Evidence for rapid evolution of phenology in an invasive grass


Correspondence: Ari Novy, United States Botanic Garden, 245 First Street, SW, Washington, DC 20024, USA.

Tel.: +1 202 225 1269; Fax: +1 202 225 1561; e-mail: anovy@aoc.gov


Evolutionary dynamics of integrative traits such as phenology are predicted to be critically important to range expansion and invasion success, yet there are few empirical examples of such phenomena. In this study, we used multiple common gardens to examine the evolutionary significance of latitudinal variation in phenology of a widespread invasive species, the Asian short-day flowering annual grass Microstegium vimineum. In environmentally controlled growth chambers, we grew plants from seeds collected from multiple latitudes across the species' invasive range. Flowering time and biomass were both strongly correlated with the latitude of population origin such that populations collected from more northern latitudes flowered significantly earlier and at lower biomass than populations from southern locations. We suggest that this pattern may be the result of rapid adaptive evolution of phenology over a period of less than one hundred years and that such changes have likely promoted the northward range expansion of this species. We note that possible barriers to gene flow, including bottlenecks and inbreeding, have apparently not forestalled evolutionary processes for this plant. Furthermore, we hypothesize that evolution of phenology may be a widespread and potentially essential process during range expansion for many invasive plant species.


Biological invasions may be enabled by biotic or abiotic characteristics of invaded habitats, traits of the introduced species or some combination (Catford et al., 2009; Gurevitch et al., 2011). Theoretical and empirical studies suggest that evolution during the invasion process may be an important but underappreciated facet of biological invasions (Baker, 1974; Lee, 2002; Novak, 2007; Lankau et al., 2009; Dormontt et al., 2011). Colonization by any species of novel habitats generally results in exposure to novel selective regimes (Suarez & Tsutsui, 2008), founder effects, genetic drift, and/or hybridization events (Bossdorf et al., 2005). Therefore, rapid evolution may be a key feature of range expansions for both native and introduced species (Maron et al., 2004; Montague et al., 2007; Whitney & Gabler, 2008; Xu et al., 2010). Evolution may even be necessary before a new arrival can become invasive, thus potentially explaining the lag time that sometimes occurs during the invasion process (Crooks, 2005). Furthermore, an understanding of the evolutionary processes that allow for persistence of founding populations, despite predictions that such populations should fail (e.g. Lynch et al., 1995), is of fundamental importance in elucidating the processes that lead to successful invasion. However, despite the often-cited possibility that invasions are enabled by evolutionary processes, there are relatively few empirical studies examining the phenomenon (Colautti et al., 2009).

Studies of life-history evolution and community ecology theory have both been leveraged to suggest that evolution of phenology (i.e. the seasonal timing of reproduction and other life-history events) ought to be an important aspect of range expansion and invasion success (e.g. Griffith & Watson, 2006; Wolkovich & Cleland, 2011). Phenology has been shown to respond to various selective pressures (Griffith & Watson, 2006; Franks et al., 2007). In particular, genetically controlled phenological timing has been associated with fitness benefits through interaction with frost avoidance (Kuser & Ching, 1980), climate change (Bradshaw & Holzapfel, 2001), growth rates (Blair & Wolfe, 2004), defence responses (Meyer & Hull-Sanders, 2008), reproductive rates (Brown & Eckert, 2005), plasticity (Lavergne & Molofsky, 2007), and trade-offs with size at reproduction (Colautti et al., 2010). Furthermore, the quantitative genetic nature of flowering timing in plants (Chardon et al., 2004) increases the likelihood of observing evolution of phenology in nature.

In this study, we evaluated the role of rapid evolutionary processes as an enabler of biological invasions by examining patterns of variation in two key life-history traits predicted to trade-off with one another: reproductive phenology and size at reproduction. We interpret our findings as evidence of rapid evolution of the life-history trade-off between flowering time and size at reproduction in the invasive grass Microstegium vimineum. We further posit that the decrease in genetic diversity which probably accompanied invasion did not prove sufficient to forestall evolution in this species and that evolution of phenology may be a common process associated with plant invasions.

Materials and methods

Study species

Microstegium vimineum (Trin.) A. Camus (stiltgrass, family Poacaeae, order Poales) is a C4 annual grass native to Asia, where it is found in various habitats, including forest margins and riparian areas (Chen & Phillips, 2007). It was first recorded in North America in 1919 in Tennessee (Fairbrothers & Gray, 1972), but is now invasive in more than 20 U.S. states (USDA & NRCS, 2005), and it continues to spread rapidly. Microstegium vimineum flowers in the fall in response to short days (Judge, 2006) and produces abundant seed that is dispersed by water, animals, recreational activities, mowing and timber harvests. Invasive M. vimineum is often first found along roads, trails, and streams, but it can also colonize full-sun forest openings, shaded forests, and riparian areas (Cheplick, 2010; Flory, 2010), where it forms dense and persistent populations. It is highly shade tolerant and can produce seed under very low-light conditions (Horton & Neufeld, 1998; Cheplick, 2010). The abundance of M. vimineum in the field is highly correlated with light availability (Cheplick, 2010; Flory, 2010), and it grows best under high-light and high-moisture conditions in experimental microcosms (Droste et al., 2010). Microstegium vimineum produces chasmogamous flowers, borne on spikes that are terminal on the culm, and cleistogamous flowers, borne on spikes contained within the leaf sheaths of the upper two or three culm segments (Chen & Phillips, 2007; Cheplick, 2007). Chasmogamous flowers are capable of both self-pollination and cross-pollination (i.e. facultatively outcrossing) as both stigmas and anthers are simultaneously exposed at maturity. Cleistogamous flowers are fully self-pollinated as they are enclosed by the leaf sheath in which they are contained. The species appears to be predominantly selfing, with one study measuring overall cleistogamous seed production at 62% (Gibson et al., 2002), although actual selfing rates may be higher as facultative outcrossing of chasmogamous flowers results in selfing as well as outcrossing.

Collection of plant material

We collected M. vimineum seeds from 10 U.S. populations, representing the majority of latitudinal variation in its invasive range (Table 1). Because M. vimineum grows in dense stands and produces lateral tillers, it was not possible to separate individual plants and positively identify maternal origin of each seed. Therefore, for each population, hundreds to thousands of seeds were randomly collected from individual stems within a circular plot of 20 m diameter. Seeds were air dried and stored in paper bags at room temperature until sowing. We randomly selected seeds for sowing from our bulk seed samples to approximate the genetic diversity of our sampling locations among growth chamber grown plants.

Table 1. Collection locations in the United States for the 10 invasive Microstegium vimineum populations sampled and their mean time to anthesis under the northern and southern light treatments
PopStateNearest townLatitudeLongitudeMean days to anthesis (± SE)
1South CarolinaHopkins33°48′28″ N80°51′55″ W119.5 (± 1.4)102.7 (± 0.4)
2North CarolinaChapel Hill35°53′24″ N79°00′56″ W123.8 (± 2.3)105.7 (± 0.9)
3VirginiaFort Belvoir38°42′22″ N77°08′48″ W103.8 (± 1.2)93.1 (± 0.5)
4MarylandWhittman38°47′43″ N76°17′40″ W103.7 (± 0.6)86.7 (± 0.5)
5VirginiaGreat Falls38°57′44″ N77°16′44″ W99.1 (± 1.7)84.6 (± 1.2)
6DelawareDelaware City39°34′22″ N75°34′50″ W94.9 (± 2.5)83.8 (± 1.1)
7West VirginiaMorgantown39°39′45″ N79°59′00″ W89.5 (± 2.3)75.6 (± 1.8)
8PennsylvaniaMurraysville40°26′05″ N79°41′50″ W85.6 (± 0.9)70.1 (± 0.5)
9New JerseyFlemington40°30′44″ N74°53′04″ W79.0 (± 0.7)71.1 (± 0.5)
10ConnecticutOrange41°18′18″ N72°59′54″ W84.1 (± 1.0)70.1 (± 0.7)

Experimental design

To quantify genetic variation in phenology among the invasive M. vimineum populations, we grew all populations under common garden conditions in growth chambers. To evaluate phenological responses generally, as opposed to under one specific latitudinal habitat, we manipulated day length in four 8.9 m2 controlled growth chambers (model GC-96-11-CW-C3; Environmental Growth Chambers Inc., Chagrin Falls, OH, USA). Two growth chambers were set to simulate growing season day length conditions at the northern extreme of M. vimineum's invasive range, approximately 42° N latitude, whereas the other two chambers simulated growing season conditions at the southern extreme, approximately 34° N latitude (USDA & NRCS, 2005). The growth chambers were set to simulate light conditions beginning on June 1 and progressing as under natural conditions for the duration of the experiment. Day length progressions were determined using U.S. Naval Observatory tables (http://aa.usno.navy.mil/). Humidity was set to a constant 70% and temperatures were set to 26 and 22 °C for day and night, respectively, in all growth chambers.

We germinated randomly selected seeds from each population in individual four inch plastic pots, filled with Fafard Growing Mix 2 (Conrad Fafard Inc., Agawam, MA, USA). Pots were then randomized into five blocks in each chamber to control for within chamber environmental heterogeneity, and separated by at least 5 cm to prevent plants from rooting in neighbouring pots and to minimize light competition among plants. Each chamber contained five randomized blocks with two plants from each of the 10 populations in each block, for a total of 100 plants per chamber and a total of 400 plants across all four chambers. Plants were watered every other day and were fertilized with dilute 20-20-20 NPK liquid fertilizer (Scotts Co., Maysville, OH, USA) and iron chelate (Becker Underwood, Inc., Ames, IA, USA) bi-weekly.

Data collection

We visually inspected all plants daily for signs of flowering. We recorded the date of first anthesis (i.e. the first emergence of anthers) for each plant, and tabulated the number of days from germination to anthesis. Plants were allowed to grow until senescence, which was defined as all flowering complete, seeds fully mature, and with more than 70% (visual assessment) of the plant brown. As each plant reached senescence, it was removed from the growth chamber. Above- and below-ground biomass were separated, roots were washed to remove soil, and all biomass was dried at 60 °C to constant mass and weighed. Above- and below-ground mass were measured separately and summed to calculate total plant biomass.

Data analysis

To determine whether latitude of M. vimineum origin determined phenology, we analysed days to anthesis and plant biomass using mixed model analyses of variance (anovas; Proc Mixed, SAS Institute, Cary, NC, USA). Population origin, north/south light treatment, and their interaction were modelled as fixed effects, and chamber, block, and block nested within chamber were modelled as random effects. Significant effects of population origin on flowering time or biomass measurements of M. vimineum would indicate genetic determination of these traits. Genetic effects are likely to be relatively uniform within M. vimineum populations due to high rates of inbreeding that result from self-compatibility and a large proportion of the flowers being cleistogamous (i.e. obligately inbreeding). Cheplick (2007) found that M. vimineum biomass allocation to cleistogamous flowers was over twice that of allocation to chasmogamous flowers in edge habitats and approximately 15% higher in shaded habitats. Even the terminal, chasmogamous inflorescences are likely to promote inbreeding, due to the plant's receptivity to self-pollen. We also performed regression analyses (SigmaPlot 11.0; Systat Software, Inc., San Jose, CA, USA) to compare days to anthesis and biomass responses to the latitude of source populations. Regression analyses were conducted separately for each treatment, but were pooled and averaged for each seed origin across block and chamber replications.


Of the 400 individuals planted in the growth chambers, 398 plants were included in the dataset for days to anthesis and 373 were included in the dataset for biomass. The 27 plants not included in the biomass dataset did not die. Rather, these plants suffered mechanical injuries during the course of the experiment that interfered with normal growth progressions and, hence, final biomass. In only two of these cases did the damage occur before anthesis, and so only those two plants were not included in the dataset for days to anthesis.

Overall, we observed a clear cline in time to anthesis (Table 1 and Fig. 1) and biomass (Fig. 2), based on latitudinal origin of populations, with more northern populations flowering earlier and producing less biomass under both northern and southern photoperiods. All populations reached anthesis later under northern photoperiods than they did under southern photoperiods (Table 1). There were significant effects of both population origin and the north/south light treatment on time to anthesis and all biomass measurements (Table 2). All variables except for aerial biomass exhibited significant interactions between population origin and light treatment. The only significant random effect was block nested within chamber for total and aerial biomass (Table 2). Under the northern light treatment, the average time to anthesis ranged from 79.0 to 119.5 days across the populations, corresponding to critical photoperiods at anthesis of 10 h : 54 min to 12 h : 56 min, depending on seed origin. For the southern light treatment, the average time to anthesis ranged from 70.1 to 105.7 days across populations, corresponding to critical photoperiods at anthesis of 11 h : 42 min to 12 h : 48 min, depending on seed origin. Days to anthesis was negatively related to population latitudinal origin in both the northern (r2 = 0.847; < 0.001; Fig. 1a) and southern (r= 0.835; < 0.001; Fig. 1b) light treatments.

Table 2. anova results for the fixed effects of population origin, light treatment, and their interactions, and the random effects of experimental chamber, block, and block nested within chamber on Microstegium vimineum days to anthesis, total biomass, aerial biomass, and root biomass
Source of variationNum d.f.Den d.f.Days to anthesisTotal biomass (g)Aerial biomass (g)Root biomass (g)
Fixed effects
Population9373217.91 < 0.0001 30.23 < 0.0001 26.06 < 0.0001 35.51 < 0.0001
Treatment137371.00 < 0.0001 6.66 0.0103 6.80 0.0095 4.85 0.0282
P × T93732.94 0.0022 1.97 0.0421 1.590.11662.47 0.0095
Covariance parameterEstSEEstSEEstSEEstSE
  1. Bold indicates significant differences (α = 0.05). ‘Est' is the covariance parameter estimate and ‘SE’ in the standard error of the covariance parameter estimate. ‘n/a’ specifies that Wald Z values could not be calculated due to negative covariance estimates, which indicates that the random effect was not significant.

Random effects
B (C)n/an/a 0.1293 0.0747 0.08729 0.04960.00310.0039
Figure 1.

Relationship between latitude of population origin and days to anthesis for Microstegium vimineum under the northern (a) and southern (b) light treatments. Bars indicate standard errors.

Figure 2.

Relationship between latitude of population origin and Microstegium vimineum performance under the northern and southern light treatments as measured by root biomass (a, b), above-ground biomass (c, d), and total plant biomass (e, f). Bars indicate standard errors.

Similar patterns were observed in biomass responses for both light treatments. For the northern light treatment mean root biomass ranged from 0.997 to 2.180 g (r= 0.748, < 0.001; Fig. 2a), above-ground biomass ranged from 3.485 to 5.870 g (r= 0.633, = 0.006; Fig. 2c), and total biomass ranged from 4.482 to 7.882 g (r= 0.633, = 0.002; Fig. 2e). For the southern light treatment mean root biomass ranged from 0.734 to 1.690 g (r= 0.712, = 0.002; Fig. 2b), above-ground biomass ranged from 2.815 to 4.715 g (r= 0.604, = 0.008; Fig. 2d), and total biomass ranged from 3.711 to 6.370 g (r= 0.704, = 0.002; Fig. 2f).


Evolution of phenology

Our results clearly demonstrate a strong latitudinal cline for the number of days required to reach anthesis and the amount of biomass produced by M. vimineum populations collected from throughout its invasive range. Plants from higher latitudes flowered earlier and produced less biomass than plants from more southern populations. Growing plants in a common environment in growth chambers allowed us to demonstrate that these traits are most likely under genetic control, while replication of the experiment under two distinct light regimes confirmed that these trends are generalized findings, independent of specific local light regimes. Possible alternative explanations for the phenotypic patterns we recorded here include maternal or epigenetic effects. However, we believe these kinds of effects to be unlikely in this case. As Montague et al. (2007) noted, maternal effects are unlikely for the types of traits measured in this experiment. Maternal effects are only expected when seed dispersal distances are less than pollen movement (Galloway, 2005), a prerequisite clearly not met by the highly selfing M. vimineum. In addition, maternal effects are most likely expressed as early life-history traits such as propagule quality (Rossiter, 1996) as opposed to the late life-history traits studied here (days to anthesis and biomass production). However, to demonstrate that maternal effects are unlikely for the materials used in this study, we did examine early life-history traits in seven of the seed lots for which we had ample seed material. For these seven populations, there were no significant differences in germination or seed weights based on seed origin (P > 0.05, see also Flory et al., 2011). To date, the heritable epigenetic effects known to exist in plants are primarily triggered by various forms of biotic and abiotic stresses and generally lead to changes in genomic stability, such as increased transpon activity (Hauser et al., 2011). Although heritable epigenetic effects could have adaptive significance in some cases, it is difficult to imagine a scenario in which they could be responsible for the phenotypic cline observed here. In any case, even an epigenetically mediated effect that becomes a stably heritable element still represents a form of adaptive evolution. As such, our observed population differences indicate divergent phenological and biomass allocation characters under genetic control. Moreover, we conclude that clinal variation in the traits we measured is most likely due to adaptive evolution, as a result of differing selective pressure to complete flowering and seed maturity before the end of the growing season (i.e. cold temperatures arresting seed maturity) at differing latitudes. We suggest that such evolution of phenological traits has permitted the rapid expansion of M. vimineum invasions into more northern habitats and such evolutionary patterns may be a common trait of invasive plant species.

Although we could not measure fitness consequences of flowering time and biomass variation directly in this study, adaptive evolution of phenological timing is the most likely explanation for what we observed here. The only other possible explanations to explain such clinal patterns would be a nonadaptive process such as isolation-by-distance (IBD) or that native M. vimineum propagules were transported, via multiple introductions, from latitudes in Asia to environmentally equivalent latitudes in North America (preadaptation). IBD could result from the northern phenotype (i.e. shorter time to flowering at lower biomass) being introduced in the northern United States and the southern phenotype (i.e. longer time to flowering at higher biomass) being introduced in the southern United States. If these two phenotypes dispersed towards each other, eventually overlapped and began crossing (and these traits were neutral), we could expect to observe such a cline as a result of IBD. However, it is unlikely that flowering timing and biomass determination are neutral traits, though direct measurement of fitness or reciprocal transplant experiments would be required to confirm nonneutrality. The introduction and dispersal patterns that would be necessary in an IBD or preadaption scenario are also unlikely given the available herbarium records in North America. The plant was first noticed in the southeastern United States by the 1910s, and then radiated northward and westward (Fairbrothers & Gray, 1972). Though we cannot preclude the possibility of multiple introductions, even such introductions would almost certainly have been discrete events, located at major shipping locations, of plants from a limited latitudinal distribution in Asia, as the plant has been reported to be introduced as packing material for ceramics imported to North America from central China (Dorman, 2008). Furthermore, steady range expansion of M. vimineum across North America, particularly northward, has been noted in recent years (Mehrhoff, 2000).

Most interestingly, the evolutionary patterns we observed here must have arisen over a 100 year period or less. We are aware of three other genera of invasive plants for which similar flowering phenology clines have developed after invasion: Lythrum salicaria in North America (Montague et al., 2007), two Solidago species in Europe (Weber & Schmid, 1998) and Impatiens glandulifera in Europe (Kollmann & Bañuelos, 2004). Interestingly, both the invasive Solidago species and L. salicaria are self-incompatible, whereas I. glandulifera is self-compatible but protandrous (i.e. male flowers maturing before female flowers, often to promote outcrossing). Therefore, M. vimineum is the first invasive plant species identified that has evolved clinal flowering time phenology in its invasive range but does not possess biology favoring, or requiring, outcrossing. In fact, M. vimineum's biology promotes selfing due to cleistogamy; although, we should note that Maron et al. (2004) observed clines in biomass and fecundity for the invasive apomict Hypericum perforatum. Microstegium vimineum may also have evolved clinal phenology in a shorter period of time than Solidago spp., L. salicaria and I. glandulifera, which were all introduced in their invasive ranges by the early 1800s (Weber & Schmid, 1998; Blossey et al., 2001; Kollmann & Bañuelos, 2004). Furthermore, unlike these other species, M. vimineum is a wind-pollinated, monocot. With the addition of M. vimineum to the examples of invasive plants that have evolved flowering phenology in their invasive ranges, we now have evidence of such evolution in a wide variety of plant families and orders (Lythraceae, Myrtales; Balsaminaceae, Ericales; Asteraceae, Asterales; Poaceae, Poales), various reproductive biologies (highly selfing to obligate-outcrossing, and both wind and insect pollination), and several different habitat types (field, wetland and forest understory; e.g. Weber & Schmid, 1998; Kollmann & Bañuelos, 2004; Montague et al., 2007). This may indicate the presence of a general trend of phenological evolution during processes of plant invasion.

Microstegium vimineum seems to have undergone a lag phase from the time it was introduced in the early 1900s until it was recognized as an invasive species in the late 1980s (Barden, 1987). As a fecund annual with a relatively short-lived seed bank, M. vimineum possesses the potential for rapid adaptation, given adequate genetic diversity. At the minimum, it has cycled through approximately 100 generations in the invasive range, though evolution is likely to have taken place over a much shorter period of time in areas where the plant has only existed for a few decades (e.g. New England). Apparently, a tendency to inbreed has not impaired this species' ability to evolve clinal phenology, as it has done so even more rapidly than the outbreeding species that have evolved similar patterns. Interestingly, in our results there was a general trend towards smaller variance in days to anthesis for populations from the extremes of the invasive range, compared with the centre of the range (Table 1 and Fig. 1a,b). This could be a result of limited genetic diversity at the edges of the range, possibly due to decreased gene flow or stronger selection under the more extreme climate regimes expected at the northernmost range extents.

We have also conducted microsatellite (Simple Sequence Repeat) marker analysis on over 30 populations of M. vimineum from its native and invasive ranges (preliminary data published in Novy et al., 2012). We found that genetic diversity, measured both by heterozygosity and effective number of alleles, was lower in the invasive range, a clear indication of at least some degree of founder effect. Therefore, despite an initial bottleneck and high levels of inbreeding, M. vimineum has been able to evolve clinal variation in phenology over approximately 100 generations.


Both above- and below-ground M. vimineum biomass decreased with increasing latitudinal origin of populations. Because M. vimineum biomass is strongly correlated with seed production (total chasmogamous and cleistogamous seeds, r2 = 0.90, = 24, S. L. Flory, unpublished data), reduced biomass from more northern populations probably reflects decreased seed production relative to more southern populations, which was also found for invasive populations of L. salicaria (Colautti et al., 2010). It has long been appreciated that for short-day flowering plants, local survival of a plant species depends on the production of viable seed before frost, or other inhospitable climate conditions, arrests metabolism (e.g. Allard, 1932). As optimal flowering time, where reproductive output is maximized before seasonal climatic conditions become unfavourable, will vary with photoperiod latitudinally, short-day flowering plants must evolve appropriate critical photoperiods for each local habitat to maximize reproductive success. For M. vimineum, this suggests evolution of a life-history trade-off between flowering time and size at reproduction.


Our results demonstrate rapid evolution of phenology in the highly invasive grass M. vimineum, whereby flowering time and biomass allocation are strongly correlated with the latitude of population origin. We hypothesize that adaptive evolution via selection on flowering time is implicated, at least in part, for the northward spread of this species in the eastern United States. Moreover, M. vimineum is a nonclonal, inbreeding, annual grass, and the other invasive plant species that have demonstrated a similar pattern of clinal evolution in phenology include clonal, obligately outcrossing, and perennial species in widely divergent families and orders. This suggests that rapid evolution of phenology may be widespread and critically important in the range expansion of many invasive plant species, despite potential limitations to gene flow and probable historical bottlenecks.


We thank Jonathan Atwell, Peter Smouse and anonymous reviewers for valuable comments and suggestions on earlier versions of this manuscript. This project was funded by the USDA-NIFA McIntire-Stennis programme provided through the New Jersey Agricultural Experiment Station at Rutgers University.