Global ‘evolution gardens’ show plants can adapt to warming—until heat pushes them past a tipping point

In a fenced research plot at the edge of Israel’s Negev Desert, most of the tiny mustard plants never made it to seed. The same mix of seeds, planted that autumn in an alpine garden high in the French Alps, turned the soil into a dense green mat.

For five years, scientists repeated that exercise at more than 30 outdoor sites across three continents, planting identical packets of wild plant seeds from Spain, Sweden, Iran and dozens of other locations into radically different climates. They then watched evolution play out in real time.

Their verdict: evolution can help plants keep pace with a warming world — but only up to a point. Beyond a certain heat threshold, even a species primed for rapid adaptation starts to fail.

A synchronized evolution experiment across three continents

The findings, published March 26 in Science, come from what researchers describe as the largest synchronized evolution experiment ever attempted in a multicellular organism. Led by Xing Wu and plant evolutionary geneticist Moisés Expósito‑Alonso, the international GrENE‑net consortium sowed about 3.5 million seeds of the model plant Arabidopsis thaliana in more than 30 outdoor “evolution gardens” from 2017 to 2022.

“We conducted a replicated, globally synchronized evolution experiment with the plant Arabidopsis thaliana for 5 years in over 30 outdoor experimental gardens with distinct climates across Europe, the Levant, and North America,” the authors wrote.

Arabidopsis, a weedy annual from the mustard family, is the workhorse of plant genetics. It grows quickly, produces lots of seeds and has thousands of sequenced wild strains, or accessions, collected from across Eurasia and North Africa. That genetic catalog allowed GrENE‑net researchers to track how different lineages rose or fell in the field as weather conditions varied.

At each site — from the Lautaret alpine garden in France to temperate fields in Germany and the United States to drylands in Israel — collaborators scattered mixed seed lots created from about 231 natural accessions of Arabidopsis. Each garden established replicated trays or plots and let the plants grow under fully natural conditions, exposed to local temperatures, rainfall patterns, soils and microbes.

Every year, teams counted surviving plants and collected a single flower from each individual that managed to reproduce. Those flowers were pooled and their DNA sequenced, a technique known as pooled sequencing, to estimate how the frequency of thousands of genetic variants changed from one generation to the next.

The result was a global time series of evolution under real‑world climate conditions.

Rapid adaptation—often predictable—until the climate gets too extreme

Across many sites, the plant populations evolved quickly. Certain genetic lineages surged in frequency within just a few generations, while others dwindled. Those changes were often parallel among replicate plots in the same climate, suggesting strong and predictable natural selection, but diverged sharply between cold and hot locations.

In general, lineages whose home climate resembled a garden’s conditions tended to do better. Strains from warmer, drier regions gained ground in hot gardens, and cool‑adapted strains increased in cooler sites — a pattern consistent with local adaptation.

Yet when the team quantified each lineage’s “favorite” temperature, a striking pattern emerged. For climate specialists whose performance peaked in a narrow range, the estimated optimum temperature was, on average, about 1.9 degrees Celsius cooler than the current climate at their home collection sites.

That lag is similar in magnitude to the global warming recorded over the past century.

“Together, our results show that many Arabidopsis lineages are already best suited to climates somewhat cooler than those they currently experience,” Expósito‑Alonso said in an interview. “That means they are, in a sense, chasing a moving target.”

Genes linked to stress response, dormancy and flowering time

The experiment also sheds light on the genetic machinery behind climate adaptation. By scanning the genome for regions where allele frequencies tracked temperature and rainfall, the researchers identified dozens of candidate genes.

Those include genes involved in heat and stress response, such as CAM5 and several heat shock factors; CYP707A1, which influences the plant hormone abscisic acid and affects seed dormancy and germination timing; and TWIN SISTER OF FT, a key regulator of when plants flower in spring. The patterns support a picture of climate adaptation as a polygenic process shaped by many small genetic changes affecting stress tolerance, germination and flowering time.

‘Evolutionary rescue’—and an eco-evolutionary tipping point

Crucially, the team did not just monitor which genotypes thrived. They also tracked whether whole populations collapsed or recovered as selection took hold.

In 8 of the 30 gardens with full genomic data, the number of plants followed a U‑shaped curve: an initial decline, followed by a rebound in later years. Those recoveries coincided with large, non‑random shifts in allele frequencies, consistent with what evolutionary biologists call evolutionary rescue — a process in which adaptive evolution prevents population extinction in a changing environment.

“Eight out of 30 experimental gardens showed average significant signs of population recovery … with U‑shaped trajectories reminiscent of evolutionary rescue,” the Science paper reported.

But that hopeful pattern did not hold everywhere. In the warmest and driest gardens, particularly those with mean annual temperatures around 15 degrees Celsius or higher, many populations dwindled and never recovered. The genetic changes there looked less orderly.

To quantify that, the team calculated how well they could predict early allele frequency changes in each tray from simple models of climate adaptation. In moderate climates, higher predictability went hand in hand with higher survival over five years. In the hottest climates, populations only had better‑than‑even odds of persisting when those early evolutionary responses were highly predictable. Below that threshold, extinction risk rose sharply.

“This reminds us of eco‑evolutionary tipping points,” the authors wrote, “where in extreme environments natural selection increases mortality and overpowers the efficiency of evolutionary adaptation, leading to erratic evolutionary trends.”

A news release from AnaEE‑France, which operates several of the French field sites, framed the result more bluntly: “In the hottest environments … selection pressure became so extreme that it exceeded the populations’ adaptive potential. Scientists term these ‘eco‑evolutionary tipping points.’”

Implications for conservation and agriculture

The study arrives as governments negotiate new biodiversity targets under the United Nations Convention on Biological Diversity and refine climate adaptation plans under the Paris Agreement. Many of those strategies implicitly assume that species will adjust their ranges or evolve new tolerances as temperatures rise.

The GrENE‑net experiment suggests that assumption has limits.

Even for Arabidopsis — a short‑lived, highly variable species that may be better equipped to adapt than many trees, corals or large mammals — evolution only buffered climate stress up to a point. Where heat and drought pushed conditions beyond the plants’ adaptive range, natural selection hastened declines rather than averting them.

For conservation biologists, the findings underscore the importance of preserving genetic diversity within species, not just the number of species. Seed banks, protected areas that encompass a range of climates and “assisted gene flow” efforts that move genotypes among populations all aim to maintain or bolster the standing variation that fueled adaptation in the experiment.

For agriculture, the work highlights genetic pathways — like heat shock responses, hormone‑controlled germination and flowering time — that breeders and biotechnologists may target as they develop crop varieties resilient to heat waves and erratic rainfall. It also suggests that relying on crops to naturally adapt to rapid warming, without active breeding and deployment, may be risky.

The authors caution that their study focuses on a single species over only a handful of generations, and that more work is needed in long‑lived plants and in systems where species interactions such as pollination and competition play a larger role. Still, they argue, the global Arabidopsis experiment offers rare empirical evidence that evolutionary theory’s warnings about rapid environmental change are playing out in real time.

“Rapid climate adaptation may be possible,” the paper concludes, “but understanding its limits across species will be key for biodiversity forecasting.”

In the alpine garden, where snow still lingers in spring, the mixed Arabidopsis plots largely managed to keep pace, reshuffling their genetic deck to match the chill. In the desert plots, the same genetic deck was not enough. As conditions warmed and dried, the experiment suggests, the plants crossed an invisible line — a threshold beyond which evolution itself could no longer buy them time.