Volume 35, Issue 5 p. 1145-1158
RESEARCH ARTICLE
Free Access

To grow or survive: Which are the strategies of a perennial grass to face severe seasonal stress?

Thomas Keep

Thomas Keep

INRAE, UR P3F, Lusignan, France

Search for more papers by this author
Jean-Paul Sampoux

Jean-Paul Sampoux

INRAE, UR P3F, Lusignan, France

Search for more papers by this author
Philippe Barre

Philippe Barre

INRAE, UR P3F, Lusignan, France

Search for more papers by this author
José-Luis Blanco-Pastor

José-Luis Blanco-Pastor

INRAE, UR P3F, Lusignan, France

Search for more papers by this author
Klaus J. Dehmer

Klaus J. Dehmer

Leibniz Institute of Plant Genetics and Crop Plant Research (IPK), Malchow/Poel, Germany

Search for more papers by this author
Jean-Louis Durand

Jean-Louis Durand

INRAE, UR P3F, Lusignan, France

Search for more papers by this author
Matt Hegarty

Matt Hegarty

IBERS-Aberystwyth University, Aberystwyth, UK

Search for more papers by this author
Thomas Ledauphin

Thomas Ledauphin

INRAE, UR P3F, Lusignan, France

Search for more papers by this author
Hilde Muylle

Hilde Muylle

Flanders Research Institute for Agriculture, Fisheries and Food (ILVO) - Plant Sciences Unit, Melle, Belgium

Search for more papers by this author
Isabel Roldán-Ruiz

Isabel Roldán-Ruiz

Flanders Research Institute for Agriculture, Fisheries and Food (ILVO) - Plant Sciences Unit, Melle, Belgium

Search for more papers by this author
Tom Ruttink

Tom Ruttink

Flanders Research Institute for Agriculture, Fisheries and Food (ILVO) - Plant Sciences Unit, Melle, Belgium

Search for more papers by this author
Fabien Surault

Fabien Surault

INRAE, UR P3F, Lusignan, France

Search for more papers by this author
Evelin Willner

Evelin Willner

Leibniz Institute of Plant Genetics and Crop Plant Research (IPK), Malchow/Poel, Germany

Search for more papers by this author
Florence Volaire

Corresponding Author

Florence Volaire

CEFE, Univ Montpellier, CNRS, EPHE, IRD, Université Paul Valéry Montpellier 3, INRAE, Montpellier, France

Correspondence

Florence Volaire

Email: [email protected]

Search for more papers by this author
First published: 03 February 2021
Citations: 20

Funding information

T. Keep received the support of the Nouvelle-Aquitaine region (50% of funding) and the Institut National de la Recherche Agronomique INRAE (50% of funding) in France. This work was funded in the frame of the project GrassLandscape awarded by the 2014 FACCE-JPI ERA-NET + call Climate Smart Agriculture. Funding was granted by the EC (grant agreement nº 618105), the Agence Nationale de la Recherche (ANR) and INRAEe (métaprogramme ACCAF) in France, the Biotechnology and Biological Sciences Research Council (BBSRC) in the United Kingdom and the Bundesantalt für Landwirtschaft und Ernährung (BLE) in Germany.

Handling Editor: Katherine McCulloh

Abstract

  1. More severe seasonal stresses resulting from climate change affect the survival of perennial plant species. The growth–survival trade-off exemplified in dormant species is a key issue to understand adaptation. As the validity of this trade-off has yet to be tested in non-dormant species, it was assessed by exploring the intraspecific variability of strategies to face drought and frost within perennial ryegrass.
  2. Three common gardens compared 385 European perennial ryegrass populations along a latitudinal environmental gradient over 3-years. Persistence, productivity and physiological traits were recorded under contrasting seasonal environments.
  3. Decoupling plant responses, that is, growth under favourable summers/winters and plant survival under harsh summers/winters, showed a general trade-off between growth potential and dehydration survival. Three groups of perennial ryegrass populations were identified according to their contrasting strategies: (a) year-round productive but stress sensitive populations from wet areas; (b) drought-tolerant populations with low summer growth potential from drought-prone areas and (c) frost-tolerant populations with low winter growth potential from frost-prone areas. Overall, the populations surviving drought best were more resource conservative, whereas populations of the other groups were more resource acquisitive. However, such overall functional patterns were less meaningful than seasonal variations of resource acquisition potentials. The predicted potential biogeographical distribution of these groups suggests shifts of areas of suitability under climate change over the next decades in Europe. Dehydration escape and dehydration tolerance through reduction of growth potential in summer may become the strategies best adapted to an increasingly large area of Europe.
  4. The large intraspecific variability of phenological adaptations within perennial ryegrass reveals that the seasonal modulation of growth potential is crucial to plant adaptation under severe chronic abiotic stresses. The global plant economics spectrum cannot account for contrasting seasonal trade-offs, which points out the importance of integrating phenological traits as key components of plant strategies. The identification of the trade-off between growth potential and frost or drought stress survival in this non-dormant species provides key knowledge to understand the future regional distribution of this major species for grassland ecosystem services.

A free Plain Language Summary can be found within the Supporting Information of this article.

1 INTRODUCTION

Climate models project an increase in probability of extreme weather such as more intense droughts under global warming scenarios (IPCC, 2018). Extreme climatic events are expected to impact ecosystem functioning and species turnover (Smith, 2011) since they can affect plant resilience, that is, their ability to recover following severe stress (Hodgson et al., 2015). In both ecology and agronomy, understanding community resilience and plant adaptive strategies is a timely research avenue (Kooyers, 2015).

A general pattern of functional strategies proposes a fast–slow continuum from plants with a high resource uptake optimizing growth in rich environments, to plants with a low resource uptake optimizing survival in poorer environments (Reich, 2014). The ability of a plant to survive harsh environmental conditions therefore implies a cost in terms of growth potential (Stearns, 1989). This growth–survival trade-off is a central ecological concept showing that organisms have limited options (Sibly & Calow, 1989) in response to drought (Benavides et al., 2015; Volaire et al., 2014) and frost (Koehler et al., 2012).

Plant adaptation to drought involves a few major strategies, such as dehydration escape, dehydration avoidance and dehydration tolerance including summer endo-dormancy (Ludlow, 1989; Volaire, 2018). These strategies parallel the ecological trade-off between an active high growth rate under moderate drought (dehydration avoidance) and a low growth rate and high survival (dehydration tolerance) under severe stress. The optimization of the costs–benefits balance between fast growth and plant survival has been demonstrated in species with seasonal endo-dormancy enhancing survival under harsh winters or summers (Gillespie & Volaire, 2017; Lang, 1987). Interestingly, frost and drought stresses may lead to similar plant adaptive responses as both cause cellular dehydration (Kong & Henry, 2018). For instance, under both kinds of stress, a high investment in protective compounds, such as water-soluble carbohydrates in reserve organs (Sanada et al., 2007; Volaire, 1995), enhances stress tolerance but reduces resource allocation to growth in grasses.

Drought and frost survival in the perennial grass Dactylis glomerata was already found to be negatively correlated with growth potential, that is, with endo-dormancy level in summer and winter, respectively (Bristiel et al., 2018). However, the validity of a trade-off between growth potential and seasonal dehydration survival should also be tested on non-dormant species to assess its generality.

Perennial ryegrass (Lolium perenne L.), hereafter termed ‘ryegrass’ is the most economically important pasture grass species in many temperate areas (Blackmore et al., 2016). This cool season C3 grass has a large biogeographical distribution, including regions with both frost and heat stresses, from Anatolia to the Iberian Peninsula and from Northern Africa to Scandinavia (Blanco-Pastor et al., 2019). So far, L. perenne has not been described as summer nor winter dormant (Gillespie & Volaire, 2017) and therefore provides a good model to examine the growth–survival trade-off in a non-dormant species.

This study aimed to explore the possibility that adaptive strategies of ryegrass largely rely on trade-offs between growth and stress survival. Three common gardens across Europe for 3 years tested 385 populations of L. perenne originating from a large biogeographical range (Europe to Middle East). Phenological and ecophysiological strategies (seasonal growth patterns and dehydration escape, avoidance and tolerance) were investigated along with the functional economics spectrum representing resource conservative versus resource acquisitive strategies (Wright et al., 2004). The following questions were addressed. (a) Does the comparison of growth under favourable summers/winters and plant survival under harsh summers/winters confirm the seasonal growth–survival trade-off theory in the intraspecific diversity of the non-dormant species L. perenne? (b) Do distinct functional characteristics exist within the L. perenne diversity and are they coupled with ecophysiological strategies conferring stress survival? (c) Are these strategies associated with climatic conditions at the sites of origin of populations, and if so, can they be used to predict the future potential biogeographical distribution of ryegrass diversity under increasing seasonal stresses induced by changing climate?

2 MATERIALS AND METHODS

2.1 Plant materials and experimental design

This study tested 385 natural populations of ryegrass taken from European genebanks (Figure 1; Appendix S1). They were sown in common gardens in three locations: Poel Island (PO) in Germany (53.990°N, 11.468°E) on 8 April 2015, Melle (ME) in Belgium (50.976°N, 3.780°E) on 2 October 2015 and Lusignan (LU) in France (46.402°N, 0.082°E) on 9 April 2015. In all three locations, 1-m2 plots for three replicates of each population were set in three complete blocks. Plots were regularly fertilized particularly so that differences between sites were mainly climatic and cut to simulate common practice expected to maximize yearly vegetative growth of meadows (Appendix S2).

Details are in the caption following the image
(a) Spatial distribution of sites of origin of the 385 ryegrass natural populations in study and of locations of the three common gardens in which populations were grown. The 1989–2010 norm of isothermality, that is, mean temperature diurnal range over annual temperature range (BIOCLIM derived variable bio3) is displayed as map background. (b) Aerial view of the common garden at LU

2.2 Climate variability in trial sites

From the seasonal weather condition records (Appendix S3), the wettest summer, the driest summers, the mildest winter and the winters with frost-related stress were identified across sites and years. The actual evapotranspiration (AET) of crop species is known to be reduced below potential evapotranspiration (PET) when the relative soil water content (Appendix S3) drops below 40% of field capacity (Allen et al., 1998); drops below 20% have furthermore been shown to induce plant mortality for certain grasses and trees (Caspersen & Kobe, 2001; Poirier et al., 2012). During summers 2016, 2017 and 2018 at LU, the average relative soil water content fell below 20% of field capacity and the experienced drought stress could thus be considered as notable. At PO, no drought was experienced during summers 2016 and 2017 with mean relative soil water content equal to 88% and 97%, respectively, of field capacity. At ME, moderate drought stress occurred, notably during summer 2017 when average relative soil water content fell below 27% of field capacity mainly because of a water deficit during the previous spring. Frost occurrences have been evidenced to induce stress in temperate grass species; for example, 15 frost days in winter were shown to induce mortality in Dactylis glomerata L. and Festuca arundinacea (Poirier et al., 2012). The number of frost days (minimum temperature of −1°C or below) that occurred during the winters varied from 9 to 41 across years and sites. Winters 2015–2016 and 2018–2019 at LU had the smallest number of frost days (9 and 12, respectively) and were considered as mild winters.

2.3 Phenotypic traits

Traits were recorded overall at the plot level (Table S1, Appendix S4).

2.3.1 Responses to seasonal abiotic stresses

Persistence was computed from soil coverage visual scores (1–9 scale) before and after the period of interest. Summer persistence (SumPer) was assessed by changes in soil coverage between end and beginning of the summers 2016 and 2018 at LU (the summers with high water stress, see Appendices S3 and S4 for water stress justification and Figure S1 for pictures of three plots before, during and after the drought at LU 2016). Winter persistence (WinPer) was assessed by changes in soil coverage between end and beginning of the winters 2016–2017 and 2017–2018 at LU, 2016–2017 at ME and 2016–2017 at PO (winters with at least 17 frost days, Appendix S4). Stay-green (STG) visual scores (1–9 scale) were recorded during the dry summer of 2016 at LU to evaluate the ability to maintain photosynthetic activity and delay of leaf senescence under drought stress (Abdelrahman et al., 2017).

2.3.2 Aerial biomass growth in non-stressful conditions (proxy of potential growth)

The assessment of vegetative growth was based on the measurement of micro-sward canopy heights. Canopy height measurement is commonly used as a proxy of grass sward standing biomass with a correlation between the former and the latter nearing 0.8 (Viljanen et al., 2018).

Potential Vegetative spring growth (PVG) was the canopy height measured after a 2-month period of vegetative growth from the start of the wet and warm spring of 2016 at LU. Potential summer growth (PSG) was the canopy height increase over the wet summer of 2016 at ME. Potential winter growth (PWG) was scored just after the mildest winter (2018–2019 at LU). Autumn vegetative growth (AVG was measured after a month of vegetative growth during the wet and mild autumn of 2016 at ME. Potential annual growth (PAG) was the sum of canopy heights measured before each cut during the most favourable site × year (2016 at ME) as an indicator of ‘resource acquisition’ at plant level (instead of its usual proxy at leaf level, the specific leaf area).

2.3.3 Phenology and sexual reproduction

The reproductive phenology was described by the spike emergence date or heading date (HEA) and by the visual score (1–9 scale) of the percentage of plants heading the year of sowing (i.e. without vernalization requirements; HFY). Investment into sexual reproduction was described by the density of fertile stems corrected for the confounding effect of the spike emergence date (DST), by the visual score (1–9 scale) of aftermath heading intensity (AHD) and by the length from ground to spike base of elongated fertile stems (straw height; HST).

2.3.4 Morphology and biochemical traits

Morphology

Average leaf lamina width (LMW) and growth habit (GRH) were visually scored (1–9 scale).

Biochemistry of aerial biomass

Near-infrared reflectance spectroscopy (NIRS) analyses delivered predictions of water-soluble carbohydrates (WSC; Wiseman et al., 1960) and acid detergent lignin (ADL; Van Soest et al., 1991) in aerial biomass. ADL was considered as an indicator of ‘resource conservation’ at plant level instead of its usual proxy at leaf level, the leaf dry matter content which was shown to be inversely correlated with pasture digestibility across a range of grasslands (Pontes et al., 2007).

Chemistry of leaf lamina

In April 2016 at LU, a leaf lamina sample was collected on 30 plants in each plot. The 30 samples were pooled, dried down, ground and analysed by mass spectrometry to predict the nitrogen content of the dry leaf lamina tissue (NLI) and the isotopic signature of 13C (C13). Nitrogen content of sunlit leaves from canopy top is considered an indicator of the fulfilment of plant nitrogen needs for optimal growth (Farruggia et al., 2004). The isotopic signature of 13C (δ13C) is considered as an indicator of photosynthetic efficiency when water supply is not limiting (Condon et al., 2007) and possibly of water use efficiency (Condon et al., 2002).

2.4 Bioclimatic variables

A set of bioclimatic variables inspired by the BIOCLIM variables (https://www.worldclim.org/data/bioclim.html) was used. Fine-resolution grids (0.05° longitude and latitude) were set up over Europe and surroundings of 1989–2010 norms using EURO4M-MESAN (1989–2010) and EUMETSAT CM SAF (1989–2013) data (see Appendix S5). Values of norms at sites of origins of populations were approximated by the values of grid cells containing these sites; those values are available in Table S2.

Spatial grids of predicted norms of the same bioclimatic variables were also computed for the 2041–2070 period using climate data predicted from the implementation of the regional climate model RCA4 (Strandberg et al., 2015) to the IPCC AR5 Representative Concentration Pathway scenario RCP 8.5 (van Vuuren et al., 2011). These grids were computed at same resolution as the 1989–2010 grids.

2.5 Data analysis

Models of analysis of variance (ANOVA) were implemented to assess the significance of the population effect and to compute adjusted means of populations over replicates within trial location and in some cases over trial locations and dates (for more details, see Appendix S4). ANOVAs were performed using the R functions ‘lm’ and ‘ANOVA’ of the R (R Core Team, 2019) ‘car’ library. For all traits, the variance components were used to calculate H2 indicators equal to the proportion of inter-population genetic variance into the inter-population phenotypic variance (Appendix S4). Pearson correlations between variables were computed using the R ‘cor’ function. Linear regressions were performed using the R ‘lm’ function.

A principal component analysis (PCA) of population trait means was performed using the R ‘PCA’ function of the FactoMineR package. Traits were used as variables and populations as observations, populations with missing data for at least one trait were removed and the variables were standardized. The two bioclimatic variables bio(ad)26 (potential evapotranspiration of warmest quarter) and bio12 (cumulative annual precipitation) were added as supplementary variables (values at sites of origin of populations) as they were the bioclimatic variables best correlated to the first and second principal components, respectively.

We partitioned the populations into groups by implementing a hierarchical clustering on principal components (HCPC) with the R ‘HCPC’ function of the FactoMineR package using default parameters and without any a priori number of clusters. Variables best discriminating the groups identified by HCPC were determined by linear discriminant analyses (LDA). LDA was implemented using the ‘discrimin’ function of the R ade4 package on bioclimatic variables.

The potential climatic distribution of the HCPC groups was modelled using a multinomial log-linear regression model. To choose the most reliable explanatory bioclimatic variables and to avoid overfitting and redundancy, the best set of bioclimatic variables to include in the model was beforehand determined by implementing a multinomial model fitted via penalized maximum likelihood using 10-fold cross-validation. This was undertaken with the ‘cv.glmnet’ R function of the glmnet package using default parameters except the cross-validation option which was set to ‘class’. We retained the bioclimatic variables included in the model presenting the largest value of the penalty multiplier lambda and for which the error departed from less than one standard deviation from the minimum mean cross-validated error observed over all models. The final distribution model was fitted using a multinomial log-linear model implemented with the ‘multinom’ function of the nnet r package. The fitted multinomial model was used to project the probability of presence of groups onto fine-resolution spatial grids (0.05° longitude–latitude) of climatic data. Projections were performed on grids of near-past (1989–2010) climatic data, as well as of climatic data predicted for the 2041–2070 period. Projections were restricted to grid cells within the range of values used to fit the model.

2.6 Data accessibility

All data required to replicate the presented results are available in the associated Supporting Information.

3 RESULTS

H2 indicators varied from 0.41 for NLI to 0.99 for HEA. This indicated that a high proportion of inter-population phenotypic variation was due to inter-population genetic variation.

3.1 Overall intraspecific variability

The first two dimensions of the PCA of population traits represented 52% of the total inertia (Figure 2a). Neither of the first two principal components was significantly associated with population collection year or to seed multiplication site which also supports the idea that phenotypic variation was largely due to genetic variation. The first principal component was negatively correlated to WinPer and autumn sugar content (WSC) and positively to winter growth (PWG), reproductive investment (DST and AHD) and autumn lignin content (ADL). The second principal component was associated with strong vegetative growth (PVG, PAG and PSG), vigorous autumn re-growth (AVG), wide leaves (LMW), high flowering stems (HST), erect growth habit (GRH), early flowering populations (HEA) and was negatively correlated to SumPer (Figure 2a). The correlations of summer PET with the first and second principal components equalled 0.73 and −0.13, respectively, and that of the cumulated annual precipitation equalled −0.06 and 0.28, respectively.

Details are in the caption following the image
Principal component analysis using the ryegrass populations as observations and phenotypic traits as input variables. Two bioclimatic variables, describing climate at sites of origin of populations, were included as supplementary variables: bio12 the 1989–2010 norm of cumulated precipitation of the wettest month and bio( ad) 26 the 1989–2010 norm of cumulated potential evapotranspiration during the warmest quarter. (a) Correlation circle of traits (see Section 2 for trait acronyms) and supplementary bioclimatic variables. (b) Scores of populations on the first two principal axes; the three groups determined by the HCPC analysis on traits are displayed with different colours, G1 in blue, G2 in green and G3 in red. (c) Spatial distribution throughout Europe of population groups with same colour code as in (b)

The HCPC analysis partitioned the populations in three groups (Figure 2b). The G1 group incorporated early spike emergence (HEA) and erect habit (GRH) populations with relatively low summer (SumPer) and winter (WinPer) persistence, high growth potential (PVG, PAG, PWG and PSG), wide leaves (LMW) and long flowering stems (HST). The G2 group incorporated late flowering (HEA) and prostrate (GRH) populations with high winter persistence (WinPer), limited summer persistence (SumPer), low reproductive investment (DST and AHD), low winter growth (PWG) and narrow leaves (LMW). The G3 group incorporated populations with relatively high SumPer, low WinPer, high reproductive investment (AHD and DST), low growth potential (PAG), high lamina 13C signature (C13) and high lignin content (ADL). See Table S3 for mean trait values of each group. LDA implemented with the bioclimatic variables (Table S4) revealed that the populations from the G3 group were mainly from warm summer areas as opposed to those from the G2 group. Populations from the G1 group were mainly from wet areas of Europe. Those trends are confirmed by the spatial distribution of these three groups (Figure 2c).

3.2 Summer and winter persistence

Populations with high growth rate during low water stress summer (PSG) were the most affected by plant mortality (low SumPer) during relatively high water stress summers (Figure 3a). Likewise, populations with high winter growth potential in mild winters (high PWG) were generally more sensitive to winter hardships (low WinPer) than populations with low winter growth potential (Figure 3b). Furthermore, SumPer was negatively correlated to WinPer (Figure 3c).

Details are in the caption following the image
Scatter plots of (a) summer persistence (SumPer) against potential summer growth (PSG) and (b) winter persistence (WinPer) against potential winter growth (PWG) and (c) scatter plot of summer persistence (SumPer) against winter persistence (WinPer). See Appendix S4 for units of traits. Each dot represents one ryegrass population. Pearson correlations and their p values are displayed as well as regression lines. Dot colours are according to population groups: G1 (blue), G2 (green) and G3 (red)

SumPer was negatively associated with potential annual vegetative growth (PAG; Figure 4a), whereas WinPer was positively associated with this trait (Figure 4b). Autumn leaf lignin content was tightly negatively associated with potential annual vegetative growth (Figure 4c), suggesting that slow growing populations had a greater cell wall lignification and therefore a lower digestibility. Furthermore, the least productive populations tended to have higher leaf lamina δ13C (i.e. lower 13C discrimination; Figure 4d).

Details are in the caption following the image
Scatter plots of (a) summer persistence (SumPer), (b) winter persistence (WinPer), (c) lignin content in autumn aerial biomass (ADL), (d) spring sunlit leaf lamina δ13C (C13) against potential annual growth (PAG). (e) Scatter plot of summer persistence (SumPer) against stay-green score in summer (STG) in which the brown line is the regression line taking into account only populations from groups G1 and G2 and the black dashed line is the one taking into account all populations. (f) Scatter plot of winter persistence (WinPer) against water-soluble carbohydrate content in autumn aerial biomass (WSC). See Appendix S4 for units of traits. Each dot represents one ryegrass population. Pearson correlations and their p values are displayed as well as regression lines. Dot colours are according to population groups: G1 (blue), G2 (green) and G3 (red)

The stay-green in summer (STG) was poorly correlated with SumPer when considering all populations (Figure 4e). However, considering only populations from groups G1 and G2, SumPer became quite strongly positively correlated to STG score (Figure 4e). The populations with the highest STG scores were mainly from North-Eastern Europe. Water-soluble carbohydrate content at autumn cut (WSC) was positively correlated to WinPer (Figure 4f).

3.3 Potential climatic distribution of the ryegrass groups

The bioclimatic variables retained in the optimized multinomial model that best predicted the climatic distribution of the three groups were summer PET, precipitation seasonality, precipitation of wettest 14–15 days period and mean diurnal temperature range. The average misclassification error rate across 50 iterations of random 10-fold cross-validation was of 28% and varied from 15% to 40%. Figure 5 displays the spatial projections of the models fitted for the three groups for near-past (1989–2010) climate and foreseen future climatic conditions (2041–2070, RCP 8.5/RCA4). The spatial areas predicted as potentially suitable with regard to climate for each of the three groups shift dramatically from the 1989–2010 period to the 2041–2070 period. The area with high probability of suitability for the G3 group widely expands northwards in central Europe. Climatic conditions in large areas of the Mediterranean rim and the Balkans shift out of the model range due to extremely high summer temperatures and low precipitation, which were beyond the extremes of the 1989–2010 data used to fit the model. The area with high probability of suitability for the G2 group (matching with frequent winter frost areas) notably moves northwards. The area with high probability of suitability for the G1 group (matching with high rainfall areas) contracts in the Southern regions (Italy, the Balkans) and Eastern-Europe, but it expands in more Northern regions from Germany to Sweden. Regions that were too cold to be within the range of any model in the 1989–2010 period, such as regions of Norway and Scotland, appear to become suitable for populations of the G2 or G1 groups in the 2041–2070 period.

Details are in the caption following the image
Spatial distribution of the probability of presence of the three groups of ryegrass. (a, c, e) spatial distribution predicted from 1989 to 2010 climatic norms; (b, d, f) spatial distribution predicted from 2041 to 2070 norms expected from the implementation of the RCA4 climate model to the IPCC AR5 RCP 8.5 scenario. Colour of areas varies from white to dark shade of either blue, green or red following an increase in the probability of presence of the G1 (plots a and b), G2 (plots c and d) and G3 (plots e and f) groups, respectively. Yellow: areas with climatic data out of the calibration range of the multinomial models. Light brown: areas out of the range of used climatic grids

4 DISCUSSION

This study spanning several years and sites allowed the comparison of ryegrass populations under contrasting seasonal environments. Their growth potential under favourable seasonal conditions was compared to persistence/mortality under severe drought or frost to identify the intraspecific variability of ecological strategies (Moran et al., 2016).

4.1 An intraspecific trade-off between growth potential and dehydration survival

This study showed negative relationships between the potential of ryegrass populations to grow during either a wet summer or a mild winter and the ability to survive and persist during the same seasons when drought or frost was severe enough to kill tillers. Based on intraspecific diversity, the hypothesis of seasonal growth versus stress survival trade-offs for a species known to exhibit neither summer nor winter dormancy was thus confirmed. This ‘field-based’ study illustrates that the dehydration tolerance mechanisms impose a cost for resource acquisition to fuel growth. It is likely that the strength of the trade-off could have been greater if the experiment had involved even more contrasting environmental conditions (e.g. Mediterranean site, rain manipulation, etc.) as found for 18 populations of Dactylis glomerata across Scandinavian and Mediterranean sites (Bristiel et al., 2018). Our results are consistent with the growth versus drought survival trade-off reported in summer dormant perennial grasses (Bristiel et al., 2018; Ofir, 1986; Volaire, 1995) and with the growth versus frost survival trade-off in winter dormant herbaceous perennials (Horvath et al., 2003; Pembleton & Sathish, 2014). In many species, the growth slowdown or cessation during the life-threatening season is associated with plant dormancy (Vegis, 1964). However, complete dormancy associated with foliage senescence induced even under favourable wet summers or mild winters (Volaire & Norton, 2006) was not detected in ryegrass in any of the tested environmental conditions. The observed seasonal patterns of reduced potential growth rate can therefore be ascribed to incomplete dormancy in this species (Volaire & Norton, 2006). The seasonal modulation of growth appears crucial to plant adaptation under severe chronic abiotic stresses since a low growth potential was beneficial to survival in both harsh summers and winters.

Our study provides traits at the population level that can be regarded as more representative of overall plant functioning than the classical leaf trait proxies (Cornelissen et al., 2003). High potential annual growth (PAG) was regarded here as indicative of resource acquisition while high lignin content in vegetative biomass was considered as indicative of resource conservation (Gardarin et al., 2014). As expected, the least productive populations were the most lignified, thus demonstrating their higher resource conservation. More lignified tissues were also found associated with greater embolism resistance and dehydration tolerance within a grass species (Volaire et al., 2018). This result supports the ‘fast–slow’ leaf economics spectrum (Reich, 2014). According to that spectrum, populations able to withstand drought appeared to be overall more ‘resource conservative’, whereas frost-resistant populations appeared to be more ‘resource acquisitive’ (Figure 4). Nevertheless, as they both survive dehydration stress by reducing growth (albeit at different seasons), this suggests that the overall fast–slow economics spectrum cannot directly explain the relationships between growth potential and stress survival if phenological and seasonal regulations are not integrated (Chuine, 2010). The frost-resistant populations appeared to be more ‘resource acquisitive’ because incomplete dormancy during a mild winter is less detrimental to potential annual growth than incomplete dormancy during a wet summer. The populations able to withstand drought or frost were either ‘resource conservative’ or ‘resource acquisitive’ according to the season, whereas those sensitive to both drought and frost were invariably ‘resource acquisitive’.

4.2 The growth–survival trade-off rests on contrasting seasonal strategies

The adaptive value of plant traits with regard to abiotic stresses can be explored through the link between ecological strategies, that is, the general trade-offs of the plant economics spectrum and ecophysiological strategies (Volaire, 2018).

Summer persistent populations of ryegrass had a greater reproductive investment (AHD and DST). High spike density and strong aftermath heading in the populations best surviving drought confirms former results (Barre et al., 2018) and can be ascribed to the strategy of dehydration escape (Berger et al., 2016). Furthermore, according to Williamson (2008), later emerging tillers could be too immature to survive summer drought, whereas the better established wintered tillers maintained in a vegetative state through a flowering repression mechanism could allow perennation. Such mechanism may allow plants to combine a dehydration tolerance strategy with well-established vegetative tillers that survive drought periods and a dehydration escape strategy with reproductive tillers that mimic the behaviour of annual species by perpetuation as desiccation-tolerant seeds (Berger et al., 2016). Dehydration tolerance in summer persistent populations may also be partly favoured by a higher intrinsic water use efficiency in spring (lamina 13C) in agreement with other studies (Migalina et al., 2014).

Drought persistence is associated with ‘staying green’ in summer when only populations originating from sites with low aridity (groups G1 and G2) are considered. Populations with high ‘stay-green’ also had a late heading date suggesting that a long period of spring vegetative growth before flowering may ensure greater root establishment and thus greater soil resource acquisition capabilities in summer. However, high ‘stay-green’ in dry summers was not found in populations from more arid areas (group G3). To ‘stay green’ and keep growing in dry conditions may contribute to deplete soil water and thus make plants extremely vulnerable to an extended and extreme drought (Zhao et al., 2017). This behaviour can be assigned to dehydration avoidance, a strategy that does not enable plant survival under severe drought (Yates et al., 2019). Populations of ryegrass from areas of intense summer drought exhibited a low growth potential in a wet summer. This could be considered as an ‘incomplete dormancy’ which has been shown to enhance dehydration tolerance in tall fescue and cocksfoot (Volaire & Norton, 2006). Our results suggested a trade-off between dehydration avoidance of populations from moderate summer drought areas and dehydration tolerance ensuring drought survival of populations from the driest areas.

Regarding frost survival, our study decoupling the measurements of growth potential in favourable winters and frost survival in harsh winters showed the crucial role of reduced growth potential for dehydration survival in a perennial grass. High winter persistence was also associated with high accumulation of soluble carbohydrates in autumn, in accordance with the role of carbohydrate reserves in winter hardiness already noticed in perennial grasses (Chatterton et al., 1989). Soluble sugars were indeed shown to enhance dehydration tolerance as they contribute to cell osmotic adjustment which sustains cell integrity in surviving meristems (Ding et al., 2019). It has been questioned whether frost survival was associated more to winter dormancy (i.e. growth cessation) or to winter hardiness (Castonguay et al., 2006). Our results suggest that low growth potential and high winter hardiness are coupled, notably through the seasonality of carbon resources allocation.

4.3 Strategies of adaptation to climate: Which types of ryegrass in future climates?

This study identified three major groups of ryegrass adapted to their climatic environments.

The G1 group gathers year-round productive populations which are frost and drought sensitive. They originate from sites favourable for plant growth harbouring highly productive meadows (Smit et al., 2008). With early growth, high growth rate, tall fertile stems, wide leaves and erect growth habit, these populations are potentially adapted to interspecific and intraspecific competition in plant communities where competition is primarily for light capture (Aerts, 1999). The G2 group gathers populations that are dehydration avoidant (‘stay green’) under summer drought and dehydration tolerant under frost. They originate from frost-prone sites. Their reduced winter growth points to a natural selection of some level of winter dormancy (Sanada et al., 2007), whereas their high autumn-soluble carbohydrate content, known to protect cell membranes (Livingston et al., 2009), contributes to winter hardiness. They also have a low reproductive investment, which may favour aerial water-soluble carbohydrates accumulation in autumn and thus winter persistence (Williamson, 2008). The G3 group gathers populations that combine dehydration escape and tolerance and frost susceptibility. They originate from relatively dry and hot areas such as the Mediterranean basin. They have a low summer growth potential but high winter growth potential, high aftermath heading and reproductive investment. This typology may be representative of more general patterns of climatic adaptation in temperate perennial grasses since three groups with similar strategies and environmental distributions were also identified in Festuca rubra s.l. and Festuca ovina s.l. (Sampoux & Huyghe, 2009).

The spatial distribution of the area of suitability of each ryegrass group may notably shift during next decades. In the 2041–2070 period, populations with trait adaptation to dry hot conditions (G3 group) could become well suited to large areas in Eastern Europe in which populations with trait adaptation to colder and wetter conditions are presently prevalent (G2 group). This prediction is coherent with changes in grassland phenology across Europe foreseen by Chang et al. (2017). Moreover, populations adapted to frequent winter frost (G2 group) could become suited to more northern areas in Scandinavia that are presently too cold for ryegrass (Ergon et al., 2018). However, despite lower exposure to frost events in high latitudes, winter survival could remain at risk due to complex interactions between environmental factors and cold acclimation (Dalmannsdottir et al., 2017). Our projections also predict that climatically not stressful areas suitable for the G1 group should shrink in southern Europe. On the other hand, such areas should expand in more northern regions in accordance with the increase in grassland productivity potential also foreseen by other studies (Chang et al., 2017; Ergon et al., 2018). Furthermore, the future climate of a notable part of the Mediterranean area should shift out of the range of our calibration, due to extremely high foreseen summer potential evapotranspiration. At present, annual or biennial Lolium taxa (L. rigidum, L. multiflorum) are prevalent in regions with extremely hot and dry summers. Changing climate could thus induce the replacement of perennial ryegrass by less persistent Lolium taxa in Southern Europe since complete summer dormancy, which enhances survival to extreme drought in perennial grasses, has not been found in perennial ryegrass (Norton et al., 2016). However, novel combinations of adaptive traits not presently found in the three delineated groups may be necessary for adaptation to unprecedented combinations of climatic constraints (Trnka et al., 2011).

The climate distribution models are expected to predict the realized niche (Hutchinson, 1957) of ryegrass groups. However, their spatial distribution in foreseen future climate should be refined by taking into account climate × soil interactions, comprehensive changes in plant community distribution, biotic interactions and air CO2 concentration (Brinkhoff et al., 2019).

This study suggests that foreseen climate changes will require migration and adaptation of the natural diversity of ryegrass. The timing of life cycle events, that is, the phenology, has been stressed as essential for climatic adaptation of plant species (Chuine, 2010). This is supported by the key role of seasonal growth timing found in our study, which determines adaptation and persistence under severe stresses. The current local natural diversity of a species may not always be sufficient to adapt to the rapidly changing climate. Therefore, assisted migration of useful natural or recombined diversity (Ferrarini et al., 2016) and reinforcement of the European biogenetic network of grassland (Wolkinger & Plank, 1981) may be necessary to adapt perennial ryegrass to changing climate at regional scale and thus to contribute to maintaining grassland ecosystem services.

ACKNOWLEDGEMENTS

Thanks to the curators from the genebanks and staff from European agronomic research institutes who contributed to in situ collections in 2015. Thanks to the staff from IPK, ILVO and INRAE who set up the common garden experiments and recorded the phenotypic data. Climate data were processed by CERFACS (httpscerfacs.fr) from EURO4M-MESAN and EUMETSAT CM SAF grids for 1989–2010 observed data and from EUROCORDEX grids corrected for bias for the RCA4 model projections (1971–2010). L. Perenne is a species covered under the Multilateral System of the International Treaty on Plant Genetic Resources for Food and Agriculture. All genetic materials used in this study were made available to the authors after signature of a Standard Material Transfer Agreement (SMTA) that ensures the respect of the Nagoya Protocol. We declare no conflict of interest.

    AUTHORS' CONTRIBUTIONS

    J.-P.S. planned and designed the research; F.S., I.R.-R., E.W., M.H., H.M., T.R. and K.J.D. helped provide genetic resources and established experiments; T.K., F.V., J.-P.S., T.L., P.B., J.-L.D. and J.-L.B.-P. analysed the data; T.K., F.V. and J.-P.S. wrote the manuscript; T.K. and F.V. contributed equally. All authors contributed critically to the drafts and gave final approval for publication.

    DATA AVAILABILITY STATEMENT

    All data required to replicate the presented results are available in the associated supplementary files archived in the figshare repository (https://figshare.com/browse) and accessible using the following https://doi.org/10.6084/m9.figshare.12640019.