Natural depletion of H1 in sex cells causes DNA demethylation, heterochromatin decondensation and transposon activation

Transposable elements (TEs), the movement of which can damage the genome, are epigenetically silenced in eukaryotes. Intriguingly, TEs are activated in the sperm companion cell – vegetative cell (VC) – of the flowering plant Arabidopsis thaliana. However, the extent and mechanism of this activation are unknown. Here we show that about 100 heterochromatic TEs are activated in VCs, mostly by DEMETER-catalyzed DNA demethylation. We further demonstrate that DEMETER access to some of these TEs is permitted by the natural depletion of linker histone H1 in VCs. Ectopically expressed H1 suppresses TEs in VCs by reducing DNA demethylation and via a methylation-independent mechanism. We demonstrate that H1 is required for heterochromatin condensation in plant cells and show that H1 overexpression creates heterochromatic foci in the VC progenitor cell. Taken together, our results demonstrate that the natural depletion of H1 during male gametogenesis facilitates DEMETER-directed DNA demethylation, heterochromatin relaxation, and TE activation.


24
Large proportions of most eukaryotic genomes are comprised of transposable elements (TEs), 25 mobile genetic fragments that can jump from one location to another. For example, TEs 26 comprise approximately 50% of the human genome (Lander et al., 2001;Venter et al., 2001), 27 and more than 85% of the genomes in crops such as wheat and maize (Schnable et

132
Heterochromatic transposons are preferentially expressed in the vegetative cell 133 To measure the extent of TE activation in the VC, we performed RNA-seq using mature pollen TEs that are transcribed at significantly higher levels in pollen than rosette leaves (fold change > 137 2; p<0.05, likelihood ratio test), and hence likely to be specifically activated in the VC ( Figure   138 1-source data 1) (Slotkin et al., 2009).

139
The VC-activated TEs are primarily located in pericentromeric regions and exhibit features of significantly overrepresented (p<10 -9 and 0.01, respectively, Fisher's exact test; Figure 1C).

147
Transposon derepression in the VC is caused by DME-directed DNA demethylation 148 To assess whether TE activation in the VC is caused by DME-mediated DNA demethylation, 149 we examined DNA methylation in VC and sperm at the 114 activated TEs. We found that these   Vegetative-cell-expressed H1 impedes DME from accessing heterochromatic transposons 169 We next tested our hypothesis that the lack of histone H1 in the VC (Hsieh et al., 2016) allows 170 heterochromatin to be accessible by DME. We first examined the developmental timing of H1

181
To understand how H1 affects DME activity, we ectopically expressed H1 in the VC. To ensure 182 H1 incorporation into VC chromatin, we used the pLAT52 promoter, which is expressed from To evaluate the effect of VC-expressed H1 on DNA methylation, we obtained genome-wide 194 methylation profiles for VC nuclei from a strong pVC::H1 line (#2; Figure 2B) and WT via  Figure 2D). However, a substantial proportion of loci that are targeted by DME 201 show hypermethylation in pVC::H1 VC ( Figure 2D). DME targets also show preferential test), indicating that most H1 hyperDMRs are DME targets ( Figure 2E-H). 212 Our results demonstrate that H1 hyperDMRs are primarily caused by the inhibition of DME.

213
However, only 3066 out of 11896 (26%) VC DME targets have significantly more CG 214 methylation in the VC of pVC::H1 than WT (p<0.001, Fisher's exact test; Figure 1-source 215 data 2), indicating that VC-expressed H1 impedes DME at a minority of its genomic targets.

216
These H1-impeded DME targets are heterochromatic, significantly enriched in H3K9me2 217 compared with H1-independent DME targets ( Figure 2I). To further examine the link with 218 heterochromatin, we aligned all VC DME target loci at the most hypomethylated cytosine, and 219 separated them into five groups by H3K9me2 levels ( Figure 2J). pVC::H1-induced 220 hypermethylation peaks where DME-mediated hypomethylation peaks, but is apparent only in 221 the most heterochromatic group (highest H3K9me2) of DME target loci ( Figure 2J). Taken 222 together, our results demonstrate that developmental removal of H1 from the VC allows DME 223 to access heterochromatin. 225 Given the importance of H1 removal for DME-directed DNA demethylation, we investigated   of leaf nuclei showed that H1 co-localizes with H3K9me2 in highly-compacted 261 heterochromatic foci, known as chromocenters ( Figure 5A). Furthermore, we found that 262 chromocenters become dispersed in the nuclei of h1 mutant rosette leaves ( Figure 5B). These 263 observations demonstrate that H1 is required for heterochromatin condensation in plants. 264 We then examined whether ectopic H1 expression can condense the heterochromatin in VC  Heterochromatin decondensation during male gametogenesis seems to be gradual: 271 chromocenters are observed at early microspore stage, but become dispersed in late microspore 272 stage, when H1 is depleted (Figures 2A and 5C). We observed strong and weak chromocenters 273 in 27% and 59%, respectively, of late microspore nuclei, whereas no chromocenters were   is dependent on DME but not H1 (Figures 3D and 6). Group II comprises TEs in which H1 326 absence is required for DME demethylation and activation ( Figure 6). For TEs in Group III, 327 H1 depletion and DME demethylation are both required for activation, but DME activity is not 328 affected by H1 (Figure 6). Group IV TEs are activated by H1 depletion and are not targeted by 329 DME (Figure 6). Groups III and IV demonstrate that H1 can silence TEs independently of 330 DNA methylation. Group III also demonstrates that DNA methylation and H1 cooperate to  Therefore, at least some TEs may be hijacking an essential epigenetic reprogramming process.     with TPM (transcripts per million) more than 5 in the Kallisto output (data used in Figure 3D). 392 To identify TEs and genes that are suppressed by H1 in the VC, we analyzed RNA-seq data 393 from WT and pLAT52::H1.1-mRFP line #2 (short as pVC::H1 unless specified otherwise) 394 pollen using Kallisto and Sleuth as described above. Significant differential expression was 395 defined with a fold change at least 2 and a p-value less than 0.05. H1-repressed TEs were listed 396 in Figure 1-source data 1.

407
As all CG hypomethylation in the VC in comparison to sperm is caused by DME (Ibarra et al.,  Similarly, ends analysis of TE transcripts was performed using the annotation of VC-activated 435 TEs described above (Figure 1-source data 1). DNA methylation data from (Ibarra et al.,  In Figure 2J, DME sites were aligned at the most demethylated cytosine, and average CG

444
This resulted in 13610 DME sites, which were separated into five groups according to 445 H3K9me2 level : < 2.5, 2.5-4.3, 4.3-6.5, 6.5-10.5, and > 10.5 ( Figure 2J). 446 The most demethylated cytosine within each site was identified if it had the greatest differential 447 methylation in sperm than VC among cytosines in the CG context (sperm -VC > 0.2, and 448 Fisher's exact test p < 0.001) and was sequenced at least 10 times.

450
Differential methylation at a 600-bp region centered upon the TSS of H1-repressed TEs was 451 calculated between VCs of pVC::H1 and WT ( Figure 4A). TEs whose differential methylation 452 is significant (Fisher's exact test p < 0.001) and larger than 0.2 (in CG context), 0.1 (in CHG 453 context), or 0.05 (in CHH context) are illustrated in the upper panel in Figure 4A.  Author information 486 Sequencing data have been deposited in GEO (GSE120519). The authors declare no competing 487 financial interests. Correspondence and requests for materials should be addressed to X.F.

488
(xiaoqi.feng@jic.ac.uk).        H3K9me2 level at VC DME targets that are significantly hypermethylated in pVC::H1 (H1-799 inhibited) or not (H1-independent), respectively. Difference between the two groups is significant 800 (Kolmogorov-Smirnov test p < 0.001). (J) VC DME targets were grouped according to H3K9me2 801 levels, aligned at the most demethylated cytosine (dashed lines), and plotted for average CG 802 methylation difference as indicated in each 10-bp interval (left). Similarly, CG methylation in 803 pVC::H1 and WT VCs was plotted for the group with the lowest and highest H3K9me2, 804 respectively. Spm, sperm. 805 The following figure source data and supplements are available for Figure 2: 806     and co-localizes with strong CCs. All bars, 5 µm. 867 of TEs in each group is shown on the top. Significantly less heterochromatic than TEs in other 872 groups ( Figure 3G), Group I TEs are activated by DME-directed DNA demethylation. Group II 873 TEs rely on H1 depletion to allow DME demethylation and activation. Group III TEs are 874 demethylated by DME but require H1 depletion to allow transcription (ie. pVC::H1 represses these 875 TEs without affecting DME). Group IV TEs are not demethylated by DME; their activation is 876 solely dependent on the depletion of H1. TEs belong to each group are listed in Figure 1-source 877 data 1. Red lollipops denote DNA methylation. 878 879