Local inhibition of PRC2 activity by H3.3K27M drives DNA replication defects through misregulation of the JNK pathway

Substitution of lysine 27 with methionine in histone H3.3 is a recently discovered driver mutation of pediatric high-grade gliomas. Mutant tumor cells show decreased levels and altered distribution of H3K27me3. How these chromatin changes are established genome-wide and lead to tumorigenesis only in specific tissues remains unclear. Here we show that H3.3K27M-mediated alterations in H3K27me3 distribution result in ectopic DNA replication and cell cycle progression of germ cells in Caenorhabditis elegans. By genetically inducing changes in the H3.3 distribution, we demonstrate that both H3.3K27M oncohistone incorporation and pre-existing H3K27me3 act locally and antagonistically on Polycomb Repressive Complex 2 (PRC2) in a concentration-dependent manner, explaining the observed H3K27me3 distribution in mutant cells. The altered heterochromatin patterns lead to extensive misregulation of gene expression. Through unbiased genetic screening, we found that inhibiting JNK pathway components, which are overexpressed in H3.3K27M cells, suppresses the ectopic DNA replication and cell cycle progression without rescuing the altered H3K27me3 distribution. Moreover, we show that JNK inhibition suppresses the replicative fate in human tumor-derived H3.3K27M cells, thus establishing C. elegans as a powerful model for the identification of potential drug targets for treatment of H3.3K27M tumors.


39!
Establishing specific chromatin landscapes to regulate access to the genetic material is an 40! effective and dynamic mechanism to control cell fate and preserve cell identity. Nucleosomes are 41! the basic structural and functional units of this chromatin regulation. They consist of an 42! octameric core of histone proteins that provide a structural scaffold for the organization of DNA. histone H3 that is mainly associated with regions of high nucleosome turnover 3 , and loss of H3.3 48! results in severe sterility or lethality phenotypes in most organisms 4,5 . H3.3 is highly conserved 49! in plants, animals and fungi and is distinguished from canonical H3 by only a few key amino 50! acids that are important for the association with the H3.3-specific histone chaperones HIRA and 51! DAXX [6][7][8] .

52!
Despite the importance for chromatin biology, only a few mutations in histone genes are 53! directly associated with specific diseases. One notable example is the recently discovered 54! replacement of lysine 27 with methionine in histone H3 or, more commonly, H3.3 that acts as a 55! driver mutation of specific types of pediatric diffuse intrinsic pontine gliomas (DIPGs) 9 and 56! some cases of acute myeloid leukemia 10 . In tumor cells carrying this oncohistone, trimethylation 57! of lysine 27 on histone H3 (H3K27me3) is strongly and globally depleted from chromatin 11-13 . 58! H3K27me3 is a mark characteristic for facultative heterochromatin, and is associated 59! with transcriptionally repressed regions 14 . H3K27 methylation is deposited, recognized and 60! propagated by the Polycomb Repressive Complex 2 (PRC2), a multi-protein complex 61! ! 4! responsible for maintaining the silent state of the genes during development and cell 62! differentiation 15,16 . PRC2-mediated H3K27me3 propagation is the result of dynamic interactions 63! between the PRC2 complex and pre-existing H3K27me3, which allosterically activates PRC2 64! and facilitates the spreading of the mark to neighboring nucleosomes 17-21 .

65!
The observed depletion of H3K27me3 in H3.3K27M tumor cells was initially explained 66! by an increased affinity of PRC2 to the K27M-containing H3.3, resulting in PRC2 trapping on 67! the nucleosomes containing this oncohistone 11,13,22 . Sequestration of PRC2 can explain how 68! H3.3K27M, present only in a small fraction of all nucleosomes, acts as a dominant negative 69! factor to affect H3K27me3 levels genome wide. However, retention of the mark in some 70! genomic regions of mutant cells is inconsistent with the hypothesis that PRC2 is immobilized by 71! H3.3K27M containing nucleosomes, and instead suggests that part of the PRC2 pool remains 72! active to maintain H3K27me3 levels at some PRC2 targets 11,12 . Moreover, ChIP-seq analysis 73! showed that PRC2 is excluded from, rather than immobilized on H3.3K27M-containing 74! nucleosomes, and residual PRC2 activity is one of the factors promoting tumorigenesis 23 .

75!
Regardless of the exact mechanism of PRC2 inhibition, several studies revealed H3K27M-76! induced changes of H3K27me3 levels at gene promoters and enhancers, showing that active 77! promoters, which are enriched in H3.3 incorporation and H3K27ac, tend to lose H3K27me3 78! ! 11! can overcome the inhibitory effect of the oncohistone incorporation. This observation is 222! consistent with the local stimulatory role of H3K27me3 for PRC2 activity described recently 18,20 .

223!
In the H3-like K27M mutant, oncohistone incorporation appears more even across the 224! genome (Fig. 3d, right panel). This changed incorporation pattern results in further local 225! reduction of the H3K27me3 on the autosomes, reflected by enlargement of cluster 3 of the heat 226! map (Fig. 3d, right panel). However, some autosomal regions retain high H3K27me3 levels 227! despite increased oncohistone occupancy (Fig. 3d, 38 . Interestingly, H3K27me3 levels are gained or lost over the entire 269! gene body (Fig. 4c). Extensive misregulation affects many specific gene groups and pathways, 270! but analysis of significantly enriched GO terms revealed that the most affected categories fall 271! into kinase-related signal transduction proteins (Fig. 4d). However, from this data it is difficult to 272! identify expression changes in specific genes that are causal for the observed phenotype.

273!
To overcome this limitation and identify key genes involved in aberrant oocyte replication 274! within these pathways in an unbiased way, we performed a random mutagenesis screen for 275! genetic suppressors of H3.3K27M-induced sterility (Fig. 5a). We identified a serine 287 to 276! asparagine substitution in KGB-1 (KGB-1 S287N) as a potent suppressor of endomitosis that 277! restored fertility in worms carrying the oncohistones (Fig. 5b). We confirmed this finding with 278! an additional KGB-1 S287N allele generated by CRISPR/Cas9 mutagenesis that restored fertility 279! to similar levels. KGB-1 is a homologue of mammalian Jun amino-terminal kinase (JNK), a 280! stress-activated MAPK subfamily serine-threonine kinase 39 . In C. elegans, KGB-1 positively 281! regulates the activator protein-1 (AP-1) transcription factor FOS-1 40 , which in turn controls 282! oocyte ovulation via IP3 signaling 41 Fig. 5c) 41,42 . 291! In H3.3K27M mutant germ lines, the kgb-1 gene loses H3K27me3 signal, and kgb-1 292! expression levels are significantly elevated in H3.3K27M mutant germ lines (Fig. 5d-e). The 293! KGB-1 S287N mutation does not restore the nuclear distribution of H3K27me3 observed in the 294! oncohistone mutant (Fig. 5f), indicating that it rescues a defect downstream of the chromatin 295! changes and may affect the stability or activity of the enzyme. Sequence alignment and 296! comparison to the structure of mammalian JNK-1 indicated that the S287N suppressor mutation 297! localizes in the serine/threonine kinase domain of the KGB-1, and is likely exposed on the 298! surface of the protein, thus potentially serving as a phosphorylation target that regulates the 299! activity of the enzyme (Supplementary Fig. 8) 43,44 . To test whether the sterility is driven by 300! diminished activity of the JNK pathway, we reduced the expression levels of kgb-1 by RNAi, 301! which resulted in a partial rescue of the endomitosis phenotype and in a significant increase in 302! fertility of H3.3K27M worms (Fig. 5g). To further validate the JNK pathway as potential target 303! for antagonizing the effects of the H3.3K27M mutations, we treated worms with the JNK 304! inhibitor SP600125 45-47 . Consistent with the RNAi experiments, the drug was able to 305! significantly increase the number of fertile animals in the H3.3K27M population, while not 306! affecting wild type worms (Fig. 5h). Taken together, our results demonstrate that H3.3K27M-307! mediated redistribution of H3K27me3 directly results in gene expression changes, but that the 308! aberrant oocyte replication and endomitosis is driven by misregulation of specific 309! serine/threonine kinases such as JNK. We show that targeting this kinase by RNAi or by 310! chemical inhibitors is an effective way to prevent the aberrant entry of oocytes into the cell cycle 311! and restore fertility in H3.3K27M mutant worms. 312! ! 15!

JNK inhibition is effective on human glioma-derived H3.3K27M mutant cells. 313!
Interestingly, JNK is one of the known targets for inhibiting proliferation of glioblastoma cells 45-314! 47 . This suggests that in addition to the similarities in the H3.3K27M-induced chromatin changes 315! between nematodes and vertebrates, similar downstream gene networks are misregulated to drive 316! replicative cell fates. To explore this possibility, we treated three glioma cell lines, SF8628 317! derived from an H3.3K27M tumor, and SF9402 and SF9427 derived from tumors with wild type 318! H3.3, with the JNK inhibitor SP600125, using assays that were previously described to test 319! differences between the same K27M-negative and -positive glioma cell lines 48  In this study, we provide a general model for the concentration-dependent in cis interplay 357! between pre-existing H3K27me3 and H3.3K27M incorporation that explains the genome-wide 358! H3K27 methylation landscape in H3.3K27M mutant cells. We find that oncohistone 359! incorporation antagonizes PRC2 and leads to a local loss of H3K27me3, but that domains of high 360! H3K27me3 levels are maintained even upon oncohistone incorporation (Fig. 2, 3 enzyme, but the more pre-existing methylation is located in close proximity, the easier it is for 381! the PRC2 to resume its activity and propagate the mark.

382!
The proposed model of PRC2 regulation by the local levels of oncohistone and 383! H3K27me3 appears simpler than the chromatin changes described for mouse and human cell 384! culture models. This may be explained by the findings that DNA methylation, which is largely 385! absent in C. elegans, also influences PRC2 activity in human cells 56 . H3K27me3 can also spread 386! in "far-cis" via long range contacts in human cells 57 . Such long-range interactions appear less 387! common in C. elegans nuclei 58 . The simplicity of the spatial organization of the worm genome 388! and the absence of DNA methylation therefore allowed us to establish the local dependencies 389! between pre-existing methylation, oncohistone incorporation and PRC2 activity. greatly differ between cell types 59,60 . Therefore, the resulting H3K27me3 landscapes upon 400! mutation of H3.3 will also differ, and it is tempting to speculate that these differences will result 401! in different, cell type-dependent patterns of PRC2 inhibition, and cell type-specific phenotypes.  (Fig. 2)

Tumorigenesis is driven by specific pathways downstream of chromatin changes 433!
The chromatin reorganization caused by oncohistone incorporation and ectopic PRC2 regulation 434! results in a global misregulation of the transcriptome in mutant cells. We found that mutated C. 435! elegans germ cells showed ectopic expression of several hundred genes, which fall into diverse 436! categories such as kinases and ion transporters (Fig. 4). Similarly, transcriptional profiling of 437! tumor-derived tissues and single cells showed that many cancer-related pathways were 438! misregulated 9,38 . To single out key factors that link the chromatin changes and the downstream 439! phenotypes, we took advantage of the powerful C. elegans genetics for unbiased screening. 440! Surprisingly, we were able to identify a single point mutation in the C. elegans JNK homologue 441! KGB-1 that significantly rescued the aberrant replication phenotype, making misregulation of the 442! JNK pathway an important contributor to the replicative fate induced by the oncohistone (Fig. 5).

443!
The JNK pathway is a known regulator of calcium levels in C. elegans gonads and linked to 444! calcium signaling in human cells 41,42,65  and replicative fate has been established in many types of gliomas 45-47 . We found that inhibition 458! of JNK suppresses proliferation of cells derived from K27M tumors more strongly than cells 459! derived from non-K27M gliomas (Fig. 6)