Solanum is an economically important plant genus that includes the cultivated potato (S. tuberosum L.). It has served as the model-system for understanding the genetic basis of tuberization in plants, but the genus includes distinct types of underground organs that can be broadly classified into two functional groups: rhizomatous species that focus on lateral spread, and species with USOs focused on starch/water storage. In Solanum, USOs are organs that are conspicuously specialized for storing complex carbohydrates and/or water, and include stem tubers in the wild potatoes in Solanum section Petota (Spooner et al., 2016; Spooner et al., 2019; Figure 1D), root tubers in the Asterophorum (Gouvêa and Stehmann, 2019) and Carolinense clades (Wahlert et al., 2015; Figure 1E), and swollen hypocotyls in the Regmandra clade (Bennett, 2008; Figure 1F). The principal function of these USOs is the storage of different sugars, nutrients and water needed by the plant to undergo dormancy and/or to allow the plant to recover from disturbance.

Rhizomatous species in Solanum, in contrast, have belowground stems that enable lateral spread, from which both above ground stems and roots arise (Serebrjakov and Serebrjakova 1965; CLO-PLA database trait definition (Klimešová et al., 2017); Figures 1A, B). Our definition of “rhizomatous” also includes species known as “root sprouters” or “root suckers”; the latter have horizontal roots with adventitious buds from which a whole plant is able to resprout (Miyazaki and Ito, 2004; Pattison et al., 2019), and which are sometimes also referred to as species with “gemmiferous roots” or with “endogenous buds” (Wahlert et al., 2015; Figure 1C). Distinguishing between species that have true rhizomes and those that are root sprouters is particularly difficult from the literature as these structures are generally not clearly or adequately described (plants are sometimes merely referred to as “clonal”), and difficult to observe on herbarium specimens. In both cases, however, these underground horizontal structures in Solanum facilitate lateral and clonal dispersal, as well as enabling plants to resprout from disturbance events.

A list of all 1,232 currently accepted species for the genus Solanum was obtained from the expertly curated database Solanaceae Source (SolanaceaeSource.org, November 2020). Trait data were derived from species descriptions in taxonomic monographs and from herbarium specimens. All species were categorized as either non-geophytic (931 species, 75% of the total), rhizomatous (180 species, 15% total), or with USOs (121 species, 10% of the total; Supplementary Table S1; Supplementary Datasheet S1). A total of 301 Solanum species (24% of the total) were found to be geophytic.

In Solanum some species possess both underground storage organs and rhizomatous structures. This includes some species of the Carolinense clade, root sprouters that sometimes also possess root tubers (Wahlert et al., 2015). Nearly all members of the Petota clade have underground rhizomes, which are frequently referred to as “stolons,” a term which in the functional root trait literature usually refers to above-ground lateral stems (Klimešová et al., 2019) but which has been applied to underground rhizomes connecting tubers in the past (Bell and Tomlinson, 1980). Rhizomes in tuberous Solanum species show considerable variation in length, varying from moniliform with tubers formed in a string along a short rhizome to species with >1 m long rhizomes with few large tubers at the end (Spooner et al., 2004). Here, we treat all these tuberous Solanum species with rhizomes under the category of USOs and consider them distinct from rhizomatous species because the latter do not store large amounts of carbohydrates or water. Some species considered rhizomatous have been shown to have carbohydrate reserves, such as the “tuberized” rhizomes with thickened cortical walls found in S. elaeagnifolium and its relatives (Knapp et al., 2017), but they represent a very different system of carbohydrate storage compared to the localized and concentrated starch storage found in USOs such as the stem tubers of potatoes. The categories we use here for USOs and rhizomes may differ from how these terms may be applied in other groups (e.g., Chomicki, 2013) but allow us to explore our questions in Solanum.

To account for the phylogenetic relatedness between the study species, we used a recently published species-level phylogeny that included 742 Solanum species (60% of those currently accepted; (Gagnon et al., 2022). Cultivated species (n = 19) where wild populations are not known or clearly labelled as cultivated were excluded because their distributions reflect anthropogenic commensalism rather than evolutionary patterns related to environment (S. tuberosum, S. lycopersicum, S. melongena, S. muricatum, S. aethiopicum, S. macrocarpon, S. lasiocarpum, S. betaceum, S. sessiliflorum, S. quitoense, S. scabrum, S. aviculare, S. crispum, S. laciniatum, S. laxum, S. pseudocapsicum, S. seaforthianum, S. wendlandii, S. mammosum). Species used by humans, but which are also known to occur in the wild and where wild populations can clearly be distinguished were included (e.g., S. torvum), but we took care to exclude all cultivated records.

We downloaded taxonomically verified occurrence data for all input species from specimen records in SolanaceaeSource (accessed 21st of April 2021; 115,496 occurrence records) and the Australian Virtual Herbarium (29,305 occurrence points; 5th June 2019, https://doi.org/10.26197/5cf786115b9ef). Occurrence data were cleaned using several steps implemented in R (v. 4.1.0 (R Core Team, 2021), to remove specimens 1) unidentified to species, or that were considered as ornamental or crop species (see section Phylogeny below); 2) indicated as cultivated in their label data; 3) lacking latitude/longitude data; 4) with erroneous coordinates (seconds greater than 59 in value); 5) in the sea or large bodies of water; 6) with imprecise coordinates (seconds missing for both latitude and longitude); 7) where coordinates did not match with the country given in label data. In addition, the R package “CoordinateCleaner” (Zizka et al., 2019) was used to remove duplicate specimens, specimens with equal lat/long coordinates, specimens with coordinates corresponding to zero latitude or longitude, as well as specimens that fall within 1 km of a list of c. 10,000 biodiversity institutions. Maps were also generated and visualized to ensure that distributions corresponded to known distributions described in the taxonomic literature, removing aberrant points. This resulted in a dataset with 80,525 occurrence records for 1,169 species (94.8% of all Solanum species); 217 out of 224 geophytes were kept in this dataset (98%, see Supplementary Table S1). Average and median number of occurrence records for each geophytic category is summarized in Supplementary Table S1, and 77% of all the retained species in Solanum had >5 occurrence records. Spatial filtering was done for all species with >5 records to reduce collection bias and outlier effects in the dataset, where we ensured that there was at least 10 km distance in between occurrence records within the same species.

We selected eight environmental variables with a strong impact on the global distribution of geophytic plants with focus on heat, frost, temperature seasonality, precipitation seasonality, drought, fire, soil sand content, and topography (Prentice et al., 1992). Heat was represented by maximum temperatures of warmest month (bio5), minimum temperatures of coldest month as a proxy for cold stress (bio6), temperature seasonality by temperature annual range (bio7), precipitation seasonality by coefficient of variation (bio15), and drought by the annual Moisture Index (MI; the ratio of annual precipitation to annual potential evapotranspiration). All temperature and precipitation variables were derived from CHELSA climate data at a 30 arcsec spatial resolution (c. 1 km). MI was calculated using the Priestly-Taylor formulation provided by the SPLASH algorithm of (Davis et al., 2017), because it has been shown to perform substantially better in reproducing eddy-flux measurements of actual evapotranspiration under unstressed conditions compared to Penman-Monteith formula (Maes et al., 2018). MI values lower than one indicate there is less precipitation than evapotranspiration, resulting in drought stress for plants thereby providing more information than values solely based on precipitation.

Fire was included as an environmental variable by using the 95th quantile of fire size (q95size), based on a global map calculated by Archibald et al. (2013), at a 0.5° spatial resolution (∼55 km). Fire size is one of the five key characteristics of fire regimes. Inherent challenges of calculating environmental variables related to fires that occurred more than 50 years ago resulted in a large amount of missing data in the q95size layer, so we used custom R scripts to attribute the value of the closest cell if the occurrence point was less than 55 km away from a neighboring cell. Variation in soil density was included by including the proportion of sand particles (>0.05 mm) in the fine earth fraction at a depth of between 0–5 cm from SoilGrids 2.0, at 30 arcsec resolution (Poggio et al., 2021). Soils with higher sand content are hypothesized to allow easier growth and expansion underground for rhizomes (Herben and Klimešová, 2020). Finally, topographic complexity was measured using the vector ruggedness metric (VRM) at 30 arcsec resolution (Amatulli et al., 2018), to test whether USOs and rhizomatous species show differential responses to complex mountainous terrain, where rhizomes would be expected to dominate in flatter terrains with looser soils where underground lateral expansion might be easier. VRM varies from 0 for flat surfaces, to 1 for the most rugged regions (Amatulli et al., 2018).

For the ecological analyses, we only retained occurrence records for which values of all environmental layers were available, resulting in a final dataset with 47,083 records, representing 1,151 species of Solanum (Supplementary Table S1). This dataset was used in the PCA, PERMANOVA, Phylogenetic ANOVAs, and for the calculation of niche breadth.

We calculated range sizes using the package “ConR” (Dauby et al., 2017) with alpha hull values (α = 2) for spatially non-filtered occurrence data based on Baldaszti (2021). Range sizes could not be calculated for 116 species that had fewer than 2 occurrence records; these were removed from all analyses. Our final dataset for range size and niche breadth included 1,062 species of Solanum with 782 non-geophytic and 280 geophytic species (Supplementary Table S1). Niche breadth was calculated using spatially filtered dataset as the sum of the scaled difference of the maximum and minimum value of all eight environmental variables. For species with >10 occurrence records, a subsampling process was introduced to reduce the impact of climatic outliers following Baldaszti (2021).

Prior to analyses, both range size and niche breadth were log-transformed to meet assumptions of normality required for post hoc analyses. To determine whether mean range sizes and niche breadth were significantly different among the three groups (e.g., non-geophytes, species with USOs, and rhizomatous species), we carried out a non-parametric Kruskal–Wallis test using the package “rstatix” (Kassambara, 2021). In addition, post hoc pairwise comparisons between the three groups were carried out to determine which pairs were significantly different from one another, using the Dunn test with Bonferroni adjustment.

We examined the correlation between climatic niche breadth (within the individual climatic variables and overall) and range size using a linear model with log-transformed values for both range size and niche breadth. The “smart” package (Warton et al., 2012) was used to determine whether the relationship between range size and niche breadth across the three growth form groups (USOs, rhizomes and non-geophytes) were similar. Statistical tests from the “Standardized Major Axis Tests and Routines” (SMATR) allowed us to determine whether there were any significant differences in the slope, as well as shifts in elevation or along the x-axis in the linear models calculated for each growth form separately. We also used robust T and adjusted p-values for multiple comparison tests. We calculated 95% confidence intervals for each slope by carrying out 1,000 bootstrap replicates using a custom R script.

Ancestral trait reconstruction analysis was run using stochastic character mapping (SIMMAP) to examine the evolutionary lability of the underground organs across the phylogeny of Solanum using the “phytools” package (Revell, 2012). We treated underground organ trait data as unordered and equally weighted for all three categories. A likelihood ratio test with the Akaike Information Criterion was used to identify the best fitting transition rate model: equal transition rate among characters states (ER), all rates different model (ARD), and symmetrical rate model (SYM). Based on the likelihood ratio test and the AICc criterion, we found the ARD model had the best fit. A total of 200 simulations were run to obtain a posterior probability distribution of ancestral states across the phylogeny with 702 tips kept for which environmental data were available.

Significant differences between the means of range sizes were detected with the Kruskal–Wallis test, with rhizomatous species having larger range sizes than all other species (Table 1; Supplementary Figure S3). While the SMATr analysis was not significant for the shift test, it did identify significant differences in the slope and in the elevation of the linear models of the three groups, with rhizomatous species having a slightly stronger correlation between range size and niche breadth than non-geophytic and USO species (Figure 4; Table 1). Finally, no significant differences in niche breadth were found between the three groups (Table 1; Supplementary Figure S3).

Ancestral trait reconstruction analyses indicated distinct patterns of underground organ evolution across Solanum, with frequent origins of rhizomatous species and only occasional origins of USOs (Figure 5). Rhizomatous species evolved on average 28 times, and reversals back to non-geophytic growth forms were observed on average 26 times (Figure 5). In contrast, USOs evolved on average 5 times across the entire phylogeny of Solanum, with no reversals back to a non-geophytic growth form (Figure 5). No transitions between rhizomatous and USO species were observed (Figure 5).

Our results confirm previous findings (Howard et al., 2019) that the evolution of geophytic growth forms is strongly associated with temperature seasonality, but we also find association with drought (Figure 2; Supplementary Figure S2) with significant differences seen between geophytic and non-geophytic species (Tables 1, 2; Figure 2). Geophytic Solanum species generally occur in environments that are more prone to drought and seasonal disturbance, as well as with more extreme minimum and maximum annual temperature (Figure 2), in comparison with their non-geophytic congeners (Figures 2E–H). This indicates that geophytic growth forms may act as potential eco-evolutionary enablers for lineages to enter environments dominated by harsh conditions, such as drought and temperature seasonality.

One of the most novel and interesting results in our study is the apparently clear environmental differences between the two distinct types of geophytic organs in Solanum. Rhizomatous species are strongly associated with environments that experience severe but less seasonal drought and larger fires. Fires are hugely influential in shaping ecosystem assembly and functional diversity (Pausas et al., 2018), with plants adopting various strategies to overcome, survive and persist different fire regimes, including the ability to resprout from underground bud banks (Pausas et al., 2016). The interplay between drought and fire strongly influences the frequency of resprouting species in the Cape Floristic region, where prolonged drought periods combined with fire favor resprouters (Ojeda, 1998). The presence of both drought and fire may also explain the number of rhizomatous Solanum species in Australia (Figure 2). In addition to fire, sandy soils and low topographic complexity are associated with rhizomatous species in Solanum (Table 2; Figure 4), likely due to poorly consolidated soils aiding lateral expansion.

In contrast, our results indicate that the evolution of USOs is associated with environments with less severe but more prolonged drought and cold stress (but not frost) and in which fire are not as large or severe. USOs in Solanum are mostly found in the large Petota clade and the environmental signal of USOs is hence driven mostly by the strong phylogenetic aggregation in our data. This can be seen in the distribution of species with USOs in American montane regions (Table 2; Figures 2, 4) reflecting species diversity patterns of the Petota clade whose species are concentrated in South America in regions like the Andes, or the Sierra Madre Mountain range in Mexico (Spooner et al., 2004; Spooner et al., 2016; Spooner et al., 2019). The link between USOs and tropical montane regions follows our expectations that USO species have evolved to persist in environments with intermediate levels of disturbance and/or limitations in growth and biomass production. Plant growth in tropical montane regions is not limited by strong seasonal changes in light availability and temperature as it is in high latitude temperate areas, but rather by water availability (including precipitation seasonality, bio15 in Table 2), that can occur due to shifts in elevation or strong rain-shadow effects. Further exploration of functional and morphological diversity of USOs in Solanum, especially in smaller clades presenting root tubers and swollen caudices may reveal additional patterns. The swollen caudices found in the Regmandra clade (here defined as USOs) occur in species occurring in dry seasonal environments in the coastal fog-deserts of western coasts of South America (Bennett, 2008), and root tubers are found in members of the Carolinense clade from the Chaco dry woodlands that experience frost (Wahlert et al., 2015). Root tubers in the Asterophorum clade, however, are found in species occurring in damper soil conditions in the Brazilian Atlantic forests (Gouvêa and Stehmann, 2019) in forest edges prone to disturbance.

Another seldom-explored aspect that might influence the distribution of geophytes is the pressure of herbivory (but see Meller et al., 2022). Placements of buds belowground protects meristems and allows plants to resprout after a herbivory event (Ott et al., 2019) involving, for example, mammalian grazing or insect attack on new shoots. In Solanum, defences against herbivory can be both mechanical (prickles and stellate hairs) and chemical (alkaloids and terpenoids). These may prove deterrents, but mammals are known to graze on prickly solanums (e.g., Pringle et al., 2014) and specialist herbivores are known to overcome chemical defences (e.g., Brown, 1987). Rhizomatous species of Solanum are especially common in the Eastern Hemisphere Spiny (EHS) clade, a group with abundant prickles on leaves and stems (Symon, 1981; Bean, 2004), and studies with the rhizomatous S. carolinense of the Carolinense clade (Nihranz et al., 2019) have demonstrated the importance of belowground organs in storing information about environmental stressors such as herbivory. Induced defenses transmitted through rhizomes affect offshoot growth and herbivore resistance, but are compromised by inbreeding (Nihranz et al., 2019), suggesting complex inter-seasonal patterns of plant-animal interaction in at least this rhizomatous Solanum species. Leaves in many members of the Petota clade that all have USOs contain high levels of steroidal alkaloids that are likely to play a role in plant defence (Spooner et al., 2004; Spooner et al., 2016; Spooner et al., 2019), perhaps through activation of chemical cascades as has been shown in Nicotiana (e.g., Kessler et al., 2010).

The question remains whether underground organs enable niche shifts or whether they expand niches by enabling species to persist and reproduce across a wider range of environmental conditions. Our results show that rhizomatous species have significantly larger mean range sizes (Table 1) than do either non-geophytes or species possessing USOs as defined here, a pattern similar to monocots (Howard et al., 2019). There is also a stronger relationship between range size and niche breadth in rhizomatous species (Figure 4). The increase in range size is not, however, related to increase in overall niche breadth when compared to species with USOs and non-geophytes (Table 1). This pattern could be explained by the fact that rhizomatous species are better able to disperse laterally by vegetative means leading to larger range sizes, but their ecological niche breath is no wider than that of non-rhizomatous species, suggesting that range size is strongly constrained by ecological niche but less so by dispersal.

Niche shifts in geophytes could be facilitated by expansion of niche breadth through exaptation (Gould and Vrba, 1982; Donoghue, 1989). Escape from recurrent episodes of aboveground disturbances such as fire and herbivory through development of underground buds could allow such species to expand into a different set of environments, with different stressors. Fire resistance is thought to have led to the evolution of geoxyle trees in African savannas and thus allowed them to later persist in and invade more frost-prone environments (Davies et al., 2017; Lamont et al., 2017).

We show striking differences in evolutionary lability of underground growth forms in Solanum. For example, the rhizomatous habit evolved independently more than 27 times across Solanum with the 25 of these occurring in the EHS clade where plants are predominantly root-sprouters. The lability is most apparent when seen in contrast to the conserved nature of USOs, which have evolved in only five independent lineages, with no reversals to a non-geophytic growth form (Figure 5). The lability in the rhizomatous habit has been observed before in other studies across monocots (Howard et al., 2019) and in temperate herbs in eastern Europe (Herben and Klimešová, 2020), whereas USOs are often confined to particular lineages (Howard et al., 2019).

The conserved nature of species with USOs suggests that these organs are difficult to evolve, possibly due to complex set of genomic and molecular processes required for the development of storage tissues in underground environments. Anatomical and molecular mechanisms regulating storage root formation in plants have been particularly well-studied in crop species, including in potatoes [reviewed by Chen and Tang (2017)], and have shown that tuber formation is complex. In potatoes, the key genes involved in tuber initiation have been shown to have evolved as a result of expansion and neo-functionalisation of FLOWERING LOCUS T (FT) proteins, also involved in the initiation of flowering (Abelenda et al., 2011; Navarro et al., 2015). The involvement of FT genes in initiation and formation of USOs have also been found in other distantly related geophytes, including in the Nymphaeales, monocot, and eudicot lineages (reviewed by Tribble et al., 2021a). Comparative transcriptomic studies have also identified other genes that enable USO formation, by allowing the plant to increase storage of water and carbohydrates (e.g., starch biosynthesis, lignin biosynthesis, cell wall modifications; Hearn et al., 2018; Tribble et al., 2021a). In comparison, much less is known about the genetic control of underground rhizomes or the development of adventitious buds on rhizomes. QTL studies in Sorghum have shown that rhizomatous phenotypes seem to be controlled by a few genes (Paterson et al., 1995; Yim and Bayer, 1997; Washburn et al., 2013; Hu et al., 2011). There is an increasing interest in the genetic control of rhizomes due to their role in the establishment of perenniality and contributions to weediness in crops (Hu et al., 2011). These factors are also of considerable interest in development of perennial crops from annual progenitors (Kantar et al., 2016; Ciotir et al., 2019; Kong et al., 2022).

Underground structures in Solanum are still poorly described relative to aboveground morphology. Plant collectors tend to focus on parts pressable on herbarium sheets, and excavation of tubers or rhizomes is often time-consuming [for an example, see discussion of S. echegarayi in Knapp et al. (2023)]. Documentation of underground structures in Solanum is expanding with increased field collection and the collaboration between ecologists and botanists in examining plant organs and their functions in the environment. Root tubers in the Asterophorum clade, for example, were only described less than 5 years ago (Gouvêa and Stehmann, 2019), and occur in species that are neither rare nor uncommon. We expect that knowledge of underground structures will evolve as botanists and ecologists better document their occurrence. Our dataset used here is likely to improve with time and collaborative effort.

We demonstrate that in Solanum, geophytes not only occupy environments with strong temperature seasonality and drought but show clear functional and ecological diversification between USO and rhizomatous growth forms. Such studies outside monocots are still surprisingly rare and our work provides a unique insight into how underground growth forms can evolve within a globally distributed clade that inhabits a wide range of habitat types.

More in-depth approaches for characterizing the functional ecology of underground organs in Solanum is a priority, but current available databases on functional root traits contain little data for this genus, despite its global distribution and agronomic importance. For example, only 12 of 1,232 species of Solanum (about 1%) have data available in the Global Root Trait database (GRootT), which seeks to build a large-scale, standardized and curated database of key root traits for plants (Guerrero-Ramírez et al., 2021). The available data is principally limited to traits related to resource acquisition, longevity, and root depth, and provide limited information on how these underground organs allow species to spread, compete and persist in the environment. Ultimately, developing a dataset of underground traits combined with aboveground traits will allow integration of datasets across different plant groups (Howard et al., 2019), and achieve better evolutionary explanatory power. Our dataset for Solanum, spanning a wide variety of habitats worldwide, is a significant first step in enabling further phylogenetically relevant macro-ecological and evolutionary studies.