Absence of warmth permits epigenetic memory of winter in Arabidopsis

Abstract

Plants integrate widely fluctuating temperatures to monitor seasonal progression. Here, we investigate the temperature signals in field conditions that result in vernalisation, the mechanism by which flowering is aligned with spring. We find that multiple, distinct aspects of the temperature profile contribute to vernalisation. In autumn, transient cold temperatures promote transcriptional shutdown of Arabidopsis FLOWERING LOCUS C (FLC), independently of factors conferring epigenetic memory. As winter continues, expression of VERNALIZATION INSENSITIVE3 (VIN3), a factor needed for epigenetic silencing, is upregulated by at least two independent thermosensory processes. One integrates long-term cold temperatures, while the other requires the absence of daily temperatures above 15 °C. The lack of spikes of high temperature, not just prolonged cold, is thus the major driver for vernalisation. Monitoring of peak daily temperature is an effective mechanism to judge seasonal progression, but is likely to have deleterious consequences for vernalisation as the climate becomes more variable.

Fig. 1

FLC downregulation occurs in the field in both VIN3-independent and dependent manners. a Locations of field sites in Norwich, North Sweden (Ramsta) and South Sweden (Ullstorp). b Temperatures experienced by the plants in the different sites. Dashed line indicates 15 °C. In Norwich, glasshouse confinement is probably responsible for buffering low temperatures (see Methods). c FLC and VIN3 expression for Col FRI^SF2, as measured in the field corresponding to plot above. FLC (top): Thick lines show measured FLC mRNA where c−e thickness of the line represents s.e.m. VIN3 (below): measured expression of VIN3 mRNA, errors = s.e.m. d Comparison of FLC downregulation between different sites. e FLC downregulation in Norwich for ‘wild-type’ (Col FRI^SF2) and a mutant impaired in VIN3 expression (vin3-4 FRI). For the wild-type, the data are separated according to presence (late timepoints) or absence (early timepoints) of VIN3 expression, as measured in panel c. RNA levels in c−e normalised to UBC, PP2A, and internal control. Straight lines (c, e) show the best fit exponential decay profile, fitted by least-squares. n = 2–6, average ≥ 5. Map adapted from https://commons.wikimedia.org/wiki/File:Blank_map_of_Europe.svg by maix, under CC BY-SA 2.5 license

Fig. 2

FLC and VIN3 expression respond to temperature extremes rather than averages. a Temperature profiles of treatments used in the experiment. b, c VIN3 and FLC mRNA from Col FRI^SF2 plants treated with ‘cold’ temperature conditions of panel a. For VIN3, plants were sampled 5–7 h after dawn. b n = 3−6, average > 5.3. c n = 16–21, average > 19. d FLC mRNA in the mutant vin3–4, otherwise as for c. n = 9–21, average > 17. e FLC mRNA in Col FRI^SF2 during (T0) and after removal (T7, T20) from the 14.2 °C constant (14-const) or fluctuating (14-fluct) conditions shown in a, Tx = x days after return to warm, W = weeks in cold. n = 2 shown as grey circles. Error bars = s.e.m. In b−e RNA levels were normalised to UBC, PP2A

Fig. 3

Modelling VIN3 dynamics reveals its temperature inputs. a Minimal model of VIN3 temperature sensing assuming a single thermosensor r whose levels quantitatively regulate VIN3 transcription. p₁= 0.91 day⁻¹, d₁ = 100 day⁻¹, d₂ = 7 day⁻¹. See Methods for more details. b Slow VIN3 upregulation in the cold for unspliced RNA, sampled 5–7 h after dawn. n = 3–6, average > 4.7. c, d VIN3 unspliced and spliced RNA levels upon return to warm (22 °C) following 4 weeks at 5 °C. At time 0, plants were sampled in the cold and subsequently transferred to warm. n = 2–3, average > 2.6. e Slow VIN3 mRNA upregulation in the cold, sampled 5–7 h after dawn. Colours correspond to vernalisation treatment as shown in legend to b. n = 3–6, average = 5. Model results shown in d, e. In b−e, RNA levels were normalised to UBC, PP2A; error bars = s.e.m.

Fig. 4

VIN3 expression is regulated in a circadian manner. a, b VIN3 spliced and unspliced expression over 12 h. Plants grown in 20 °C night, 22 °C day 16 h photoperiod for 1 week and transferred to vernalising 8 h photoperiod with constant or fluctuating 14.2 or 8 °C profiles (as shown in Fig. 2a) for 4 weeks. The dark grey background indicates night-time. Error bars are s.e.m., n = 2–3, average > 2.7. c−f, VIN3 and various clock gene mRNA levels sampled on consecutive days at 8 °C after 4 weeks constant 8 °C, 8 h photoperiod (purple). On the first day of sampling the plants were transferred to constant light conditions (blue). The lighter grey background indicates subjective night. RNA levels normalised to UBC, PP2A. n = 1–3, average > 2.1. g Summary of VIN3 regulation showing the slow regulator (B), the new component (A) found to be necessary from modelling which gives the absence of warm regulation and the circadian clock, which could itself also be used as a temperature input