LATS2 Positively Regulates Polycomb Repressive Complex 2

LATS2, a pivotal Ser/Thr kinase of the Hippo pathway, plays important roles in many biological processes. LATS2 also function in Hippo-independent pathway, including mitosis, DNA damage response and epithelial to mesenchymal transition. However, the physiological relevance and molecular basis of these LATS2 functions remain obscure. To understand novel functions of LATS2, we constructed a LATS2 knockout HeLa-S3 cell line using TAL-effector nuclease (TALEN). Integrated omics profiling of this cell line revealed that LATS2 knockout caused genome-wide downregulation of Polycomb repressive complex 2 (PRC2) and H3K27me3. Cell-cycle analysis revealed that downregulation of PRC2 was not due to cell cycle aberrations caused by LATS2 knockout. Not LATS1, a homolog of LATS2, but LATS2 bound PRC2 on chromatin and phosphorylated it. LATS2 positively regulates histone methyltransferase activity of PRC2 and their expression at both the mRNA and protein levels. Our findings reveal a novel signal upstream of PRC2, and provide insight into the crucial role of LATS2 in coordinating the epigenome through regulation of PRC2.


Supplementary Results and Discussion for Fig C-J Tissue or context specificity of LATS2-dependent H3K27me3 targets
We attempted to explore the potential physiological relevance of the LATS2-responsive PRC2 signal. To more precisely evaluate the LATS2-dependent epigenetic signature, we analyzed ChIP-seq and RNA-seq data based on properties of their promoters. To this end, we categorized promoters of protein-coding genes into three classes based on their CpG content [1]: 1) High-CpG promoter (HCP), associated with both housekeeping genes and other genes with complicated expression patterns during development. This class often includes genes with bivalent states, i.e., those that possess both H3K27me3 and H3K4me3 marks in their promoters; 2) Low-CpG promoter (LCP), associated with genes that are under control of tissue-specific transcription factors and are often repressed by DNA methylation. This class exhibits tissue-specific expression pattern during differentiation; and 3) Intermediate-CpG promoter (ICP), i.e., genes which are neither HCPs nor LCPs. (HCP, LCP and ICP) (Fig Ca). Promoter analysis of potential LATS2-dependent epigenetic signatures, i.e., 'H3K27me3-loss' module from HeLa-S3 study, revealed that this module significantly correlated with the LCP class (p = 5.2E-117) (Fig Cb). Consistent with this, fluctuations in gene expression were larger in the LCP class than in HCP or ICP (Fig Cc). LCP genes are associated with tissue-specific transcription programs during differentiation. Indeed, the LATS2-responsive PRC2 target genes in HeLa-S3 cells correlated with genes harboring the H3K27me3 mark specifically in human ES cells (vs. human fibroblasts, used as a reference sample) (p = 2.0E-173) (Fig Cd).
Next, to reveal the tissue or context specificity of LATS2-dependent H3K27me3, we explored tissue-specific expression patterns of 'H3K27me3-loss' genes using Body Atlas tools of NextBio statistical platform. Intriguingly, LATS2-dependent H3K27me3 target genes were expressed at relatively low levels in embryonic cells, but were overexpressed in the nervous system (Fig Da and Db). These observations indicate that the expression pattern of the LATS2-PRC2 signal is converted to an active state during neurogenesis. Indeed, the expression profile of LATS2 KO HeLa-S3 cells (relative to wild type) positively correlated with that of differentiated neurons relative to neural stem cells (Fig Dc and S5 Table), i.e., the expression profile of cells with LATS2 correlated with that of neural stem cells. Further analysis revealed that the expression of genes with LATS2-dependent H3K27me3 marks in HeLa-S3 was highly induced during the differentiation of ES cells into mature neuron cells (p < 0.001) (Fig Dd). The tissue-specific induction of 'H3K27me3-loss' genes in neural differentiation process was further supported by another GSEA using previous microarray studies for various in vitro differentiation systems: significant positive enrichment of 'H3K27me3-loss' genes was observed in only the neural differentiation experiment (one of six differentiation systems, Fig E). These results suggest that potential LATS2-responsive H3K27me3 targets are repressed in dedifferentiated cells, and then gradually induced in the proneural stage.

Association between LATS2 and PRC2 in brain tumorigenesis
Because Lats2 may contribute (via PRC2) to maintenance of the dedifferentiated state in the nervous system, we hypothesized that LATS2 can contribute to tumorigenesis in neural cells. First, to characterize the expression of LATS2 in various types of cancer, we visualized the relative expression level of the LATS2 gene using 8,415 RNA-seq datasets from TCGA. In many types of cancer, LATS2 tended to be expressed at low levels in tumor samples relative to normal solid tissue, consistent with the known tumor-suppressive properties of LATS2. By contrast, in GBM, LATS2 was expressed at higher levels in tumor samples than in normal tissue (Fig Fa, upper panel). To characterize the expression of LATS2 in detail, we examined array-based transcriptome data from 583 samples, including ten normal brain samples, from the TCGA portal site.
In this dataset, we again observed significantly higher LATS2 expression in primary tumor samples (Fig Fb, left panel). Common GBM patients can be divided into four expression subtypes [2]; LATS2 expression was elevated in all but the proneural subtype (Fig Fc), which has a transcription signature related to neurogenesis and exhibits a more differentiated expression pattern than the other subtypes. Intriguingly, LATS1 did not exhibit a similar expression pattern in GBM (Fig Fa,  We next performed GSEA to determine whether genes potentially targeted by H3K27me3 or genes related to neurogenesis are differentially expressed in LATS2-low GBM (i.e., whether positive effects of LATS2 on PRC2-mediated repression of genes for neural differentiation were disturbed). To obtain gene expression profiles that depend on LATS2, we first divided GBM cases into two groups: those with higher expression of LATS2 than the median for normal brain (n = 411, 'LATS2-high') and those with lower-than-median expression (n = 54, 'LATS2-low'). We performed GSEA using this profile. Consistent with our hypothesis, known H3K27me3 targets were highly expressed in LATS2-low GBM samples (Fig Ga). Moreover, some neuronspecific transcripts were positively associated with LATS2-low GBM samples (Fig Gb), whereas the stem-cell signature was enriched in LATS2-high GBM samples (Fig Gc). Furthermore, we discovered an association between LATS2 expression and clinical prognosis. Kaplan-Meier curves and estimates of survival data revealed that patients with higher-than-median LATS2 expression exhibit poorer clinical prognoses than patients with lower-than-median LATS2 expression (n = 466; p = 0.00512) (Fig Gd). Importantly, the difference in the survival probability simply reflects the difference between the proneural subtype and the other subtypes. We observed a similar tendency in survival probability in an analysis using only proneural samples (n = 115) (Fig Ge). These results support the impact of the LATS2 signal via PRC2 in both development and dedifferentiation of the nervous system.

Transcriptome analysis of Lats2 KO MEFs
Based on the observation and correlation of LATS2 with PRC2 related signals in nervous system, we next examined if LATS2 plays fundamental roles through PRC2 in this tissue. Indeed, Lats2-deficient mice exhibit embryonic lethality due to a defect in development of the central nervous system [3]. The insight of the association of Lats2 with PRC2 during neurogenesis is further supported by a previous study in Drosophila: a mutant of Wts, the Drosophila homolog of Lats1/2, phenocopies the effect of Polycomb group (PcG) mutants on dendrite maintenance [4]. To examine the functional link between Lats2 and PRC2 in mice, we next performed transcriptome analysis of Lats2 KO MEFs (Fig Ha and Hb). GSEA revealed that potential PRC2-regulated genes were elevated in Lats2 KO MEFs (p-value < 0.001) (Fig Hc and Hd). Furthermore, the up-regulated genes in Lats2 KO MEFs were significantly correlated with those of upregulated genes in Eed KO ES cells during differentiation (Fig He). These results suggest that Lats2 depletion in MEFs causes dysregulation of PRC2 function.
We further investigated whether the dysregulation of repressive epigenetic mechanisms caused by Lats2 KO is related to differentiation processes or maintenance of stemness.
We performed GSEA on the expression profile of Lats2 KO MEFs, using signatures of differentiation or stemness defined in a previous meta-analysis of human ES cells and differentiating stem cells [5]. In agreement with the study, Lats2 KO was associated with a significant enrichment of genes that are expressed during developmental processes (Fig Ia and Ib). This result suggests that Lats2 may repress genes involved in development (via PRC2). Indeed, some well-known transcripts associated with differentiation, such as those of homeotic genes, were up-regulated in Lats2 KO MEFs (Fig Id and Ie).
The canonical Hippo pathway contributes to various processes related to both development and dedifferentiation. The significant overlap of up-regulated genes in Lats2 KO MEFs with genes for differentiation that we observed Figure S8, may be simply a result of canonical Hippo-YAP/TAZ signaling, which is caused by activation of Yap/Taz due to Lats2 depletion (a schematic; Fig Ja). To determine whether the Lats2 KO signature was Yap/Taz-dependent, we compared gene expression in Lats2 KO MEFs and rapidly growing wild type MEFs in log-phase, which have active Yap/Taz (validated in Fig Jb). Gene Ontology (GO) analysis of ontological terms related to differentiation processes revealed that 'Hippo-OFF' cells, i.e., growing cells, in which the intrinsic Hippo pathway is turned off and Yap/Taz is active, show associations with differentiation processes less significantly than Lats2 KO cells. (Fig Jc, S6 and S7 Table). This result indicates that the association of up-regulated genes in Lats2 KO MEFs with differentiation does not simply reflect the output of the canonical Hippo pathway.
In summary, a series of comparative analyses of LATS2-responsive H3K27me3 targets, cancer genomics data and Lats2 KO MEFs present a possibility of an attractive novel function of LATS2 kinase in specific tissues and cellular contexts, especially neurogenesis, which will be clarified in our future research.      (a) Box-and-whisker plots of mRNA expression of LATS2 (top) and LATS1 (bottom) in normal tissues and human cancers. mRNA expression data of LATS2 and LATS1 for various cancers and normal tissue samples were obtained from TCGA pan-cancer cohorts. Cancers with at least one normal sample were analyzed and visualized. (b) Box-andwhisker plots of LATS2 (left) and LATS1 (right) mRNA expression in normal brain and primary GBM. LATS2 mRNA expression data for GBM and normal tissue samples were obtained from level 3 preprocessed expression data from Agilent 244K custom gene-expression G4502A_07_2 microarrays. Statistical significance of differences between normal brain and primary tumor was evaluated by Wilcoxon rank-sum test. (c) Box-and-whisker plots of LATS2 mRNA expression in normal brain and predefined gene-expression subtypes of GBM. Statistical significance between normal brain and each subtype was evaluated by paired Wilcoxon rank-sum test.       (c) Enrichment analysis for canonical pathways related to differentiation processes. DEGs (≥2-fold, pvalue <0.05) in Lats2 KO MEFs and in a Hippo-inactive state (growing in log-phase/high-density culture of wild type MEFs) were subjected to NextBio analysis. Lats2 KO profile shows associations with differentiation processes more significantly than Yap/Taz-dependent profile. See also supplementary S6 and S7 Table.