Introduction
The genus Dasylirion comprises a group of perennial, long-lived plants commonly called sotol (singular) or sotoles (plural). They belong to the family Asparagaceae and are characterized by their rosette-forming leaves and their dioecious reproductive system, with individuals bearing either staminate or pistillate flowers and obligate sexual reproduction. The species of this genus are polycarpic, thus capable of flowering multiple times throughout their lifespan. The basal stem or ‘head’ is typically large and rich in sugars, being the reason they are used for fermentation in the beverage industry. The natural distribution of the genus ranges from the southern United States through Mexico, and extends as far south as the state of Oaxaca (Reyes-Valdés et al. Reference Reyes-Valdés, Palacios, Rivas-martínez, Robledo-olivo, Antonio-Bautista, Valdés-dávila, JÁ, Benavides-mendoza, Grumezescu and Holban2019). Of the 23 currently accepted species, all are naturally found within Mexico (POWO 2026).
One notable species is D. cedrosanum, which, according to Bogler (Reference Bogler1995), constitutes a complex of waxy-leaved populations occurring in the Mexican states of Coahuila and Zacatecas, exhibiting possible hybridization with neighbouring Dasylirion species. Besides Coahuila and Zacatecas, the native range of D. cedrosanum includes the state of Durango (POWO 2026). Specimens collected in Coahuila, mainly by David Bogler and documented in the Missouri Botanical Garden database (Tropicos 2026), are found between 25°22′ and 27°00′ N latitude and 101°02′ and 102°00′ W longitude, encompassing the central to southern regions of the state.
Dasylirion plants have been utilized as a food source in North America since prehistoric times, primarily due to the high sugar content of their stems (Poinar et al. Reference Poinar, Kuch, Sobolik, Barnes, Stankiewicz, Kuder and Pääbo2001). It has been suggested that Dasylirion, along with other wild plants of the Asparagaceae family, provided most of the caloric intake for inhabitants of the current territory of Texas, approximately 10,000 years ago (Leach and Sobolik Reference Leach and Sobolik2010). In addition to their nutritional value, the leaves and floral scapes of sotol have historically been used as materials for weaving and other forms of construction (Reyes-Valdés et al. Reference Reyes-Valdés, Palacios, Rivas-martínez, Robledo-olivo, Antonio-Bautista, Valdés-dávila, JÁ, Benavides-mendoza, Grumezescu and Holban2019).
The most intensive contemporary use of several Dasylirion species is in the production of a distilled alcoholic beverage known as sotol, which has held a Denomination of Origin since 2002 for the Mexican states of Coahuila, Chihuahua and Durango (IMPI 2002). Sotol is consumed domestically and exported to various international markets, and its production process is similar to that used for tequila and mezcal. Its production may follow either artisanal or industrial (technified) methods (Reyes-Valdés et al. Reference Reyes-Valdés, Palacios, Rivas-martínez, Robledo-olivo, Antonio-Bautista, Valdés-dávila, JÁ, Benavides-mendoza, Grumezescu and Holban2019). Current annual production in Mexico is approximately 500,000 L, with an estimated growth rate of 5% per year (Barragán Aviles and Hernández Ortiz, Reference Barragán Avilés and Hernández Ortiz2024).
Unlike the production of tequila and many mezcal types, which are largely based on cultivated Agave populations, sotol production depends almost entirely on wild harvesting (Juárez-Morales et al. Reference Juárez-Morales, Martínez-salvador, Chávez-mendoza, Villarreal-guerrero, Pinedo-álvarez, Santellano-estrada, Corrales-lerma, Hernández-quiroz and Vega-mares2023). This reliance on natural populations poses a significant threat to the size and genetic diversity of the genus (Reyes-Valdés et al. Reference Reyes-Valdés, Palacios, Rivas-martínez, Robledo-olivo, Antonio-Bautista, Valdés-dávila, JÁ, Benavides-mendoza, Grumezescu and Holban2019). With the continued growth of the sotol industry, this pressure on natural populations is expected to intensify. Consequently, the implementation of Dasylirion cultivation practices is of critical importance to mitigate these ecological impacts. Establishing sustainable cultivation practices will require the development and adoption of effective planting techniques. In our experience, Dasylirion cultivation is entirely feasible; these plants have already been successfully grown in various regions of the world for ornamental purposes (Reyes-Valdés et al. Reference Reyes-Valdés, Palacios, Rivas-martínez, Robledo-olivo, Antonio-Bautista, Valdés-dávila, JÁ, Benavides-mendoza, Grumezescu and Holban2019). As we will show in the following sections, the seeds can be readily germinated, and the plants can initially be established under laboratory and greenhouse conditions.
An important yet unresolved question for sotol planting is whether selecting specific seed genotypes from wild populations can yield plant material better suited for cultivation. One of the primary objectives in the selection of Dasylirion genotypes should be to maximize growth potential under planting conditions. While other traits, such as sugar concentration, are also relevant, initial efforts should prioritize identifying individuals with high biomass accumulation rates. A central question, therefore, is whether Dasylirion populations exhibit sufficient genetic variation in early growth performance, as has been reported for Pinus pinea L. provenances, in which genetic variation enabled the selection of high-performing genotypes for growth traits (Carrasquinho and Gonçalves Reference Carrasquinho and Gonçalves2013).
While suitability for production will depend on several traits of the adult plants in field conditions, a stage that can take more than 12 years, it is of crucial importance their capacity to grow actively during their first months of life, when they should be maintained in greenhouse conditions. Before being transplanted to field conditions, they must have enough size and texture to be suitable for the environmental change (Villa Castorena et al. Reference Villa Castorena, Catalán Valencia, Inzunza Ibarra and Román López2012). In fact, the quality of the plants grown in greenhouse conditions will be a key factor for a successful field plantation. Therefore, it is important to find genotypes with an outstanding capacity for growth in greenhouse conditions during the 18- to 21-month period prior to transplantation, a time span recommended by Reyes-Valdés et al. (Reference Reyes-Valdés, Gómez-Espejo and Antonio-Bautista2026).
The central hypothesis of this research is that genetic variation for early plant growth dynamics can be found in natural populations of sotol, making it possible to select outstanding genotypes for the greenhouse stage. The general aim is to contribute to a transition from wild plant utilization towards planting projects with productivity objectives, allowing the sotol production processes to become more sustainable by preserving natural genetic variation and population sizes.
The general approach is to evaluate stem growth in Dasylirion half-sib families, which are easily obtained due to its dioecious reproductive system. Seeds collected from a single floral scape form a half-sib family, sharing a common maternal parent while being pollinated by multiple male individuals. Within this framework, the specific objective of the present study is to investigate genetic variation for early stem growth in sotol by evaluating a sample of half-sib families.
Materials and methods
Plant materials
Seed collection sites were located within the documented range of D. cedrosanum, as recorded in the Missouri Botanical Garden database (Tropicos 2026). Although no other species has been reported in this area, population mixture with other sotol species cannot be discarded, given migration and phenotypic resemblances. The seed samples used in this study originated from northeastern Mexico (Fig. 1a), specifically around 26°11′ N latitude and 101°39′ W longitude, and were collected from 2012 to 2022. This region of Coahuila lies within the Chihuahuan Desert and the physiographic province of the Sierra Madre Oriental. The climate ranges from warm semi-arid (BSh) to warm desert (BWh), according to the Köppen–Geiger classification (Kottek et al. Reference Kottek, Grieser, Beck, Rudolf and Rubel2006). A more detailed view of the maternal plant distribution is shown in Fig. 1b, focusing on the state of Coahuila. Collection sites span from the central-southern to the southeastern part of the state.
Geographic origin of seed sources for 30 half-sib families and a control population of Dasylirion Zucc. (a) Distribution across Mexico. (b) Distribution within the state of Coahuila. Some sampling points for the half-sib families are geographically close and may appear overlapping on the map.
The individual maternal plants of 30 half-sib families were sampled within the municipalities of Saltillo, Ramos Arizpe and Parras, all of them from southeastern Coahuila. The control material, RP1, was collected from a distinct northern location within the municipality of Cuatro Ciénegas. The source site of this control population is well centred in the area of distribution of D. cedrosanum, our seed stock is large enough to be used in several research projects, and it is being utilized as a reference for ongoing work on plant growth, nutrition and tissue culture. All collection sites fall within the documented natural distribution range of D. cedrosanum. In addition, we previously reported the matK sequences of two specimens: one on the southernmost point of our sampling area (Ortiz-Covarrubias et al. Reference Ortiz-Covarrubias, Orozco-sifuentes, Mendoza-Rodríguez, Villlarreal-quintanilla, Martínez, Hernández-godínez and Reyes-Valdés2022) (GenBank accession MZ355826), and the other one in the northernmost point, corresponding to the area of our control material RP1 (GenBank accession PX897973). In both cases, the matK DNA sequence corresponds to D. cedrosanum.
In all cases, floral scapes were manually harvested, and seeds were cleaned and stored at 4°C until use. Prior to germination testing, seeds were surface-sterilized by immersion in 70% ethanol for 1 minute, followed by 2% sodium hypochlorite for 15 minutes, and rinsed with distilled water for 1 minute. Germination tests were conducted in Petri dishes to ensure a minimum germination rate of 50% for all seed lots.
Sowing and nursery cultivation
Following surface disinfection as previously described, seeds were sown in 180-cavity trays filled with a substrate composed of peat moss and vermiculite in a 3:1 ratio (v/v). Trays were placed in an environmental chamber (Lab-Line Biotronete) maintained at a temperature range of 24–26°C under a 12-hour light/12-hour dark photoperiod.
Once seedlings reached approximately 10 cm in height, which happens around 45 days from sowing, they were individually transplanted into 1.2-L plastic nursery bags and transferred to a greenhouse maintained at an average temperature of 24°C, equipped with air extractors and evaporative cooling pads. The 31 materials (30 half-sib families and the control population) were randomly distributed within the greenhouse, with 21 plants per material (Fig. 2). This design constituted a completely randomized layout, with the treatments corresponding to the different materials (i.e. genotypes).
Young Dasylirion plants growing under controlled greenhouse conditions.
Plants were irrigated three times per week and fertilized twice during the two-year experimental period using Nitro Tabs®, which contain a balanced mix of nitrogen, phosphorus, potassium, sulphur, iron and zinc. Each plant was assigned a unique identification code for individual tracking and data collection.
Data collection
Stem diameter was measured in individual plants eight times over the course of approximately 21 months. This trait was measured in millimetres (mm) using a digital calliper. Because young stems are typically ovoid in shape, two perpendicular measurements were taken for each plant: one along the major axis and one along the minor axis. The average of these two values was used as the stem diameter for each plant at each time point.
Data analysis
All data management and analyses were conducted using the R language and environment for statistical computing (R Core Team 2024). Data from 522 plants were used for the analyses, with an average of 16.84 plants per genotype. The general growth trend was explored using scatter plots generated with the ggplot2 package. During the 21-month period, growth appeared to be approximately linear, and thus a linear modelling approach was adopted to analyse temporal stem diameter trends.
A longitudinal analysis based on a mixed model was used to gain knowledge about general trend and genotype variation in growth rates with 3,460 data points. The mixed model was fit through the lmer function of the R package lme4 (Bates et al. Reference Bates, Mächler, Bolker and Walker2015), aided by the lmerTest package (Kuznetsova et al. Reference Kuznetsova, Brockhoff and Christensen2017). A random-slope mixed model was fitted, with the intercept constrained to zero because stem growth was assumed to start at germination with a diameter of zero. The model included time as a fixed global slope through the origin, a random slope for genotypes, and a random slope for plants nested within genotypes. The mathematical expression of the model was:
\begin{equation*}{y_{ijk}} = {{\beta}} {t_k} + {u_i}{t_k} + {v_{ij}}{t_k} + {\epsilon_{ijk}},\end{equation*}where yijk is the stem diameter in plant j from genotype i in time k, tk is the growth time, β is the population average growth rate (fixed effect), ui is the random deviation from growth rate in genotype i, vij is the random deviation in growth rate for plant j within genotype i and ϵijk is the residual error.
The model was fit for two situations: the 30 full-sib families plus control genotype, and only the 30 full-sib families. A fixed global slope was tested through Satterthwaite’s method with Type II analysis of variance, implemented in the lmerTest package. A likelihood ratio test against the null model was used to reveal the statistical significance of the variance among genotypes for growth slopes.
Predicted values for stem growth rate across genotypes, along with their conditional variance, were obtained from the respective mixed model. Generalized heritability was estimated through the Cullis method (Cullis et al. Reference Cullis, Smith and Coombes2006). This estimator is based on the expression H 2 = 1 – Att/(2Vg), where Att is the average pairwise prediction error variance, and Vg is the genetic variance. The function h_cullis from the agriutilities R package (Aparicio and Bornhorst Reference Aparicio and Bornhorst2026) was used to perform calculations.
Since the distribution of the estimated heritability does not have a known statistical shape, we relied on a parametric bootstrap with 200 simulations and the aid of the mertools R package to estimate its confidence interval. Simulations allowed us to get an empirical distribution and extract the quantiles for 95% confidence intervals.
Results
Growth trend
The growth trend through the 21-month period showed an approximate linear fit, with an overall growth rate of 0.67 mm per month. Furthermore, growth rate variation among the 31 genotypes can be graphically appreciated. Figure 3 presents scatter plots of stem diameter growth across the different half-sib families and the control population, with their least squares fitted lines. Time is expressed in months based on the average Gregorian calendar month (30.437 days). Although the linear trend fits well for all genotypes, slope differences can be perceived. In fact, the control material RP1 shows a remarkably low growth rate, along with some families like SM140. In contrast, some families such as UAA2002 and PG02 depict high growth rates. In most cases, variation among plants within genotypes can be readily observed, especially at the final stages of the 21-month growth period.
Stem growth trends across a 21-month period and their fitted lines, for 30 half-sib families of Dasylirion and a control population (RP1).
Figure 3 Long description
Scatter plots display stem diameter growth in millimeters over a 21-month period for 30 half-sib families and a control group labeled RP1. Each plot has time in months on the horizontal axis, ranging from 0 to 21 and stem diameter in millimeters on the vertical axis, ranging from 0 to 30. The plots include least squares fitted lines indicating growth trends. Overall, the plots show a general increase in stem diameter over time. The control group RP1 exhibits a lower growth rate compared to other families. Some families, such as UAA2002 and PG02, show steeper growth trends, while others like SM140 have flatter trends. The data points are marked with symbols and the fitted lines are drawn through these points to indicate trends. Each subplot is labeled with identifiers such as CA02, CA03 and so on, up to UA2007. The fitted lines generally show a positive trend, with variations in slope indicating differences in growth rates among the genotypes. No specific outliers are highlighted, but variations in growth patterns are evident across the different genotypes.
Longitudinal analysis results
The fitted models had a reasonable fulfilment of the assumptions for residual normality and homoscedasticity, through QQ normal and Tukey-Anscombe plots. The fixed effect of the model with all genotypes, which corresponds to the global slope, was estimated at 0.674 mm per month and was highly significant with p < 2.2 × 10−16. This was an expected result, given the observed linear trend for stem growth. For the random effects, the likelihood ratio tests revealed highly significant variation among genotypes. Without excluding the control population RP1, the p-value for the genotype random slope was 2.183 × 10−9, whereas including only the experimental half-sibs led to a p-value of 1.419 × 10−8.
The estimate of Cullis heritability for the whole set of genotypes was 0.725, with 95% bootstrap confidence limits of 0.533 and 0.826. Including only the 30 half-sibs led to a heritability of 0.701, with 95% confidence limits of 0.424 and 0.815.
Predicted effect results
In Table 1, the best linear unbiased predictions of stem growth rate for each genotype are presented along with the conditional standard deviations of the respective random effects. The family with the highest predicted stem growth rate was UA2002, from the municipality of Saltillo, with a value of 0.81 mm/month, followed by PG07 from the same municipality, with a prediction of 0.77 mm/month. The genotype with the lowest growth rate was half-sib family GC32, from the municipality of Parras, Coahuila, with a predicted value of 0.53 mm/month, followed by the control population RP1, with a prediction of 0.57 mm/month. The ratio of the highest (UA2002) to the lowest (GC32) predicted slopes was 1.53. Compared to the control population, the ratio is 1.42. Thus, we would expect a superiority of more than 40% in early stem growth of the best family, compared to the control population.
Ordered best linear unbiased predictions (BLUPs) of rates for monthly stem diameter growth (mm/month) for 31 Dasylirion genotypes, along with the conditional standard deviations (SD) of the respective random effects. The growth period was 21 months in greenhouse conditions
Table 1 Long description
Ordered BLUP estimates rank 31 Dasylirion genotypes by predicted monthly stem diameter growth in millimeters per month, with a conditional standard deviation for each estimate. Predicted growth spans from 0.81 for UA2002 (SD 0.03) down to 0.53 for GC32 (SD 0.03). Other high performers include PG07 at 0.77 (SD 0.04) and UA1103 at 0.75 (SD 0.03), while low performers include RP1 at 0.57 (SD 0.04). Many genotypes cluster tightly between about 0.63 and 0.66, including CA03, UA2007, CA02, CA05, and PG05, indicating small differences among mid-ranked entries. Conditional SD values are consistently small, mostly 0.03 or 0.04, suggesting similar uncertainty across genotypes. Because these are model-based predictions, small rank differences, especially within the mid-range cluster, may not reflect meaningful biological separation.
Discussion
The collection sites for the half-sib families are located within a well-defined region of southeastern Coahuila, an area previously reported within the natural distribution of D. cedrosanum (Tropicos 2026). However, mixture or hybridization with other local species cannot be discarded (Bogler Reference Bogler1995). Given the potential for long-distance pollination in the genus, either by wind or by insect pollinators (Reyes-Valdés et al. Reference Reyes-Valdés, Palacios, Rivas-martínez, Robledo-olivo, Antonio-Bautista, Valdés-dávila, JÁ, Benavides-mendoza, Grumezescu and Holban2019), populations in the sampled area are expected to experience ongoing gene flow.
The control population (RP1), although belonging to the same species reported for southwest Coahuila (D. cedrosanum), is likely genetically differentiated from the base population due to geographic isolation. The straight-line distance between the RP1 site and the geographic centre of the half-sib collection area is approximately 120 km, a distance sufficient to limit natural gene flow and promote genetic divergence.
The general pattern of stem growth during the 21-month period was linear (Fig. 3). Variation was also observed among plants within genotypes. This finding is consistent with quantitative genetics theory. In fact, for a quantitative trait, the variance among half-sib families represents one-quarter of the additive genetic variance of the population, whereas the variance among individuals within half-sib families comprises the remaining three-quarters of the additive genetic variance, together with the non-additive genetic variance (Hallauer et al. Reference Hallauer, Carena and Miranda Filho1988). Consequently, substantial genetic variation is expected among plants within families, in addition to random microenvironmental variation, even under controlled greenhouse conditions.
The linear early growth trend observed during this research agrees with the observations of Villa Castorena et al. (Reference Villa Castorena, Catalán Valencia, Inzunza Ibarra and Román López2012) for plant height and foliar area in Dasylirion. Thus, linear growth is expected during the recommended period of 18–21 months in the greenhouse, for sotol growth before transplantation to the open field (Reyes-Valdés et al. Reference Reyes-Valdés, Gómez-Espejo and Antonio-Bautista2026). For the early growing stages, greenhouse conditions are necessary until the plant reaches the size and texture to adapt to field conditions, where it will face several challenges, including the attack of herbivores. That is why early growth in greenhouse conditions is so important, even though early growth may not be strictly correlated with late growth.
The general linear growth tendency through the 21-month growth period for stem diameter in sotol had a global slope of 0.67 mm per month. The highly significant variation among genotypes for early stem growth, both in the whole set and the one comprising only the sample of half-sibs without the control population RP1, is encouraging. It suggests the potential for selecting promising Dasylirion genotypes with strong early growth performance under greenhouse conditions – a head-start advantage for growing sotol plants, and one that may persist under field conditions.
The results suggest that different regions within the natural distribution range of Dasylirion in southeastern Coahuila harbour genotypes with varying growth capabilities, potentially due to historical genetic drift or local environmental adaptation. Notably, the control population RP1 exhibited one of the lowest predicted growth rates.
Heritability is a critical component in predicting response to selection in both plant and animal breeding (Hallauer et al. Reference Hallauer, Carena and Miranda Filho1988; Falconer Reference Falconer1989). Therefore, it plays a decisive role in assessing the feasibility of developing sotol cultivars with enhanced early growth rates, particularly for stem diameter, the most relevant trait for sotol production. The estimated heritability was 0.725 when all genotypes were included in the analysis and 0.701 when only the 30 half-sib families were considered. These values indicate a high reliability of genotype evaluation and suggest that selection for increased stem diameter growth rate should be effective in this population. Broadly speaking, heritability values exceeding 0.5 are generally considered favourable for selection (Burdon Reference Burdon and Burley2004).
The slight reduction in heritability after removing RP1 may be attributable to its extreme growth rate. To further investigate this effect, we randomly selected three additional genotypes and removed them one at a time to evaluate the resulting changes in heritability. Removing TN04, a genotype with an intermediate growth rate, yielded an updated heritability estimate of 0.734, representing a slight increase. Removing GC12, a family with a relatively low but not extreme growth rate, resulted in an estimate of 0.726, indicating virtually no change. In contrast, removing UA2002, the family with the highest growth rate, reduced the heritability estimate to 0.684.
The results of our analyses offer a clear opportunity to select or breed sotol genotypes with superior early growth. Dynamic early growth in greenhouse conditions will give a head start for their future development in field conditions. A practical approach, considering the dioecious reproductive system of Dasylirion, is half-sib family selection based on seeds collected from individual maternal plants. Since a single female sotol plant can produce approximately 100,000 seeds (Sierra-Tristán et al. Reference Sierra-Tristán, Lara Macías, Carrillo-romo, Melgoza-castillo, Morales-nieto and Royo-vázquez2008), this strategy allows for the evaluation of family performance while retaining sufficient seed stock for further use. Subsequent propagation of superior genotypes could be achieved either through controlled intercrossing among selected half-sibs or by employing in vitro propagation techniques.
A practical breeding strategy could involve further cultivating the families with the highest predicted values (Table 1) under field conditions to assess the consistency of their high growth rates. After several years, seeds from the top-performing families could be harvested to establish a selected planting population. An alternative strategy would be to relocate and re-harvest seeds from the original maternal plants.
Dasylirion plants are commonly planted ornamentally in various towns across Mexico and the southern United States, suggesting early stages of domestication, a pattern observed in many native plant species (Figueredo et al. Reference Figueredo, Casas, González-rodríguez, Nassar, Colunga-garcíamarín and Rocha-ramírez2015). The genus also possesses traits associated with ‘domesticability’, including notable genetic plasticity, a key feature proposed for plants amenable to domestication (Romero et al. Reference Romero, Kolesnikova, Ezard, Charles, Gutaker, Osborne and Chapman2025). Furthermore, Dasylirion is easily propagated under cultivation; in this study, seeds germinated successfully without any pretreatment. Our results, which show significant genetic variation and heritability for early vegetative growth, suggest that Dasylirion populations from southeast Coahuila have the genetic potential to be selectively bred for increased early growth rates. This adds to the evidence supporting its potential domestication for plantation-based production. Such a strategy could reduce pressure on wild populations and contribute to the conservation of this valuable plant resource.
Finally, it is important to acknowledge a major limitation of this study: there is currently no evidence that families exhibiting superior early growth under greenhouse conditions will maintain their growth advantage under field conditions. Nevertheless, vigorous plant development during the greenhouse stage is expected to increase the likelihood of successful field establishment. A comparative field evaluation of growth performance will require more than eight years to complete and is the focus of our ongoing research.
Acknowledgements
We thank Secretaría de Ciencia, Humanidades, Tecnología e Innovación (Secihti) of Mexico for granting a graduate scholarship to G.T, Velazquez-Cruz and for V.G. Vizcarra-Manriquez. Artificial intelligence assistance was provided through the ChatGPT platform (GPT-3.5), which was used solely to refine some individual paragraphs and support R code development for final plot modifications and refining longitudinal models.
Funding statement
This research was funded by Universidad Autónoma Agraria Antonio Narro.
Competing interests
The authors have no relevant financial or non-financial interests to disclose.
