Extreme cold plays a key role in the range boundaries of plants. Winter survival is central to their persistence, but not all structures are equally susceptible to frost kill and, therefore, limiting to distributions. Furthermore, we expect intraspecific variation in cold tolerance both within and among tissue types. In a laboratory setting, we determined freezing tolerances of two overwintering propagule types – seeds and rhizomes – of the globally invasive Johnsongrass (Sorghum halepense), testing apparent emergence and electrolyte leakage as a proxy for cell death. We used 18 genotypes from agricultural and non-agricultural habitats spanning the climatic extremes occupied by Johnsongrass in the US. Single node rhizome fragments had an average LT90 of -5.1 °C with no significant variation based on home climate or ecotype. Seeds frozen at -85 °C suffered a decline in germinability to 10% from 25% at 22 °C. Population origin did not affect seed response to any temperature. However, non-agricultural seeds germinated more and faster than agricultural seeds from the coldest climates, with a reversed relationship among warmest origin seeds. Regardless of ecotype, seeds from the cold/dry and wet/warm sectors of Johnsongrass’s range germinated more and faster. Drastic differences in cold tolerance between seeds and rhizome and evidence for seeds’ local adaptation to land use and climate suggest that its spread is likely limited by winter rhizome survival, as well as adaptability of germination behavior to longer winters. These findings shed light on Johnsongrass’ dispersal dynamics and help identify future avenues for mechanistically understanding its range limitation.

Cold tolerance, invasive plants, land use change, local adaptation, range boundaries

Species range limits are often dictated by climatic tolerances at large spatial scales. For most plants, temperature and moisture availability play a leading, though not unilateral, role in defining distributions (Curtis and Bradley 2016). To characterize how specific temperature or moisture parameters (e.g., extremes, durations, seasonality) limit a plant’s range, empirical methods and spatial models are typically used (Kotta et al. 2019; Greiser et al. 2020). In the temperate biomes of the northern hemisphere, cold winter temperatures play a role in limiting some perennial plants’ northern range boundaries, but are not sole drivers, interacting with growing season conditions that shape overwintering tissue maturation (Körner et al. 2016). While winter survival is measured at the whole-plant level, it is important to consider that a variety of organs (e.g., roots, herbaceous or woody shoots) may be responsible for cold tolerance to different degrees (Washburn et al. 2013; Sage et al. 2015; Ambroise et al. 2020). This means that distribution modeling based only on species presence is not sufficient to mechanistically identify cold tolerance limits (Gardner et al. 2019), especially in herbaceous perennials which often look similar in each season of their lifespan. Specifically, summer presence of such a species does not necessarily indicate survival through the previous winter, but potentially successful re-establishment from seed. This, in turn, muddles our understanding of how the plant and its constituent tissues respond to freezing temperatures. Cold tolerance is often considered at the organismal level, but testing different organs separately is most appropriate, particularly for perennial plants.

Both sexual and vegetative structures need cold tolerance to survive between growing seasons in non-tropical climates. While individual perennation is dependent on winter cold tolerance (Sage et al. 2015), sexual reproduction is dependent on annual seed production, dispersal, and establishment. In many temperate species, the winter season also intercedes between seed abscission and germination, requiring seed cold tolerance (Leiblein-Wild et al. 2014). Given the importance of cold tolerance to both survival and reproduction – and, therefore, overall fitness – its parameters (e.g, lethal temperature estimates) can be projected onto spatial climate patterns to hypothesize, though not predict, range limits (Sánchez-Fernández et al. 2012; Gardner et al. 2019). The strength of this approach is not in accurately and causally predicting distribution (which is better served by correlative modeling of multiple drivers), but in understanding the locations at which the specific stressor (i.e., temperature) is or is not likely to limit distribution.

Perenniality is a boon to plant fitness because it reduces each subsequent year’s demand for vegetative-to-reproductive allocation (Rohde and Bhalerao 2007). However, this requires major structural (e.g., xylogenesis) and non-structural carbohydrate production (Kozlowski 1992; Slewinski 2012), whose storage is mediated by winter minimum temperatures (Li et al. 2017). This is a key to some plants’ persistence through seasonal energy reallocation and storage in rhizome tissue (Boström et al. 2013). Across climate gradients, conditions acting on perennial tissue survival select for locally hardy (i.e., cold tolerant) genotypes (Malyshev et al. 2014; Dong et al. 2019). This selection can be driven not only by macroclimates, but also by anthropogenically-induced microclimates (e.g., irrigated fields, urban heat islands, etc.) and resource subsidies (Charrier et al. 2015; Oraee et al. 2020).

Perenniality can also buffer against challenging growing conditions (e.g., cold winter temperatures, Wingler 2015), and therefore impact species distribution. An important link between perenniality, fitness, and distributions is dispersal capability, which we can better understand by comparing propagules of varying anatomy, stress tolerance, and transportability (e.g., seed vs. rhizome).Thus, if local climates and habitats (i.e., provenance) select for biologically significant differences in cold tolerance and perenniality, this raises two questions. Firstly, could ranges suitable for vegetative propagation be more provenance-limited than ranges suitable for seed propagation? Secondly, if habitat type impacts perenniality via cold tolerance, could species distribution be mediated by land use as a selection pressure? These two questions have not been explicitly studied, but we might anticipate many interactions between land management and adaptation in current and future climates (Ramesh et al. 2017; Weber et al. 2017).

Advantages of having rhizomes (perennial underground stems) are evident across many plant systems (e.g., Chen et al. 2015; Grewell et al. 2019). In particular, rhizomes are a common feature of invasive species, which lend themselves well to the study of range limits and stress tolerance due to high data availability and ease of propagation. For example, some perennial invaders are able to maintain their competitive edge in recipient communities despite initial reductions in growth due to rhizome fragmentation (Zhou et al. 2017). In other cases, perennial invasive plants emerging from rhizomes are more competitive and stress tolerant than conspecific seedlings (Mitskas et al. 2003; Acciaresi and Guiamet 2010). Other forms of modified perennial stems, such as stolons, are believed to buffer invasive plants from stress and promote colonial expansion (Roiloa and Retuerto 2016). For instance, Pompeiano et al. (2015) found metabolic adjustment of sugar and proline concentration across all organs of the rhizomatous invader Arundo donax to explain differences in cold hardiness between Hungarian and Honduran populations. Studying populations from across a climate gradient, Dietrich et al. (2018) found Dactylis glomerata rhizome cold tolerance to correlate negatively with mean precipitation at home habitats across Europe. However, it remains unclear which climate and habitat factors drive variation in rhizomatousness in perennial invasive plants, and whether these species may forgo perenniality to colonize more challenging ranges. To address this, we used a model perennial invasive grass to evaluate the effects of climate and habitat on both rhizome and seed cold tolerance.

The cosmopolitan invader Johnsongrass (Sorghum halepense) sexually reproduces annually through seed, while its perenniality is achieved by rhizome survival through the winter (Washburn et al. 2013). It has been estimated that a single Johnsongrass genet can produce 33,600 kg ha^-1 of rhizome annually (McWhorter 1972). The per-propagule establishment efficiency of rhizome over seed is another factor in its importance to the invader’s persistence (Atwater et al. 2017). At Johnsongrass’s northern range edge in southern Ontario, there are reported persistent annual populations which are presumed to be caused by failure of winter rhizome survival (Warwick and Black 1983). Given the abundant evidence for great intraspecific variation in Johnsongrass (Atwater et al. 2016, 2017; Sezen et al. 2016), it appears likely that large differences in minimum temperature across this plant’s U.S. range could have led to local adaptation of rhizome cold tolerance based on climate.

Rhizome cold tolerance may also be related to overall rhizome development, which responds to resource inputs. Fertilization and irrigation can be responsible for rhizome development changes compared to growth in non-agricultural settings (Schmid and Bazzaz 1992; Schwinning et al. 2017), therefore selection may be different in cropland habitats where more rhizome biomass is generated than in more stress-associated non-agricultural environments. This advantage in growing season resources may buffer against winter kill, similar to the way that trees rely on tissue maturation to survive freezing temperatures (Körner et al. 2016).

Specific to Johnsongrass, Atwater et al. (2017) found that, while seed was more efficient at reproducing than rhizome on a per-unit-carbon basis, rhizome was more efficient than seed per propagule. The same study found Johnsongrass plants emerging from rhizome fragments to be more sensitive to habitat variation, competition, and density than seedlings. In addition, there has been abundant evidence of ecotypic differences interacting with home climate and response to competition in Johnsongrass in the United States (Atwater et al. 2016). These pieces of information suggest that the species’ different reproductive allocation strategies could be mediated by habitat type. However, no studies had isolated winter rhizome survival – on which the invader’s perenniality depends – as affected by home climate or habitat. Fletcher et al (unpublished data) found populations from across the U.S. range to be incapable of winter survival at the northern range edge (Ithaca NY) regardless of ecotype, as opposed to 100% survival in Virginia, Texas, and New Mexico. Home climate- and ecotype-based differences in Johnsongrass’s perenniation have remained an important knowledge gap. Combining our awareness of photosynthetic differences between agricultural and non-agricultural populations (Kelly et al. 2020; Lakoba and Barney 2020) and tissue maturation’s role in cold tolerance (Körner et al. 2016), we chose to test for ecotypic differences in seed and rhizome cold tolerance.

In the broader context of plant invasion biology, we set out to test whether adaptation to different land uses can yield divergent stress adaptation in a relatively short period of time (i.e., decades to centuries). While other studies have investigated differences between geographic ecotypes and home climates as predictors of perennial plant cold tolerance (Pompeiano et al. 2015; Dietrich et al. 2018), ours is the first to compare climate origin with land use origin. To address this research gap, we subjected populations of agricultural and non-agricultural Johnsongrass ecotypes representing a wide range of home climates to sub-zero temperatures to evaluate rhizome and seed cold tolerance. Specifically, we wanted to know whether: 1) populations from colder and/or drier climates exhibit greater cold tolerance; 2) populations from non-agricultural habitats exhibit greater cold tolerance due to the lack of agricultural inputs aiding growth and non-structural carbohydrate storage; 3) cold tolerance trends within and among populations differ between seeds and rhizomes. The findings will help us further understand the implications of habitat switching in Johnsongrass as well as offer a new link between land use change, climate, and invasive species.

We found that both Johnsongrass seed and rhizome are affected by exposure to acute treatment temperature minima, but on very different scales. While rhizome emergence showed a sharp decline from ~50% emergence to non-viability in the vicinity of -5 °C, seed germinability declined very gradually from ~25% to 10% across the gradient of 22 to -85 °C. The rhizome survival threshold of approximately -5 °C confirmed Hull’s (1970) finding of sharp decreases in rhizome survival between -3 °C and -5 °C to the point of no viability, and appears consistent across the broad geographic range. Thus, we saw that overwintering potentials of seed and rhizome are vastly different, implying seed survivability in climates north of Johnsongrass’s current range, far beyond known non-perenniating populations in southern Ontario (Warwick and Black 1983). Additionally, population differences based on home climate and ecotype were observed in seed, but not in rhizome. While these differences may be adaptive, it is all but certain that successful germination following -85 °C freezing treatment is not a trait selected for in the landscape, as Johnsongrass does not experience such extremes anywhere in, or near, its range; nor are there many places on Earth with such low temperatures. Interestingly, the impacts of home MinT on seed GP and MGT were mediated by ecotype identity, which may shed light on agricultural practices selecting for traits related to temperature and, by association, day length. The twofold (50% vs. 25%) difference in baseline (control) emergence of rhizome over seed, in conjunction with the latter’s vast numerical superiority, echoes Atwater et al. (2017) conclusion that rhizome is more efficient than seed on a per-propagule basis, while seed is more efficient on a per-unit-carbon basis.

We found no differences in seed germination or rhizome emergence response – and therefore no differences in cold tolerance – to freezing treatments based on home MAP and MinT. There were, however, inherent differences in seed germination response based on home MAP and MinT. This yielded a response surface where cool/dry and warm/humid origin Johnsongrass populations germinated more and faster than cool/humid and warm/dry origin populations. It is possible that reduced and delayed germination on dry sites may be a conservative strategy selected for by drought stress, which can be especially damaging for seedlings, compared to seeds or mature plants (Schwinning et al. 2017). Meanwhile, an adaptation for proportionally higher germination in cold habitats may buffer against inevitable partial die-off in late frosts. Though associated more with growing season length than late frosts specifically, such a strategic adaptation has been found in the introduced range of the invader Ambrosia artemisiifolia (Leiblein-Wild et al. 2014). Earlier germination and leaf-out has been known to correspond to quicker recovery following freezing damage (Menzel et al. 2015). Therefore, one explanation for no difference in MGT based on home MinT could be that seedling Johnsongrass is unable to resprout following frost damage. This vulnerability may also contribute to inefficiencies in adapting to conditions at its northern range limits (Fletcher et al. 2020). Finally, earlier leaf-out as part of overall extended leaf phenology is a common strategy among invasive plants (Fridley 2012), though its ultimate utility in carbon gain is diminished at increasingly northern latitudes (O’Connell and Savage 2020).

No differences were found between agricultural and non-agricultural populations’ seed or rhizome responses to freezing treatment temperatures, indicating no differences in cold tolerance based on ecotype identity. However, we again found inherent differences in GP and MGT based on home MinT as mediated by ecotype identity. Non-agricultural populations germinated more and faster than agricultural ones when originating from colder climates; however this did not translate to any differential response to our cold treatments. Given the smooth decrease in germinability from +22 °C to -85 °C treatments across all populations, it makes sense that population differences based on a MinT range of -10 °C to +5 °C are unrelated to survival of -85 °C or even -20 °C treatments. Tolerance of the extreme cold could not have been selected for in the landscape, as Johnsongrass seed does not encounter these temperatures. Unfortunately, we were limited by available equipment to test temperatures between -10 and -85 °C. We had posited that any differences in cold tolerance between ecotypes could be driven by energy assimilation and storage from a more favorable preceding season; however, this could not have been the case for seed, as we accounted for maternal effects by using only germplasm that had been grown out in a common environment for a generation.

Cold tolerance differences between seed and rhizome were so vast that they cannot be compared by LT90 values. Seed GP approached 0.1 (analogous to LT90) around -85 °C and no colder treatment was available, meaning that a true dose response curve could not be built for seed as it was for rhizome. This extreme cold tolerance across Johnsongrass populations informed us that seed freezing is likely not range limiting. Given no origin MAP or MinT differences in rhizome LT90, we also cannot test whether the annual climate niche is significantly different between populations. Instead, our evidence points to propagule pressure and phenology as likely factors of northern range limitation. Given the relatively high winter temperatures that rhizomes cannot survive, rhizome segments likely cease to be feasible propagules for range expansion in regions with climates similar to southern Ontario, where populations persist only via seed (Warwick and Black 1983). The general uniformity of rhizome LT90s across the species range suggests that rhizome cold tolerance is unlikely to adapt to colder climates in the future. Because we used only one cold acclimation regime across treatments, it is also possible that differences in acclimation capacity are present between populations, such as found in Miscanthus by Peixoto and Sage (2016). However, we have less reason to suspect such an effect in Johnsongrass, as Miscanthus is known to have a wider overall range in cold tolerance among genotypes (Fonteyne et al. 2016). It should also be noted that, even though GP values were low overall, the stochastic nature of these reductions and the very high significance level of model effects across 18 populations tell us that these correlations are robust. It is not unexpected to observe low overall seed germinability in weedy and invasive species that have evolved seed dormancy.

Dormant and non-dormant seeds are clearly not range limiting to Johnsongrass as a species, nor limited based on ecotype or home climate. In other words, seed from anywhere in the North American range can survive winter temperature minima anywhere on the continent. Given that Johnsongrass persists in places with colder winter temperatures than the rhizome LT90 of -5 °C, thermal dynamics of soil are clearly a factor that prevents us from simply predicting cold temperature range limitation. Lack of apparent climate or ecotype adaptation of rhizome cold tolerance tells us that this may be a stable trait within the species, while an expanding “annual range” beyond the perennial range is feasible. However, even though seeds may always be cold tolerant, seedlings are likely to be much more vulnerable to stressors such as late frosts and droughts (Olson et al. 2018). Conceivably, more and faster germination as an adaptation to colder climates could be based on limited photosynthetically active radiation (PAR) of shorter summers. But why is this only seen in non-agricultural populations, while the inverse is true of agricultural ones? One driver of colder origin agricultural seed germinating less and slower than its non-agricultural counterpart could be greater winter/spring cold exposure in cropland due to bare soil (Snyder et al. 2015). It is also possible that early application of post-emergence herbicides could select for later germination. For example, a genetic link between herbicide resistance, dormancy and germination behavior is seen in the weedy grass Alopecurus myosuroides (Délye et al. 2013). Other agricultural factors could be at work, as seed phenology is known to be tremendously adaptive to cropping systems (Batlla et al. 2020). Molecular regulation of cold tolerance is currently an ongoing investigation with much progress made in understanding both the stress signaling and acclimation response involved (Ding et al. 2019).

One of the primary challenges in interpreting rhizome cold tolerance and forming hypotheses about continental distributions is the interaction of climate change with snow cover and, thereby, insulation of soils in winter. Rhizome carbohydrate storage, bud formation, survival, and phenology of spring emergence are known to be sensitive to winter snow depths (Lubbe et al. 2021). Specifically, reduced snowpack as a result of winter warming has been seen as particularly injurious to rhizomatous species populations, suppressing their competitive ability in herbaceous communities (Lubbe and Henry 2021). Winter warming, rather than summer warming, has been linked to major shifts in grassland primary production, species composition, and soil respiration (Kreyling et al. 2019), suggesting that future studies should also focus on temperature minima at population origins to study community assembly, as well as improved measures of perenniation, budbank, and multi-trait assessment (Lubbe et al. 2021).

By uncovering drastic differences in cold tolerance and between organs and populations, we are able to better understand their potential contributions to species distributions. We can begin to deduce which organs may or may not be limiting to overall plant stress tolerance and whether there are other physiological or phenological drivers of known range limits. Likewise, we can narrow possible drivers of range limitation and connect them to spatially explicit habitat parameters. However, we must be careful not to conflate experimentally isolated stress limits with distribution boundaries (Curtis and Bradley 2016). In this example of Johnsongrass seed versus rhizome cold tolerance, it becomes clearer whether or not a trait could have been acquired through selection. Namely, we see that, while seed is virtually unlimited by temperature minima in North America, we should not assume a lack of northern range limit. Likewise, we found rhizome cold tolerance to be less than what the perennial range would suggest, congruent with Curtis and Bradley’s (2016) findings across many species. Furthermore, this approach has provided us with leads in terms of which organs, life stages, and stressors to investigate further. It also allowed us to characterize seed germinability and germination performance of different Johnsongrass ecotypes over a broad range of home climates.

We see that not all propagules of a plant respond similarly to all stresses – cold temperatures being a key example. In studying and managing invasive plants, this can inform our understanding of likely dispersal vectors. Our findings on Johnsongrass, in corroboration with Atwater et al. (2017), tell us that seed is a far more efficient, stress tolerant, and easily transported propagule than rhizome. This implies that, if most Johnsongrass populations are established by seed, we should expect ample genetic variation within and among landscapes compared to stands where clonal propagation is likely to occur. This genetic diversity contributes to the known intraspecific variation across the continent (Atwater et al. 2016, 2017; Sezen et al. 2016; Fletcher et al. 2020; Kelly et al. 2020), as well as its likelihood of adaptation to new habitats. At the same time, our study suggests some hard limits to cold adaptation, which is both ecologically interesting and informative for invasion risk assessment.

We thank Valerie Thomas and Brian Strahm for providing manuscript draft feedback. We thank Dave Mitchem for consultation on laboratory methods. We thank David Haak for laboratory equipment use. We thank Edward Gaines for common garden care.

This work was partially supported by the Virginia Tech College of Agriculture and Life Sciences and the National Institute of Food and Agriculture Global Food Security CAP [2015-68004-23492 to JNB].

We would like to acknowledge support in the publication of this article from Virginia Tech's Open Access Subvention Fund.