels in a landform can determine the presence or absence of specific species (Wondzell et al., 1996). These shifts in distribution caused by small changes in environmental conditions could help explain the diversity and co-occurrence of species within Epithelantha.
The combination of three sources of information (morphology, DNA, and environmental data) has a broad potential for species delimitation. This study confirms our hypothesis that Epithelantha is an environmentally divergent genus and that small but significant changes in soil conditions have influenced the species diversification (Donoghue & Edwards, 2014). Finally, the information derived from the understanding of ecological niche is important to inform conservation efforts. An important number of the species within Epithelantha and related genera have populations restricted to <10 km 2 (Hernández et al., 2010). From this study, E. pachyrhiza is the species with the smallest distribution and more environmental restrictions, which should be considered to protect this species natural habitat. Therefore, the next step is to construct the potential distribution of species of Epithelantha, as well as understanding biotic interactions (pollinators, seed dispersers, and pathogens), which could justify its inclusion in the Norma Oficial Mexicana (NOM-059-SEMARNAT-2010) and IUCN Red List.

ACK N OWLED G M ENTS
This study is part of the first author's dissertation and is presented as a partial requirement for the Ph.D. in Biological Sciences in the

CO N FLI C T O F I NTE R E S T
The authors have declared that no competing interests exist.