Organisms are already facing climate change driven by recent anthropogenic activities. In an attempt to understand and mitigate the negative effects of climate change on wild and farmed species, recent research focused on predicting the fitness of organisms or populations in future climates. The accuracy of these predictions is, however, seldom tested. To test such predictions, we grew 800 genetic families of the annual selfing plant Arabidopsis thaliana in the same field site for two consecutive years with contrasted climates. Despite observing, in both years, a clear association between fitness and climatic distance between our field site and the climate of origin of the genetic families, the diverse set of methods we used failed to accurately predict fitness from a year to another. This impossibility resulted from different contributions of climatic factors in climate adaptation every year, which impeded the definition of a meaningful climatic descriptor across years. Our results also show that, for our study populations at least, vegetative growth is a more important trait for climate adaptation than phenology. We discuss the implications of our results for predicting the fitness of wild organisms in future climates and for breeding programs.

Plant material

The plant material we used has already been described in a previous study (Brachi et al. 2013) but we provide a short description of this material hereafter. In 2009, we collected 800 individuals of the annual plant species A. thaliana from 49 natural populations in four regions of France with different climatic regimes (Brittany: oceanic climate - 11 populations; Burgundy: continental climate - 11 populations; Languedoc: Mediterranean climate - 16 populations; North of France: semi-continental climate, 11 populations; Table S1- Figure S1 of the associated publication). We collected an average of 16 plants per population (mean=16.3 ±5.6). We collected these plants before they initiated flowering, and grew them in isolation with aratubes (https://www.arasystem.com/asn003-aratubes-360) in the greenhouse (16h photoperiod, 20°C) to prevent cross-pollination and harvest seeds. The seeds we obtained from each individual plant therefore constitute a genetic family. Because A. thaliana is a highly selfing species, each genetic family is quasi-homozygous (Platt et al. 2010). To reduce the impact of maternal effects in our experiments (Rossiter 1996) and increase the proportion of variance due to genetic factors, we grew each genetic family in isolation with aratubes for an additional generation in a common environment (i.e., greenhouse conditions, 16h photoperiod, 20°C) and harvested their seeds. These seeds are the starting material for the experiments described in this study.

Field experiments

We set the same experiment for two consecutive years in the same field site in the north of France (Villeneuve-d’ascq, France, GPS coordinates: 50°36'27.2"N, 3°08'37.0"E). For convenience, we refer to the first experiment as the ‘2011 experiment’ and the second experiment as the ‘2012 experiment’. For each experiment, we aimed at growing two individuals from each genetic family. We sowed seeds on a similar Julian day of the year for both experiments (the 27th of September 2010 for the 2011 experiment, and the 26th of September 2011 for the 2012 experiment). The following design and protocol were identical for both experiments.

For each genetic family replicate, at least four seeds were placed within an individual well of a 66-well botanical tray (TEKU, JP 3050/66) filled with damp standard soil culture (007B Florafleur, France), so that a given well contained only seeds from a given genetic family. Following sowing, we placed the trays in a cold chamber (4°C) for four days, to stratify the seeds and promote germination. We then placed the trays in a frost-free greenhouse, without additional light or heating. We treated the trays with Vectobac (8mL per liter) on the day of their placement in the frost-free greenhouse, to protect them against a possible episodic outbreak of the pest dark-winged fungus gnats. We monitored the trays daily for 17 days after their placement in the frost-free greenhouse to score germination date. Germination rates (measured as the percentage of sown wells for which at least one germination was observed) were very high with 96.4% and 96.3% of the wells containing germination for the 2011 experiment and the 2012 experiment, respectively. On day 18, when possible, we transplanted two seedlings of a specific genetic family replicate that germinated into a replicate of the same genetic family that did not germinate. On that same day, we thinned all individual wells to keep two seedlings per well and moved the trays onto the experimental field, located outside the greenhouse at the University of Lille 1 France (GPS coordinates: 50°36'27.2"N, 3°08'37.0"E). To facilitate root development (the wells have a hole allowing root development into the soil), we tilled the soil below the trays to be able to bury them slightly into the ground. The field was far enough from any surrounding buildings to have direct sunlight all day. We did not place any object or plastic covering on top of the plants, as we aimed to let the plants experience the climatic conditions at the field site (sunlight, temperature, and precipitation). A week after placement in the field, we thinned all wells to only retain a single seedling per well.

We organized each experiment into two experimental blocks, with one replicate of each genetic family within each block. Each block comprised 19 trays, with genetic families organized randomly within the 19 trays. We placed trays belonging to the same block near each other in four rows and five columns. We placed blocks approximately five meters apart. To reduce the impact of micro-environmental variations, we rotated trays daily in the cold chamber and the frost-free greenhouse. To control for possible micro-environmental variations within blocks in the field, we placed two individuals from the same genetic line (Bg-2 line, referred to as control hereafter) at the same two positions within each tray. Another set of 398 worldwide or French genetic families was also included in these two field experiments for another study that investigated the genetics underlying natural variation in glucosinolate profiles (Brachi et al. 2015).

In total, the 2011 experiment and the 2012 experiment involved the monitoring of 3,352 individual plants (1,676 plants per experiment). To score phenological and vegetative traits as accurately as possible, we monitored individual plants on the field site every two or three days during the entirety of the experiments, from the placement of trays on the field site to the production of the last mature fruit.

Estimating phenology, vegetative growth, and seed production

We measured four phenological traits (germination time, bolting time, the interval from bolting to flowering, and the length of the reproductive period), one vegetative trait (diameter at bolting), and an estimator of fitness (seed production) on all plants.

We measured germination time (GERM hereafter) as the number of days elapsed from the end of the stratification treatment to germination. We measured bolting time (BT hereafter) as the number of days elapsed between germination and the onset of bolting. We measured the flowering interval (INT hereafter) as the number of days elapsed between bolting and the opening of the first flower. We measured the length of the reproductive period as the number of days elapsed between the opening of the first flower and the maturation of the last fruit. To facilitate trait value comparison across years, we rescaled these four phenological traits in photothermal units (PTU) using a phenological model integrating both daily photoperiod length and daily temperatures measured in the greenhouse and at the field site (Brachi et al. 2010).

As an estimator of vegetative growth, we measured the maximum diameter of the rosette at bolting (DIAM hereafter) using a plastic ruler (~0.5mm accuracy), as this trait is a good descriptor of plant size in A. thaliana (Weinig et al. 2006).

The lifetime fitness of each plant (seed production hereafter) was estimated by the total length of fruits produced, which strongly correlates with seed count (Roux et al. 2004; Roux et al. 2005). Plants that germinated but did not survive to produce any mature fruit were given a total fruit length value of zero.

Climatic data

We obtained past climatic data for our 49 populations for the worldclim 2.1 database (Fick and Hijmans 2017). Specifically, we downloaded the 19 bioclimatic variables for these populations using the finest resolution data (30 seconds = ~1km resolution). These estimates are the average for the years 1970-2000.

To characterize the climate experienced by plants in our experiments, we used the meteorological reports (temperature and precipitation) from the nearest meteorological station (Lille-Lesquin, 59, France) for the years 2011 and 2012. We transformed these raw records into 19 bioclimatic variables, similar to the one mentioned above, but for our field site for the years 2011 and 2012. We performed this transformation using the function calc_biovars from the package QBMS (v 0.9.1) (Al-Shamaa 2023).

BLUEs estimation

We used raw data for most analyses unless specified otherwise. For some analyses we used BLUEs values. We calculated them as follow.

 For each experiment, we estimated the genotypic value for each trait of each genetic family within each experiment by running the following mixed model for each trait:

Trait_ij = b₁ block_i + b₂ family_j + b₂ control_ij + intercept (equation 3)

Where ‘trait’ corresponds to the trait value for a given plant, ‘block’ corresponds to differences between the two blocks, ‘family’ corresponds to variation among the 800 genetic families, and ‘control’ corresponds to the mean trait value for the two control plants located on the same tray than this plant. In this model, the family effect was considered as random, and all other effects as fixed. We extracted BLUPs (Best Linear Unbiased Predictions) for each family from this model, and obtained genotypic values (BLUEs - Best Linear Unbiased Estimates) by adding the trait mean computed from raw trait values to these BLUPs. We then perform the same analyses as described in the previous paragraph to obtain genotypic gradients of selection both at the population and genetic family levels. We fitted our mixed models using the lmer function from the lme4 package in R (v4.3.2) (Bates et al. 2015; R-Core-Team 2020) and extracted BLUPs with the ranef function from the same package.

References

Al-Shamaa, K. 2023. QBMS: Query the Breeding Management System(s).

Bates, D., M. Mächler, B. Bolker, and S. Walker. 2015. Fitting Linear Mixed-Effects Models Using lme4. Journal of Statistical Software 67:1 - 48.

Brachi, B., N. Faure, M. Horton, E. Flahauw, A. Vazquez, M. Nordborg, J. Bergelson, J. Cuguen, and F. Roux. 2010. Linkage and association mapping of Arabidopsis thaliana flowering time in nature. PLoS genetics 6:e1000940.

Brachi, B., R. Villoutreix, N. Faure, N. Hautekeete, Y. Piquot, M. Pauwels, D. Roby, J. Cuguen, J. Bergelson, and F. Roux. 2013. Investigation of the geographical scale of adaptive phenological variation and its underlying genetics in Arabidopsis thaliana. Molecular ecology 22:4222-4240.

Platt, A., M. Horton, Y. S. Huang, Y. Li, A. E. Anastasio, N. W. Mulyati, J. Agren, O. Bossdorf, D. Byers, K. Donohue, M. Dunning, E. B. Holub, A. Hudson, V. Le Corre, O. Loudet, F. Roux, N. Warthmann, D. Weigel, L. Rivero, R. Scholl, M. Nordborg, J. Bergelson, and J. O. Borevitz. 2010. The scale of population structure in Arabidopsis thaliana. PLoS genetics 6:e1000843.

R-Core-Team. 2020. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria.

Rossiter, M. 1996. INCIDENCE AND CONSEQUENCES OF INHERITED ENVIRONMENTAL EFFECTS. Annual Review of Ecology, Evolution and Systematics 27:451-476.

Roux, F., C. Camilleri, A. Bérard, and X. Reboud. 2005. MULTIGENERATIONAL VERSUS SINGLE GENERATION STUDIES TO ESTIMATE HERBICIDE RESISTANCE FITNESS COST IN ARABIDOPSIS THALIANA. Evolution 59:2264-2269.

Roux, F., J. Gasquez, and X. Reboud. 2004. The Dominance of the Herbicide Resistance Cost in Several Arabidopsis thaliana Mutant Lines. Genetics 166:449-460.

Weinig, C., J. Johnston, Z. M. German, and L. M. Demink. 2006. Local and global costs of adaptive plasticity to density in Arabidopsis thaliana. Am Nat 167:826-836.