Demographic fate of Arabidopsis thaliana cohorts of autumn- and spring-germinated plants along an altitudinal gradient


Corresponding author. E-mail:


1. Understanding how plants respond and adapt to varying environmental conditions has attracted the attention of plant ecologists for decades. To study this process, altitudinal gradients are used because of their inherent variation in environmental conditions. In the current scenario of global warming, altitudinal gradients may also represent a valuable resource to deepen our understanding about plant adaptive responses to predicted changes in environmental conditions.

2. Nowadays, the study of adaptive variation demands equal contributions from ecology and genetics. We need to assess the effects of selective pressures on phenotypic variation but also the genetic and molecular basis of phenotypic traits. The annual plant Arabidopsis thaliana represents a useful model system for investigating adaptive variation. Here, I characterized patterns of life cycle variation in natural A. thaliana populations along an altitudinal gradient to unravel how the species copes with different environmental conditions.

3. I periodically monitored thousands of plants from eight populations in NE Spain over 5 years (2007–2011) to estimate survival and fecundity schedules for autumn- and spring-germinated plants. Data were used to estimate net reproductive rate for each life cycle type. Data were regressed against altitude to detect altitudinal climatic clines for life cycle variation.

4. Survival of spring-germinated A. thaliana plants was significantly higher than that of winter-germinated plants. Plants from both cohorts exhibited similar fecundity values. The net reproductive rate of spring-germinated plants was fourfold higher than that of autumn-germinated plants. The proportion of spring-germinated plants increased significantly with altitude.

5.Synthesis. Arabidopsis thaliana can behave as a winter or spring annual plant. Nevertheless, the spring annual life cycle is clearly relevant to maintain A. thaliana populations, particularly at high-elevation locations. It is suggested that germination timing represents one of the most relevant traits to focus our efforts to understand adaptive variation in A. thaliana. The results illustrate the potential of annual plants to adjust their life cycles to varying environmental conditions encountered along a climatic gradient, which could mitigate the effects of global warming on annual plant populations.


Abiotic gradients represent unique natural laboratories to study fundamental issues in plant ecology, such as plant response to varying environmental conditions (Berry & Björkman 1980; Körner, Farquhar & Wong 1991; Angert & Schemske 2005; Byars, Papst & Hoffmann 2007; Giménez-Benavides, Escudero & Iriondo 2007; Gonzalo-Turpin & Hazard 2009; Montesinos-Navarro et al. 2011). In the case of altitudinal gradients, plants must cope with changing environmental conditions including increasing atmospheric pressure, decreasing air temperature, increasing cloudiness, and increasing UV-B radiation with increasing altitude (Körner 2007). Plants respond to these climatic shifts by adjusting key physiological processes, such as gas exchange and photosynthetic rates (Berry & Björkman 1980; Williams & Black 1993; Körner 2007), but also morphological traits and demographic rates, such as leaf shape or seed dormancy and germination patterns (Cavieres & Arroyo 2001; Cummins & Miller 2002; Ma et al. 2010; Phillips et al. 2011). The view that these changes can be adaptive has been frequently discussed among evolutionary ecologists (Turesson 1922; Clausen, Keck & Hiesey 1940; Angert & Schemske 2005; Giménez-Benavides, Escudero & Iriondo 2007; Gonzalo-Turpin & Hazard 2009; Stöcklin, Kuss & Pluess 2009; Alberto et al. 2011; Montesinos-Navarro et al. 2011). In the context of global warming effects, which have been shown to be accelerated at high elevations (Nogués-Bravo et al. 2007), altitudinal gradients may be used to understand how plants adjust their phenological traits to adapt to new environmental conditions (Jump & Penuelas 2005; Inouye 2008; Giménez-Benavides et al. 2011).

Nowadays, it is widely accepted that the study of adaptive variation demands a multidisciplinary approach. It is equally important to identify not only the effects of selective pressures on phenotypic variation but also the genetic and molecular basis of such phenotypic variation (Mitchell-Olds 1995; Feder & Mitchell-Olds 2003; Alonso-Blanco, Méndez-Vigo & Koornneef 2005; Tonsor, Alonso-Blanco & Koornneef 2005; Mitchell-Olds & Schmitt 2006; Stinchcombe & Hoekstra 2008; Ungerer, Johnson & Herman 2008; Atwell et al. 2010). This comprehensive view of evolutionary change as a multidisciplinary topic has chiefly been enhanced by important recent advances in the fields of genomics and bioinformatics (Borevitz & Nordborg 2003; Hudson 2008). Using new approaches, we are now able to identify genes and the underlying genetic and molecular mechanisms contributing to natural variation in ecologically and evolutionarily important traits. However, the selected model system plays a central role in avoiding methodological dead ends.

Arabidopsis thaliana, the mouse-ear cress, has been considered to be one of the most suitable genetic plant model systems for decades due to its well-known genetic features (see Meyerowitz 1987; Meinke et al. 1998; Weigel 2012). Recently, ambitious initiatives, such as the 1001 Genome Project, are generating genome-wide data from world-wide natural A. thaliana genotypes (Cao et al. 2011; Schneeberger et al. 2011), providing tools to identify genes and their regulation patterns in different environments. Arabidopsis thaliana may also represent an outstanding ecological and evolutionary model system because the species inhabits a wide range of anthropic and wild environments (Sharbel, Haubold & Mitchell-Olds 2000; Nordborg et al. 2005; Ostrowski et al. 2006; Schmid et al. 2006; Beck, Schmuths & Schaal 2008; François et al. 2008; Picóet al. 2008; Platt et al. 2010; Yin et al. 2010; Weigel 2012). In addition, there is ample evidence that adaptation through natural selection accounts for the species’ diverse geographic and environmental distribution from reciprocal transplants (Donohue et al. 2005a,b; Wilczek et al. 2009; Fournier-Level et al. 2011), common garden (Weinig, Stinchcombe & Schmitt 2003; Griffith, Kim & Donohue 2004; Donohue et al. 2005a,b,c; Malmberg et al. 2005; Huang et al. 2010), growth chamber experiments (Pigliucci, Pollard & Cruzan 2003; Li et al. 2006; Banta et al. 2007; Boyd et al. 2007; Kover et al. 2009; Mendez-Vigo et al. 2011; Montesinos-Navarro et al. 2011) and genome-wide scans (Hancock et al. 2011). Finally, we are increasing our knowledge of the species’ genetic geographic structure and historical demographic processes inferred from patterns of neutral genetic variation (Bakker et al. 2006; Picóet al. 2008; Lundemo, Falahati-Anbaran & Stenøien 2009; Bomblies et al. 2010; Cao et al. 2011; Gomaa et al. 2011), providing insight into the evolutionary background of study populations.

However, we still lack a very important piece of knowledge to deepen our understanding about adaptive variation in A. thaliana: the clear identification of ecologically and evolutionarily important traits (but see Wilczek et al. 2009; Fournier-Level et al. 2011), which are traits that significantly contribute to population fitness. Studies aiming to unravel the genetic and molecular bases of A. thaliana traits focus on a myriad of morphological, physiological and phenological traits, but it is very difficult to ascertain the adaptive value of all these traits. Along with large-scale reciprocal transplant experiments (Wilczek et al. 2009; Fournier-Level et al. 2011), the study of A. thaliana’s population ecology in its natural environments can definitively bridge this gap of knowledge. Several important aspects of the species’ natural history are widely assumed but quite unexplored, the life cycle being one of the most notorious examples. It is widely accepted that A. thaliana predominantly behaves like a winter annual: the plant germinates in autumn, overwinters as a vegetative rosette and flowers in early spring (Ratcliffe 1965, 1976; Baskin & Baskin 1983; Callahan & Pigliucci 2002; Griffith, Kim & Donohue 2004; Shimizu & Purugganan 2005; Montesinos et al. 2009). However, it has also been reported that the plant can also behave like a spring annual: the plant overwinters as seed and germinates, grows and flowers in spring (Donohue 2002 and references therein; Montesinos et al. 2009). At present, we still ignore the extent of among- and within-population life cycle variation (but see Montesinos et al. 2009), the reproductive success of plants exhibiting winter and spring annual life cycles (but see Donohue 2002) and the environmental and ecological factors that account for life cycle variation in the wild.

The most relevant difference between A. thaliana’s winter and spring annual life cycles lies in the extent of delayed germination. Theory states that delayed germination in annual plants can be regarded as a bet-hedging adaptation to unpredictable environmental conditions (Cohen 1966; Venable & Lawlor 1980; Philippi 1993; Venable 2007). However, the body of the theory has been developed for desert annuals occurring in unpredictable desert conditions, and delayed germination refers to germination spread out over more than 1 year (Philippi 1993). In the case of A. thaliana, delayed germination would apply to the seed fraction postponing germination until late winter or early spring to prevent overwinter mortality. In fact, high overwinter mortality of A. thaliana plants growing at high elevations has already been recorded (Montesinos et al. 2009). Hence, harsh winter conditions would be one of the main ecological forces shaping seed germination patterns. Recently, it has been shown using molecular markers that high-elevation subalpine A. thaliana populations exhibited lower effective population size than low-elevation Mediterranean populations in NE Spain, which was interpreted as the net effect of higher genetic drift due to demographic bottlenecks and overall higher demographic instability at high-elevation populations (Gomaa et al. 2011).

Here, I quantify the extent of life cycle variation in A. thaliana populations and evaluate the demographic importance of winter and spring annual life cycles for population performance. To this end, I focus on survival and fecundity schedules of A. thaliana plants that germinated in different seasons. I monitored periodically, from September to June over 5 years (2007–2011), thousands of individual plants from eight natural A. thaliana populations along an altitudinal gradient in NE Spain with contrasting environmental conditions (Montesinos et al. 2009; Gomaa et al. 2011). This survey scheme allowed the identification of individuals that germinated before and after the winter and that are autumn- and spring-germinated annual plants, respectively. Individual survival and fecundity data for winter and spring cohorts were used to compute the net reproductive rate, a surrogate of population fitness (Caswell 2001, 2011). This study provides evidence for the understanding of A. thaliana’s population behaviour in natural environments and the relationship between environmental variation and the species’ life cycle variation along an altitudinal gradient. This is essential for detecting ecologically and evolutionarily important traits across different environmental conditions, which can become the subject of study to investigate the genetic and molecular basis of adaptive variation in plants.

Materials and methods

Study organism

The annual Arabidopsis thaliana (L.) Heyhn. (Brassicaceae) is a cosmopolitan herb native to Europe (Hoffmann 2002; Platt et al. 2010). The plant is a poor competitor and occurs in a variety of anthropic and wild habitats (Picóet al. 2008; Lundemo, Falahati-Anbaran & Stenøien 2009; Bomblies et al. 2010). Populations of this self-compatible and self-fertile plant (Jones 1971) are assemblages of different stands that vary in area, number of plants and distance from other stands (Bomblies et al. 2010; Gomaa et al. 2011). Arabidopsis thaliana has a persistent seed bank with an estimated half-life of approximately 3 years (Montesinos et al. 2009). Genetically, A. thaliana populations are composed of related high- and low-frequency genotypes that in turn are strongly differentiated from other genotypes from other populations (Stenøien et al. 2005; Bakker et al. 2006; : Picóet al. 2008; Lundemo, Falahati-Anbaran & Stenøien 2009; Bomblies et al. 2010; Gomaa et al. 2011). As a result, each A.  thaliana population can be regarded as an independent unit with its own demographic history determined by founder events and subsequent expansion within the population.

Study sites

I selected a total of eight natural A. thaliana populations along an altitudinal climatic gradient in NE Spain (Fig. 1). Population names and their codes are the following: Barcelona (BAR; altitude, 429 m.a.s.l.), Poblet (POB; 656 m), Mura (MUR; 836 m), Vilanova de Meià (VDM; 975 m), Albet (ALE; 1225 m), Pallerols del Cantó (PAL; 1433 m), Bisaurri (BIS; 1450 m) and Vielha (VIE; 1620 m). Distance between populations ranged between 7.5 and 176.7 km. This set of populations is characterized by Mediterranean and subalpine climates for coastal low-elevation and montane high-elevation populations, respectively. A detailed description of vegetation type, soil type and general weather conditions for each population can be found elsewhere (Montesinos et al. 2009). All populations showed the following general demographic patterns: the highest germination peak occurs in October, the onset of flowering is in March/April and seed dispersal takes place between May and June (see Montesinos et al. 2009). During the 5 years of study, I never observed A. thaliana plants departing from this general pattern. This previous basic demographic knowledge was used to set appropriate timings for periodic demographic surveys.

Figure 1.

 Map showing the location of Arabidopsis thaliana populations of study in NE Spain. The study area is encircled in the map of Europe. Figure adapted from Montesinos et al. (2009).

Demographic surveys

In spring 2007, at each population, I selected a representative stand with fruiting individuals within each population and laid down a 4-m permanent transect (Fig. S1) including 18 plots (20 × 20 cm2 divided into 5 × 5 cm2 grid cells). Plots were divided in groups of three that were placed in a line perpendicular to the main transect and separated by a distance of 10 cm. In total, there were six groups of three plots each every 80 cm along the main transect. In autumn 2008, a second identical transect (1–7 m distant from the first transect) was laid down in a different stand except in PAL, ALE and VDM, where it was difficult to find another suitable stand to lay down the second transect.

I surveyed all populations 6–7 times per year in mid-September, late October, late December, early March, early April, mid-May and early June. Sample intervals were shorter in spring because flowering and fruiting data require more intense surveys for precise data collection. In addition, the life cycle in spring can be so fast that individuals can germinate, grow and reproduce in a month (F. X. Picó, personal observation). In autumn and winter, biological activity slows down so sample intervals can become longer with no significant loss of relevant information. In total, I conducted 28 surveys over four complete field seasons spanning 5 years: spring 2007, autumn 2007–spring 2008, autumn 2008–spring 2009, autumn 2009–spring 2010 and autumn 2010–spring 2011. At every survey, I counted, mapped and tagged with colour-headed pins (36 mm long) all individuals observed within the plots (Fig. S2). The smallest individuals I included in the surveys were very young plants bearing cotyledons and the first true leaves (c. 5 mm plant diameter). In autumn, I estimated that these individuals were 2 or 3 weeks old. I normally did not take into account younger seedlings (i.e. plants bearing cotyledons only) because it was quite easy to confound them with seedlings from other plants species at that early developmental stage in the field. However, the number of A. thaliana seedlings excluded from the data set was low (<1%) and unlikely to significantly bias the demographic patterns found in this study. When individuals occurred in dense clumps, particularly in autumn when germination peaked, I counted and tagged individual plants as well as clumps. In spring, I also counted the total number of fruits from all surviving individuals. The survey protocol described here was applied in the last three field seasons (2008–2009, 2009–2010 and 2010–2011). During the first field season (2007–2008), the survey protocol was slightly different: I used the same 20 × 20 cm2 plots but without the 5 × 5 cm2 grid cells, iron nails with coloured wire instead of colour-headed pins, and I mostly tagged clumps that were also mapped. Although the accuracy of the sampling procedure for the first field season was poorer than that applied in the rest of field seasons, the consistency of the results among field seasons gives confidence about the robustness of the methodology. Over the whole study period, I did not observe severe damage to individuals due to the insect activity sometimes observed on A. thaliana, such as sap-sucker aphids, flowering stalks destroyed by moth larvae or leaf mining or fruit depredation by fly larvae. Hence, the demographic behaviour, in terms of recruitment and survival, of monitored individuals was mainly determined by variation in environmental conditions over the whole study period.

This survey scheme allowed the identification of cohorts of autumn- and spring-germinated A. thaliana plants. It must be emphasized that any individual in a population may potentially exhibit both demographic behaviours. A pilot study to assess genetic differentiation between autumn- and spring-germinated A. thaliana plants in different study populations showed that some individuals exhibiting winter and spring annual cycles shared the same genotype (F. X. Picó & C. Alonso-Blanco, unpublished data) estimated with neutral genome-wide SNP markers (Gomaa et al. 2011). Hence, variation in germination timing within each A. thaliana population of study, which determines winter and spring life cycles, is likely to be strongly determined by variation in environmental factors in these populations.

Statistical analyses

Demographic data to evaluate survival and fecundity schedules of autumn- and spring-germinated A. thaliana plants were obtained from two permanent transects. I pooled data from the two transects within each population, which generally showed similar behaviour (Fig. S3) to increase sample size and statistical power of analyses. Surveys clearly allowed the identification of two main cohorts of individuals: those that germinated before (i.e. winter annuals) and after (i.e. spring annuals) January, which is the coldest month during which demographic activity remains very low in all study populations. Hence, winter and spring cohorts of A. thaliana plants are represented by a mixture of individuals recruited between September and December, and between January and April, respectively. Most of the winter and spring individuals were recruited in October/November and March/April, respectively.

For each population and field season, I calculated the percentage survival for each cohort. The effects of field season, population and cohort on the fate of A. thaliana plants were tested with a log-linear analysis. The analysis started with the null model YPC, F, which indicates that field season (Y), population (P) and cohort (C) had an independent effect on fate (F). New models were created to test the effect of each factor, and interactions between factors, on fate. The difference in unexplained variance between each new model and the null model follows a chi-square distribution, with the number of degrees of freedom equal to the difference between model terms.

I estimated fecundity as the total number of fruits produced per individual plant because flower abortion rates (overall mean ± SE = 3.24 ± 0.63%) and fruit depredation rates (0.96 ± 0.39%) were low over the whole study period. Statistical differences between winter and spring cohorts for fecundity were analysed with Student’s t-tests that were performed with the total number of fruits of individual reproductive plants from each cohort in spring.

For each cohort from each population and field season, I estimated the net reproductive rate (R0) as the product of survival and mean fecundity following classical life table theory (Caswell 2001, 2011). I used the number of fruits per individual instead of the number of seeds per individual to compute R0. The main reason is that, although the number of fruits and the number of seeds per fruit are significantly positively correlated in these populations (Montesinos et al. 2009), I ignore the year-to-year variation in this correlation, which might bias the results. I computed R0 for a total of 18 matched pairs of field seasons with data from both winter and spring cohorts. I analysed the statistical differences in R0 between winter and spring cohorts using paired t-tests by pooling data from all populations and field seasons.

I computed the mean proportion of observed and surviving A. thaliana plants that were autumn- and spring-germinated plants, the mean net reproductive rate of winter and spring cohorts, and the mean percentage survival and mean fruit production of winter and spring cohorts over all field seasons available for each population. I analysed the relationship between altitude and these demographic parameters with linear regression models.

For parametric tests, I did not conduct any transformation on demographic parameters because inspection of residuals showed that the assumptions of analyses, mainly homoscedasticity, were met. The log-linear analysis, Student’s t-tests, paired t-tests and linear regression models were performed using spss v.17 statistical software (SPSS Inc., Chicago, IL, USA).


I monitored a total of 12 013 A. thaliana plants over the whole study (2007–2011; Table 1). Seasonal and yearly fluctuations in plant numbers were very high in almost all populations, varying from zero to hundreds or even thousands (Table 1). In general, there were more individuals before (mean ± SE across populations and field seasons = 275.31 ± 115.89; Fig. 2a) than after winter (51.45 ± 10.16; Fig. 2a). Only populations at elevations above 1400 m (VIE, BIS and PAL) had some years with more individuals from spring than from winter cohorts (Table 1). When considering surviving individuals (i.e. surviving until reproduction) only, spring cohorts (mean ± SE across populations and field seasons = 36.03 ± 8.02) were more numerous than winter cohorts (20.35 ± 6.34), except in some years for some populations below 1000 m (BAR, POB, MUR and VDM; Table 1).

Table 1. Arabidopsis thaliana plants from winter and spring cohorts monitored for each population and year of study (overall = 12 013 individuals). For each cohort, the total number and the number of surviving individuals until reproduction are given. The number of individuals observed during the first survey in spring 2007 when the first of the study transects was laid down is also given. Data from the two transects of study were pooled. Data were not available for BIS in spring 2011. Transect 1 of POB was vandalized in winter 2010, and data for this population and field season refer to transect 2 only. Altitude (m.a.s.l.) for each population is given in parentheses next to population acronyms
VIE (1620 m)613/336/326/218/1862/2960/5534/188/76
BIS (1450 m)96251/168/58239/70140/12718/3161/118143/−−/−
PAL (1433 m)7421345/0115/40/−0/−48/1791/6790/449/41
ALE (1225 m)101128/00/−4/00/−48/336/1248/15/0
VDM (975 m)188713/13315/1270/116/252/729/1437/013/3
MUR (836 m)484/03/349/00/−102/1225/118/00/−
POB (656 m)259633/6036/36105/3015/8221/39178/14518/229/21
BAR (429 m)113551/93119/59318/10075/633457/20175/1325/00/−
Figure 2.

 Overall mean (±SE) values for autumn- and spring-germinated cohorts of Arabidopsis thaliana for different traits: (a) total number of plants, (b) percentage survival until reproduction, (c) mean number of fruits per individual and (d) net reproductive rate. Means (±SE) were computed by pooling values from winter and spring cohorts for all different populations and field seasons.

Survival varied significantly among field seasons, populations and cohorts (Table 2). Importantly, for this study, overall survival of spring cohorts (65.38 ± 6.27%; Fig. 2b) was fourfold higher than that of winter cohorts (14.63 ± 3.87%; Fig. 2b). Survival of spring cohorts varied between 16.65 ± 23.55% and 91.75 ± 3.41% among populations, and between 53.12 ± 19.59% and 72.47 ± 14.23% among field seasons. Survival of winter cohorts varied between 2.10 ± 1.71% and 45.75 ± 23.42% among populations, and between 2.93 ± 1.63% and 18.19 ± 12.84% among field seasons. The effect of all interactions between main factors on survival was also significant (Table 2), indicating that survival probability is affected by multiple sources of variation that interact in a complex way.

Table 2.   Log-linear analysis of the effects of field season (Y; 2007–2008, 2008–2009, 2009–2010 and 2010–2011), population (P; VIE, BIS, PAL, ALE, VDM, MUR, POB and BAR), cohort (C; winter and spring) and fate (F; dead and surviving until reproduction) in Arabidopsis thaliana. All models were tested against the reference model given by YPC, F (d.f. = 115 and χ= 38325.04). The difference in the degrees of freedom (Δd.f.) and chi-square values (Δχ2) between each model and the reference model are indicated
ModelsEffectΔd.f.Δχ2 P-value
YPC, YFField season (Y)36135.29<0.0001
YPC, PFPopulation (P)76762.67<0.0001
YPC, CFCohort (C)19667.14<0.0001
YPC, YF, PFY × P1030120.83<0.0001
YPC, YF, CFY × C61585.31<0.0001
YPC, PF, CFP × C822182.69<0.0001
YPC, YF, PF, CFY × P × C1167656.76<0.0001

Winter and spring cohorts did not significantly differ from each other for mean number of fruits in almost all cases (Table 3). Only three t-tests yielded significant P-values for the mean number of fruits: individuals from the spring cohort produced more fruits than those from the winter cohorts in BAR in 2007–2008, whereas the opposite was found in VIE and POB in 2009–2010 (Table 3). Overall, winter and spring cohorts produced on average (±SE) 4.43 ± 0.72 and 4.21 ± 0.58 fruits per plant, respectively (Fig. 2c).

Table 3.   Mean (±SE) number of fruits produced by Arabidopsis thaliana plants from winter and spring cohorts for each population and year of study. Altitude (m.a.s.l.) for each population is given in parentheses next to population acronyms. Dashes indicate that fecundity data were not available due to the lack of individuals surviving until reproduction. Significant comparisons (< 0.05; t-test) between winter and spring cohorts for the total number of fruits are indicated in bold face
  1. *Number of fruits from a unique surviving plant.

VIE (1620 m)12.0 ± 0.78.5 ± 1.67.0 ± 7.13.6 ± 0.5 4.0 ± 0.7 2.8 ± 0.2 20*6.0 ± 0.7
BIS (1450 m)26*9.2 ± 1.03.9 ± 0.23.7 ± 0.21.7 ± 0.84.8 ± 0.3
PAL (1433 m)1.8 ± 0.63.7 ± 0.63.3 ± 0.34.8 ± 2.02.8 ± 0.3
ALE (1225 m)3.3 ± 0.85.9 ± 1.01*
VDM (975 m)2.5 ± 0.22.9 ± 0.81*1.5 ± 0.71.9 ± 0.43.6 ± 0.81.7 ± 0.8
MUR (836 m)2.7 ± 0.85.5 ± 1.34.8 ± 1.5
POB (656 m)1.5 ± 0.11.3 ± 0.12.9 ± 0.53.0 ± 1.2 5.6 ± 0.8 1.8 ± 0.1 2.0 ± 0.02.0 ± 0.3
BAR (429 m) 5.0 ± 1.3 10.6 ± 2.1 10.6 ± 1.210.0 ± 1.31.9 ± 0.32.7 ± 0.2

Net reproductive rates (R0) differed significantly between winter and spring cohorts (t1,17 = 6.04, < 0.0001). All R0 values for spring cohorts (range R0 values = 0.19–8.40) were higher than those for winter cohorts (0.01–3.34). On average (±SE), R0 for winter and spring cohorts was 0.84 ± 0.21 and 2.67 ± 0.43, respectively (Fig. 2d).

Linear regressions between altitude and mean proportion of observed and surviving plants from winter and spring cohorts were significant (Fig. 3a,b). Only the population located at the highest altitude, VIE, had more observed plants from spring than from winter cohorts: 72% of observed plants over all the field seasons were spring-germinated individuals (Fig. 3a). The proportion of observed spring-germinated plants in the rest of populations was below 50% (range, 14–49%; Fig. 3a). In contrast, the four high-elevation populations exhibited on average more than 80% of surviving plants over the field season that were spring-germinated plants (range, 80–87%; Fig. 3b), whereas the other four populations at lower elevations had a lower proportion of surviving spring-germinated plants (range, 37–57%; Fig. 3b). Linear regressions between altitude and fecundity from winter and spring cohorts were not significant (> 0.14). Finally, linear regressions between altitude and R0 for winter and spring cohorts were not significant either (Fig. 3c,d).

Figure 3.

 Linear regressions between altitude and (a) mean proportion of Arabidopsis thaliana plants observed from spring cohorts, (b) mean proportion of plants surviving until reproduction in spring from spring cohorts, (c) net reproductive rate of winter cohorts and (d) net reproductive rate of spring cohorts. The proportion of plants from winter and spring cohorts is inversely correlated and sum to one. Dots represent the mean value across all field seasons for each population of study. When significant, regression coefficients and their respective P-values are also given.


Life cycle variation along an altitudinal gradient

The periodic demographic surveys of Arabidopsis thaliana populations described in this study illustrate how annual plants can adjust their life cycle to the changing environmental conditions encountered along an altitudinal gradient. Overall, the results clearly indicated the progressive shift from a winter to a spring life cycle with increasing altitude in A. thaliana. In general, annual plants tend to be poor competitors and need to colonize empty patches or new habitats, and variation in germination behaviour under different environments has been interpreted as a general strategy to survive across broad geographic ranges (Donohue et al. 2005a,b,c). The results of my study should be applicable to other annual plants because I have also observed similar demographic patterns of autumn- and spring-germinated cohorts in other annual crucifers co-occurring with A. thaliana in the same community of therophytes, such as Erophila verna, Hornungia petraea, and various species of the genera Cardamine and Arabis (F. X. Picó, personal observation).

On the basis of the results of my study on A. thaliana, I hypothesize that the effects of global warming on annual plants could be translated into an increase in the relative importance of winter cohorts in high-elevation populations. Given that A. thaliana, and many other annual plants, exhibit a high plasticity in germination timing, adjustments of life cycle to new environmental conditions are totally feasible. High-elevation A. thaliana populations are characterized by low genetic diversity and a low effective population size due to demographic bottlenecks and overall demographic instability (Gomaa et al. 2011). Hence, a reduction in mortality rates and an increase in autumn-germinated plants due to more favourable weather conditions could lead to an increase in genetic diversity at high elevations in a scenario of global warming. In contrast, I cannot hypothesize about the effects of global warming on ecological performance and genetic makeup of A. thaliana populations at low elevations from the results of this study. However, observations from A. thaliana populations occurring in hot and dry environments, such as the Guadalquivir Valley in S Spain (e.g. Posadas; 37º87′N, 5º11′W; 203 m.a.s.l.), show that A.  thaliana can also exhibit a strict spring life cycle of <2 months of duration (February and March) from seed germination to seed dispersal (F. X. Picó, personal observation). This observation indicates that A. thaliana is also able to adjust its life history in warmer environments.

Demographic behaviour of A. thaliana populations

In the literature, it is widely accepted that A. thaliana predominantly behaves like a winter annual (Ratcliffe 1965, 1976; Baskin & Baskin 1983; Callahan & Pigliucci 2002; Griffith, Kim & Donohue 2004; Shimizu & Purugganan 2005; Montesinos et al. 2009). This fits perfectly with the generally accepted view that annual plants mostly exhibit a typically winter annual life history (Kalisz 1991; McKenna & Houle 2000; Facelli, Chesson & Barnes 2005; Kluth & Bruelheide 2005; Dostál 2007). This would be the case for A.  thaliana if we only took germination patterns observed in this study into account. However, it is clear that A. thaliana can also behave like a spring annual, which would not be unusual because the duality of life cycles in annual plants has long been known in plant population biology (Purvis 1939; Stewart & Hull 1949; Gottlieb 1977; Primack 1979; Masuda & Washitani 1992). In the particular case of A. thaliana, the problem is that we completely lacked information about the spatiotemporal demographic behaviour of natural populations. For this reason, assumptions on A. thaliana’s life cycle have been based on partial knowledge. Given the results of this study, the definition of A. thaliana as a winter annual plant may be questionable if we consider the demographic importance of winter and spring cohorts. Beyond a mere definition, the results presented here stress the need to quantify the demographic contribution of the life cycle events that matter for the maintenance of A. thaliana populations, that is, recruitment, survival, and fecundity, instead of purely describing the timing of such events.

Spring annual behaviour seems to be a fundamental strategy to maintain viable populations in A. thaliana. On average, the net reproductive rate of winter cohorts was less than unity, meaning that populations would irremediably decline if autumn-germinated plants were the only group of individuals contributing to population growth. However, the preponderance of spring annual cycles increased with altitude. This pattern of life cycle variation with altitude is interesting because it might illustrate adaptive variation in life cycle along the altitudinal climatic gradient of study. In fact, previous experimental work on this set of populations already indicated adaptive clinal variation in different life-history traits (Montesinos-Navarro et al. 2011). I conclude that local adaptation in A.  thaliana is a process that may encompass not only architectural and developmental plant traits, but the life cycle as a whole.

What is the main environmental factor that may account for such clinal variation in A. thaliana’s life cycle? It has been shown by experimental means that genotypes from high-altitude populations exhibit faster developmental rates in early stages to accelerate growth and increase plant size before winter, and slower rates in later stages to lengthen the period of growth before flowering as a result of cooler springs at high-elevation locations (Montesinos-Navarro et al. 2011). Hence, temperature regimes during the winter months represent the main candidate abiotic factor shaping life cycle in these A. thaliana populations. I want to emphasize, however, that no significant variation between climatic conditions and demographic behaviour was found in this study (results not shown). Longer temporal series of demographic data would be needed to explore the way in which extrinsic (e.g. climatic conditions) and intrinsic (e.g. population density) factors account for plant fate and population dynamics as a whole.

Implications of results

The study of complex processes, such as adaptive responses to varying environmental conditions in a context of global warming, can only be achieved through a multidisciplinary approach. We need to find the right model systems to work with and coordinate complementary research lines to fully address the problem. Given the genetic knowledge of A. thaliana, understanding A. thaliana’s performance in natural environments allows the study of the genetic and molecular basis of adaptive variation in important life cycle traits. The results of this study suggest that germination timing is a trait of special interest in A. thaliana, which has already been shown to have enormous relevance, given the existing knowledge of its genetic basis (Bentsink & Koornneef 2002; Finch-Savage & Leubner-Metzger 2006; Bentsink et al. 2010; Chiang et al. 2011) and evolutionary potential (Donohue et al. 2005a,b). We need to assess the extent of phenotypic variation at different spatiotemporal scales in many other A. thaliana populations across the species’ distribution range (Metcalf & Mitchell-Olds 2009; Montesinos et al. 2009). The current extraordinary efforts to obtain whole-genome sequence variation for an extensive world-wide collection of A.  thaliana accessions (Cao et al. 2011; Schneeberger et al. 2011; Weigel 2012) should be complemented with equal efforts to generate long-term life cycle and population performance data across multiple environments where A. thaliana occurs.


I am grateful to Adrian C. Brennan, José M. Fedriani and Carlos Alonso-Blanco for their comments on earlier versions of this manuscript. Alicia Montesinos-Navarro, Pedro F. Quintana-Ascencio, Elena Caballero and Stephen J. Tonsor made suggestions and assisted in the field during different years of study. Funding was provided by Ministerio de Ciencia e Innovacion of Spain (Grants CGL2006-09792/BOS and CGL2009-07847/BOS) and CSIC (Grant 200630I255) to F.X.P.

Data accessibility

Data deposited in the Dryad repository: