replying to: Z. Zhang et al. Nature https://doi.org/10.1038/s41586-023-06189-z (2023)

We write in response to Zhang et al.1. Using their mutant we confirm our original findings. Zhang et al.1 analyse both F1 populations and FRI-GFP homozygous lines2 (but using different criteria for each) and report that “the cold-induced formation of nuclear FRI condensates is independent of COOLAIR”. Zhang et al.1 also find that “COOLAIR is not involved in FLC repression by prolonged cold exposure” because “loss of COOLAIR expression in either FRIΔCOOLAIR-1 or FRIΔCOOLAIR-2 had no effect on the progression of transcriptional shutdown of FLC during cold exposure or on post-cold stable silencing of FLC”.

Fig. 1: Cold-induced FRI–GFP condensate formation and FLC repression are attenuated in ΔCOOLAIR-1.
figure 1

a–c, Quantification of the FRI–GFP nuclear condensate area (spot area (a) and percentage (c)) and number per nucleus (b) in the roots of FRI-GFP and FRI-GFPΔCOOLAIR-1 homozygous F3 lines. Plants were exposed to cold treatment for 4 days at a constant temperature of 5 °C. For a and b, the open circles indicate the median of the data and the vertical bars indicate the 95% confidence interval determined by bootstrapping. n = 1,729 and 2,548 condensates (a) and n = 271 and 529 (b) nuclei in n = 15 and 26 roots, respectively. Individual data points are shown as black or red dots. Comparison of mean values was performed using two-tailed t-tests with Welch’s correction. d,e, The relative expression level of unspliced FLC (d) and spliced FLC (e) in Col FRI, FRI ΔCOOLAIR-1 and FRI TEX1.0 plants with 2 weeks of growth under the indicated temperature conditions6. Data are mean ± s.e.m. of n = 6 biologically independent experiments. f, Schematic of FLC and COOLAIR transcripts at the FLC locus. Untranslated regions are indicated by grey boxes and exons by black boxes. Head-to-head arrows indicate primers used for antisense transcript level analysis by quantitative PCR (qPCR). g–i, The relative expression level of antisense transcripts at FLC, including but not limited to CAS6, by the indicated primers (P1 (CAS) (g), P3 (h) and P4 (i)) in Col-FRI and FRIΔCOOLAIR-1 plants with 24 h of growth under the indicated temperature conditions. The primers used are indicated in f. Data are mean ± s.e.m. of n = 3 biologically independent experiments. NS, not significant; the exact P values are shown at the top of each comparison.

Source data

Full size image

We have combined the He laboratory ΔCOOLAIR-1 deletion3 with our FRI-GFP transgene2 (FRI-GFP ΔCOOLAIR-1) and repeated the analysis. In contrast to their data, we show that deletion of the COOLAIR promoter significantly attenuates cold-induced formation of FRI–GFP nuclear condensates, changing the overall size distribution4 (Fig. 1a–c and Extended Data Fig. 1a–c). This fully confirms our original findings that antisense transcription is one component of the multiple cold responsive factors regulating FRI condensates2. FRI–GFP condensates show concentration dependency and plasticity to environmental conditions, both well-known properties of condensate dynamics5.

We disrupted COOLAIR expression in our original report using a terminator exchange construct (TEX1.0)6 and, indeed, the level of FRI–GFP is reduced in this line as it is in the frl1-1 mutant (discussed in our original paper), a component of the FRI complex2. These effects are not due to transgene-induced RNA silencing of FRI-GFP (Extended Data Fig. 1d). We also confirm that the repression of spliced FLC RNA levels, which is particularly sensitive to widely fluctuating cold conditions6, is attenuated in the FRI ΔCOOLAIR-1 line (Fig. 1d,e). Over the years of study, we find that FLC downregulation in the cold, even in wild-type plants, is very dependent on growth conditions and seedling density, with growth being essential for RNA reduction. Cold-induced downregulation of FLC expression is mediated through several different mechanisms7,8,9,10,11,12,13 with COOLAIR affecting the dynamics of many of these, and not only through FRI condensation.

The ΔCOOLAIR-1 line still produces abundantly expressed antisense transcripts6 (Extended Data Fig. 1e–k). The marked robustness of antisense/non-coding expression at loci such as FLC shows the intrinsic connection of sense/antisense transcription, as has been found at many yeast loci14,15.

These antisense transcripts are inducible by short cold exposure (Fig. 1f–i) but are indeed downregulated after longer exposure (Extended Data Fig. 1f,h,i). The TEX1.0 line has different alternative antisense transcript levels compared with the ΔCOOLAIR-1 line (Extended Data Fig. 1e–k). We chose to use the TEX1.0 line in our original study because it has the lowest levels of alternative antisense transcripts6 (Extended Data Fig. 1e–k). The sequence of events during cold-induced FLC silencing is very dynamic and condition dependent due to the nonlinearity of FLC transcriptional shutdown and epigenetic silencing dynamics. This nonlinearity emerges from the complex feedback mechanisms interconnecting non-coding transcription, chromatin modifications and RNA stability.

Methods

The reference genotype Col FRI2, the FRI-GFP2 transgenic plants, TEX1.02,6 and ΔCOOLAIR-13 have been described previously. ΔCOOLAIR-1 was crossed with Col FRI and FRI-GFP to generate the FRI ΔCOOLAIR-1 and FRI-GFP ΔCOOLAIR-1 lines. Imaging and quantification of FRI–GFP condensates were performed as described previously2. The experiments under fluctuating conditions and all of the qPCR with reverse transcription analyses were performed as previously described6. A list of all of the primers used in the qPCR assay are provided in Extended Data Table 1.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.