EGFR/SRC/ERK-stabilized YTHDF2 promotes cholesterol dysregulation and invasive growth of glioblastoma

Glioblastoma (GBM) is the most common type of adult malignant brain tumor, but its molecular mechanisms are not well understood. In addition, the knowledge of the disease-associated expression and function of YTHDF2 remains very limited. Here, we show that YTHDF2 overexpression clinically correlates with poor glioma patient prognosis. EGFR that is constitutively activated in the majority of GBM causes YTHDF2 overexpression through the EGFR/SRC/ERK pathway. EGFR/SRC/ERK signaling phosphorylates YTHDF2 serine39 and threonine381, thereby stabilizes YTHDF2 protein. YTHDF2 is required for GBM cell proliferation, invasion, and tumorigenesis. YTHDF2 facilitates m6A-dependent mRNA decay of LXRA and HIVEP2, which impacts the glioma patient survival. YTHDF2 promotes tumorigenesis of GBM cells, largely through the downregulation of LXRα and HIVEP2. Furthermore, YTHDF2 inhibits LXRα-dependent cholesterol homeostasis in GBM cells. Together, our findings extend the landscape of EGFR downstream circuit, uncover the function of YTHDF2 in GBM tumorigenesis, and highlight an essential role of RNA m6A methylation in cholesterol homeostasis. EGFR is frequently constitutively active in glioblastoma (GBM). Here, the authors show that EGFR induces YTHDF2 protein stabilization, which reduces cholesterol homeostasis through an RNA m6A methylation dependent molecular mechanism to promote GBM tumourigenesis.

G lioblastomas (GBM; World Health Organization grade IV glioma) are a devastating type of brain tumor, with a median survival of 14 months, regardless of treatment 1 . The tumor cells that invade adjacent brain tissue and contribute to the intracranial dissemination of GBM, represent a major clinical problem. The proliferation and invasion capacities of glioblastomas are a consequence of the interplay of multiple layers of altered molecular regulations. The most prevailing genetic alterations are the amplification and mutation of EGFR that occur in more than 50% of GBMs 1 . EGFRvIII, lacking part of the extracellular region, is a frequent activating oncogenic mutant 2 . EGFR activation by amplification, mutation, and ligandbinding in an autocrine or paracrine manner can accelerate GBM formation and is associated with poor prognosis. EGFR signaling has been reported to confer global effects on the modulation of multiple pathways in GBM. However, the effects of EGFR activation on mRNA modification are less known. N 6 -methyladenosine (m 6 A) modification is the most prevalent internal modification existing on eukaryotic mRNA 3 . We and others have recently shown that aberrant mRNA m 6 A modifications actively assist the tumorigenicity of glioblastoma stem cells 4,5 . m 6 A-modified mRNA can be preferentially recognized by the members of the human YTH domain family, the 'reader' proteins, such as YTHDF2 6,7 . Among the mammalian YTH family proteins, YTHDF2 has been shown to destabilize m 6 Amarked transcripts through deadenylation by the CCR4-NOT complex, or through endoribonucleolytic cleavage, with the help of HRSP12-RNase P/MRP [7][8][9] . Depletion of Ythdf2 leads to mouse embryonic lethality due to the failure of neural stem/progenitor cell proliferation and differentiation 10,11 . Despite the important roles of YTHDF2 in brain development and function, it is unknown whether this reader is presented in GBM and affects GBM propagation.
Dysregulated cholesterol metabolism by cancer epigenetic regulators is another hallmark of GBM 12 . Cholesterol metabolism is under the transcriptional control of sterol regulatory elementbinding proteins (SREBPs) and liver X receptors (LXRs). SREBPs promote cholesterol synthesis and enhance the uptake of extracellular cholesterol 13 . EGFR mutation in GBM has been shown to increase the PI3K-dependent activation of SREBP1 14 . LXRs induce cholesterol efflux by regulating the expression of apolipoprotein E (ApoE), a major apolipoprotein in the CNS that mediates the transport of cholesterol, and its transporters ABCA1 and ABCG1. LXRs also suppress cholesterol uptake by enhancing the degradation of LDLR 15,16 . Although LXR agonists have been reported to cause GBM regression 12,14 , the regulatory mechanisms for LXRs expression in GBM remain unclear.
In the current study, we investigate m 6 A YTH readers in GBM, and observe a correlation between increased expression of YTHDF2 and decreased survival of glioma patients. We identify a regulatory role of EGFR activation in YTHDF2 overexpression and elucidate the underlying mechanisms in GBM cells. We also determine the biological consequences of EGFR-mediated induction of YTHDF2 on the tumorigenicity of GBM cells. Mechanistically, we uncover that YTHDF2 accelerates m 6 Adependent LXRA mRNA degradation, thereby promoting cholesterol dysregulation in GBM cells. YTHDF2 also promotes the mRNA degradation of HIVEP2 that is critical to GBM tumorigenesis.

Results
YTHDF2 expression is increased in GBMs and correlates with poor prognosis. After showing the vital role of m 6 A erasers in GBM development 5 , we are interested in the roles of m 6 A readers, especially those of YTH domain family proteins, which impact diverse biological processes, in glioma tumorigenesis. We first queried The Cancer Genome Atlas (TCGA) 17,18 , REMBRANDT 19 , French 20 , Kawaguchi 21 , and Paugh 22 datasets, and found the expression of YTHDF2, but not YTHDF1, YTHDF3, YTHDC1, or YTHDC2 correlates with poor overall survival of glioma patients in all of the datasets (Fig. 1a-c, Supplementary Fig. 1a-b). Moreover, the expression of YTHDF2 in GBMs increased as compared with normal brains in TCGA and REMBRANDT datasets (Fig. 1d).
We then examined the protein levels of YTHDF2 in human glioma specimens. YTHDF2 shows higher protein expression in tumor relative to adjacent non-tumor tissue ( Supplementary  Fig. 1c). Notably, the expression of YTHDF2 protein is significantly increased in GBM and correlates with tumor grade (Fig. 1e, f). In cultured glioma tumor cell lines and normal human astrocytes, YTHDF2 protein is highly expressed in GBM cells, especially in GBM-derived stem cells (GSCs), as compared with lower-grade glioma cells (Hs683 and SW1783) and normal human astrocytes ( Fig. 1g and Supplementary Fig. 1d). Moreover, YTHDF2 protein expression is reduced upon induced differentiation of GSCs ( Supplementary Fig. 1e). Collectively, these results suggest that YTHDF2 might be involved in the malignancy of glioma.
YTHDF2 regulates GBM cell proliferation, invasion, and tumorigenesis. To examine the role of YTHDF2 in GBM tumorigenesis, we used two distinct short hairpin RNAs targeting the 3′UTR (shYTHDF2#1) or the CDS (shYTHDF2#2) of YTHDF2 mRNA to ablate YTHDF2 expression in GSC11 and GSC7-2 cells (Supplementary Fig. 2a). Loss of YTHDF2 in the GSC cells inhibited cell proliferation, DNA replication, and invasiveness ( Fig. 2a-c, Supplementary Fig. 2b, c) in vitro, whereas overexpression of YTHDF2 in LN229 cells did the opposite (Supplementary Fig. 2d-g). Moreover, the ablation of YTHDF2 in the cells reduced tumor growth of GSC11 and GSC7-2 cells (Fig. 2d, Supplementary Fig. 2h, i) as determined by in vivo intracranial tumor assay. Immunohistochemical staining of mouse brain sections revealed that YTHDF2 ablation decreased Ki-67 expression and increased cleaved Caspase-3 expression (Fig. 2e, f, Supplementary Fig. 2j). Moreover, YTHDF2-depleted tumors displayed distinct margins with significantly reduced invasive fingers of tumor (Fig. 2g). Furthermore, depletion of YTHDF2 prolonged the survival of tumor-bearing mice (Fig. 2h, Supplementary Fig. 2k). Most importantly, adding back a shYTHDF2#1-resistant YTHDF2 to the YTHDF2-depleted GSCs rescued the malignant phenotypes of GSCs ( Fig. 2d-h, Supplementary Fig. 2h-k), suggesting that YTHDF2 is important for GBM tumor growth and invasion. In contrast, overexpression of YTHDF2 in LN229 cells enhanced the tumor growth and invasiveness and reduced the survival of tumor-bearing mice ( Fig. 2i-k). Collectively, these results suggest that YTHDF2 regulates proliferation, invasion, and tumorigenicity of GBM cells.
EGFR/SRC/ERK signaling sustains YTHDF2 expression. To determine which signaling pathway drives the upregulation of YTHDF2 in GBM, the most commonly aberrantly activated pathways that include AKT, FGFR, PDGFRα/β, WNT, TGFβR1, EGFR, SRC, and ERK1/2 were tested by using inhibitors MK-2206, AZD4547, CP673451, Wnt-C59, SB-431542, gefitinib, SU6656, and SCH772984, respectively. We found the kinase inhibitors for EGFR, SRC, and ERK1/2 consistently suppressed YTHDF2 protein expression in GSC11 and GSC7-2 cells in a 24-h period (Fig. 3a, Supplementary Fig. 3a). Further analysis showed that either activation of wild-type EGFR by EGF or induced expression of EGFRvIII could promote YTHDF2 expression (Fig. 3b, c). Interestingly, EGFR inhibition by gefitinib or depletion by siRNA reduced the expression of YTHDF2 protein without affecting the mRNA levels ( Supplementary Fig. 3b, c), suggesting a post-transcriptional regulation involved in the increase of YTHDF2. Moreover, inhibition of EGFR by gefitinib in GSC11 cells accelerated degradation of YTHDF2 protein (Fig. 3d), whereas induction of EGFRvIII expression in LN229/ EGFRvIII tet-on (doxycycline-inducible EGFRvIII) cells by doxycycline (Dox) stabilized YTHDF2 (Fig. 3e), suggesting that EGFR plays a vital role in YTHDF2 protein stabilization. Notably, hyperactivation of EGFR, which is widely found in most of GBMs including the GSCs in our current study, can activate Src 23 Fig. 3e). However, Src Y527F failed to restore the expression of YTHDF2 upon ERK1/2 inhibition, suggesting an essential role of ERK1/2 ( Supplementary Fig. 3e). In agreement of this finding, EGFRvIII expression failed to stabilize YTHDF2 when ERK1/2 was inhibited by SCH772984 (Fig. 3g). Together, these results indicate that EGFR/SRC/ERK cascade is responsible for the elevated YTHDF2 protein in GBM cells.
ERK1/2 phosphorylates YTHDF2 at serine 39 and threonine 381 to stabilize YTHDF2. We found that endogenous YTHDF2 can associate with active ERK1/2 determined by reciprocal immunoprecipitation assays, which was further enhanced by the addition of EGF (Fig. 3h, i, Supplementary Fig. 3f). To determine whether this association also occurs in vivo, we performed a proximity ligation assay (PLA) that confirmed the interaction between YTHDF2 and ERK1/2, with an enhanced effect by EGF stimulation or EGFRvIII expression which is consistent with our in vitro observation (Fig. 3j). Next, we asked whether ERK1/2 phosphorylates YTHDF2 since the ERK1/2 kinase inhibitor SCH772984 suppressed YTHDF2 expression (Fig. 3f). As predicted by GPS 3.0 (http://gps.biocuckoo.cn/) ( Supplementary  Fig. 3g), several serine and threonine sites in YTHDF2 may be potentially phosphorylated by ERK1/2. Among these serine and threonine sites, S39 and T381 phosphorylations have been documented in PhosphoSitePlus ® which extracts highthroughput data from published papers (https://www. phosphosite.org/homeAction.action) ( Supplementary Fig. 3h).
In addition, we found that S39 and T381 of YTHDF2 are conserved across different mammalian species ( Supplementary  Fig. 3i), suggesting that they are most likely the ERK1/2 phosphorylation sites. To examine this hypothesis, we conducted an in vitro kinase assay using recombinant ERK1 and GST purified YTHDF2. We found that wild-type YTHDF2 could be directly phosphorylated by ERK1, whereas both S39A and T381A mutants showed compromised phosphorylation (Fig. 4a).
Notably, S39A/T381A double mutation completely abrogated phosphorylation by ERK1 (Fig. 4a). In line with these observations, S39A or T381A mutant alone only slightly attenuated the association between YTHDF2 and ERK1/2, but the S39A/T381A double mutant results in a complete loss of the association ( Supplementary Fig. 4a). Using the panphosphothreonine and -phosphoserine antibodies, we also confirmed that S39 or T381 of YTHDF2 was phosphorylated in GSC11 cells ( Supplementary Fig. 4a). On the other hand, inhibition of the ERK downstream kinase MNK1 by CGP 57380 failed to affect YTHDF2 phosphorylation at these sites (Supplementary Fig. 4a), suggesting YTHDF2 as a direct target of ERK1/2. To assess the biological effect of the YTHDF2 phosphorylation, we measured the exogenous FLAG-tagged YTHDF2. S39A/ T381A mutant displayed an accelerated degradation and a shortened half-life from 13.45 to 6.71 h, compared with the wild type (Fig. 4b). Moreover, when S39A/T381A mutant was expressed to a similar amount as wild-type YTHDF2 (judged by the similar mRNA levels, Supplementary Fig. 4b), it displayed minimal effects on the promotion of invasiveness of LN229 cells (Fig. 4c), proliferation of GSC11 and GSC7-2 cells (Fig. 4d, e), and tumor growth (Fig. 4f, Supplementary Fig. 4c) compared with wild-type YTHDF2. Furthermore, we found that EGFR inhibition in GSC11 cells decreased threonine and serine phosphorylations of YTHDF2, while Dox-induced EGFRvIII expression in LN229 cells increased the phosphorylations (Supplementary Fig. 4d). Finally, a correlation between YTHDF2 protein expression and the expressions of p-EGFR (Y1173), p-Src (Y419), and p-ERK1/2 in human gliomas also suggested the importance of this EGFR/ SRC/ERK/YTHDF2 cascade in tumor development (Supplementary Fig. 4e, f).
Identification of YTHDF2 targets by high-throughput RNAseq and RIP-seq. It has been reported that YTHDF2 is a "reader" for m 6 A modification, and executes its function by destabilizing its target mRNAs in the cytoplasm or preventing demethylation of 5′UTR of stress-induced transcripts in the nucleus 7 . To investigate the downstream targets of YTHDF2, we performed RNA immunoprecipitation sequencing (RIP-seq) in GSC11 cells for YTHDF2 associated RNAs, which identified 3940 significantly enriched transcripts (Fig. 5a, and Supplementary Data 1). Then, we retrieved the m 6 A methylomes of GSC11 cells by MeRIP-seq from our previous study 5 and found that 3518 genes overlapped in both MeRIP-seq and RIP-seq (Supplementary Data 2), regarding as m 6 A-dependent targets of YTHDF2. In parallel, we performed a gene expression analysis by RNA-seq that showed 1412 differentially expressed genes in YTHDF2-knockdown GSC11 cells using two different siRNAs ( Fig. 5b and Supplementary Data 3). Since YTHDF2 functions in the clearance of m 6 A-modified RNAs, we combined the results from RNA-seq and RIP-seq, and found 243 transcripts associated with YTHDF2 were upregulated in YTHDF2-knockdown GSC11 cells as compared with the control (Supplementary Data 4). Among them, the Fig. 1 YTHDF2 is highly expressed in glioblastoma and predicts poor prognosis of glioma patients. a Kaplan-Meier overall survival plot showing survival rates for lower-grade glioma and GBM patients having YTHDF2 low expression (blue), and high expression (red) in the TCGA PanCancer Atlas dataset (http://www.cbioportal.org) (two-sided log-rank test). b Kaplan-Meier overall survival plot showing survival rates for glioma patients having YTHDF2 low expression (blue), and high expression (red) in REMBRANDT dataset (two-sided log-rank test). c Kaplan-Meier overall survival plot showing survival rates for glioma patients having YTHDF2 low expression (blue), and high expression (red) in French dataset (two-sided log-rank test). d Profile of YTHDF2 mRNA expression in normal, GBM, or grade I-III glioma patients in TCGA and REMBRANDT datasets. e Analysis of YTHDF2 protein expression in 2 TMAs with 14 normal cores, 10 grade I astrocytomas cores, 47 grade II astrocytomas cores, 33 grade III astrocytomas cores, and 34 GBM cores; and in 36 GBM tissues, using immunohistochemical analysis. Scale bar = 100 μm. Representative image of two independent experiments. f Summary of YTHDF2 protein expression profile in (e). IHC staining score 0-1, low; 2-8, medium; 9-12, high (Kruskal-Wallis test and the post hoc Dunn's multiple comparisons test). g Western blotting of YTHDF2 in glioma cells (Hs683, SW1783), GBM cells (T98G, U87 MG, U251 MG, LN229), and GSCs (GSC11, GSC17, GSC20, GSC23, GSC7-2, GSC6-27). Representative blot of two independent experiments. Source data are provided as a Source data file. expression levels of ATOH8, C11ORF71, CELSR3, DGCR5, HIVEP2, HSPA12A, ITPK1, KIAA1755, LXRA, PCDHA1, PRDM11, PRNP, RBSN, RNF103, RWDD2A, SIRPA, SIRT3, USP54, and ZBTB47 were negative correlated (spearman correlation <−0.3) with YTHDF2 in TCGA LGG and GBM datasets (Fig. 5c, Supplementary Data 5), suggesting that these genes could be downstream effectors of YTHDF2 in glioma. Thus, we searched the literature for the functions of these 19 genes. LXRA (synonym: NR1H3) has been reported to be a player in regulating glioblastoma cell invasion 25 , and LXR agonist has been shown to promote glioblastoma cell death 12,14 . HIVEP2 level is known to decreases in higher grade of glioma and most significantly in GBM. Transfection of HIVEP2 into glioma cells inhibited their cell growth 26   glioma have not been reported so far. Collectively, the above information suggested that LXRA and HIVEP2 could be direct targets of YTHDF2 which related to GBM cell proliferation and invasion (Fig. 5d). Indeed, the mRNAs of LXRA and HIVEP2 have several YTHDF2 binding sites (Fig. 5e, f), and the expression of LXRA and HIVEP2 negatively correlated with YTHDF2 expression in TCGA data ( Supplementary Fig. 5a). We therefore focused on LXRA and HIVEP2 for their potential roles in mediating YTHDF2-induced GBM proliferation and invasion. Next, we performed YTHDF2 cross-linking immunoprecipitation (CLIP) assay to confirm the direct interactions between LXRA and HIVEP2 mRNAs and YTHDF2 protein. LXRA and HIVEP2 mRNAs could interact with YTHDF2 protein in both GSC11 and GSC7-2 cells (Fig. 5g, Supplementary Fig. 5b), suggesting that LXRA and HIVEP2 are direct targets of YTHDF2. Moreover, knockdown of YTHDF2 increased mRNA levels of LXRA and HIVEP2 in GSC11 and GSC7-2 cells, validating our RNA-seq results (Fig. 5h, Supplementary Fig. 5c). In contrast, knockdown of YTHDF2 did not affect mRNA levels of LXRB (synonym: NR1H2), a close family member of LXRA and the predominant LXR subtype in GBM 12 (Fig. 5h, Supplementary  Fig. 5c). On the other hand, overexpression of YTHDF2 decreased LXRA and HIVEP2 mRNA levels in LN229 cells without affecting LXRB (Fig. 5i). Importantly, in the TCGA glioma dataset, we found that a low LXRA and HIVEP2 coexpression predicts worse prognosis of glioma patients compared with a high co-expression (Fig. 5j). In patients with a high YTHDF2 expression, the LXRA and HIVEP2 co-expression tends to be low and vice versa, suggesting an indeed clinically important regulation (Fig. 5k).
YTHDF2 downregulates LXRA and HIVEP2 through m 6 Adependent mRNA decay. We sought to investigate the mechanism by which YTHDF2 regulates LXRA and HIVEP2 expression. The expressions of LXRα and HIVEP2 were low in GSC11 cells, while YTHDF2 knockdown increased LXRα and HIVEP2 protein levels without affecting LXRβ (Fig. 6a, Supplementary Fig. 6a). LXRα and HIVEP2 protein levels were reduced by induction of EGFRvIII expression in LN229/EGFRvIII tet-on cells (Fig. 6b). Moreover, the decrease of LXRα and HIVEP2 protein levels in the EGFRvIII expressing LN229/EGFRvIII tet-on cells was rescued by knockdown of YTHDF2 (Fig. 6b).
YTHDF2 consists of a C-terminal YTH domain, which specifically binds to m 6 A-containing RNA. We had mapped the m 6 A methylomes of GSC11 cells by MeRIP-seq in a previous study 5 . Through analyzing the MeRIP-seq data, we found that LXRA and HIVEP2 transcripts have several significant m 6 A peaks (Fig. 6c, d). Moreover, knockdown of YTHDF2 enriched m 6 A methylated LXRA and HIVEP2 mRNAs ( Supplementary Fig. 6b), which is consistent with the m 6 A-dependent mRNA decay function of YTHDF2 7 . According to the data of RIP-seq and MeRIP-seq, there are peaks strongly located in 3′UTR of the mRNA of LXRA and HIVEP2. We then constructed luciferase reporter containing 3′UTR of LXRA or HIVEP2. We found that Dox-induced EGFRvIII expression decreased the activity of 3′ UTR of LXRA and HIVEP2, while YTHDF2 knockdown could reverse the effect ( Supplementary Fig. 6d). Furthermore, knockdown of YTHDF2 inhibited mRNA decay of LXRA and HIVEP2 as we measured the mRNA half-life of these genes by inhibition of transcription with actinomycin D in GSC11 cells (Fig. 6e, f). Thus, we determined whether YTHDF2 regulated LXRA and HIVEP2 expression through m 6 A methylation. Given that tryptophan residues at position 432, 486, and 491 located in YTH domain have been shown to be critical for m 6 A recognition of YTHDF2 27,28 , and are conserved across multiple vertebrate species and YTH family members ( Supplementary Fig. 6c), we then generated m 6 A-recognition defective YTHDF2 (YTHDF2 MUT) through W432A, W486A, and W491A triple mutations ( Supplementary Fig. 6c). We found that m 6 A-recognition defective YTHDF2 failed to enrich the mRNAs of LXRA and HIVEP2 (Fig. 6g). As a result, expression of the YTHDF2 mutant failed to decrease the levels of mRNAs and proteins of LXRA and HIVEP2 compared with wild-type YTHDF2 (Fig. 6h, i). To examine whether YTHDF2 regulates the expression of LXRA and HIVEP2 by mediating mRNA decay, we measured the mRNA half-life of these genes in LN229 cells. Wild-type YTHDF2 accelerated mRNA decay of LXRA and HIVEP2 indicating by the shorter half-life as compared with the control (Fig. 6j, k), whereas m 6 A-recognition defective YTHDF2 has no effect on mRNA decay of LXRA and HIVEP2 (Fig. 6j, k). In addition, YTHDF2 knockdown failed to regulate the mRNA expression of LXRA and HIVEP2 in m 6 A writer METTL14-depleted GSC11 cells (Supplementary Fig. 6e). These results suggest that YTHDF2 regulates LXRA and HIVEP2 dependent on the binding of YTHDF2 to m 6 A-modified mRNA. Moreover, S39A/T381A mutations of YTHDF2 compromised YTHDF2's ability to accelerate the degradation of LXRA and HIVEP2 mRNAs in YTHDF2depleted GSC11 cells ( Supplementary Fig. 6f), implying EGFR/ SRC/ERK signaling might affect these m 6 A-dependent mRNA decays. Indeed, inhibition of EGFR, SRC, or ERK could increase mRNAs of LXRA and HIVEP2 (Fig. 6l). In addition, EGF treatment in GSC11 cells or Dox-induced EGFRvIII expression in LN229/EGFRvIII tet-on cells decreased the mRNA expression of LXRA and HIVEP2 (Fig. 6m, n), whereas knockdown of YTHDF2 reversed the effects of EGF and EGFRvIII (Fig. 6m, n). Furthermore, in the YTHDF2-depleted GSC11 cells with EGF b Proliferation of shCtrl and shYTHDF2 GSC11 cells was assessed by 5-bromo-2'-deoxyuridine (BrdU) incorporation for 3 h. The percentage of BrdU positive cells was quantified. c In vitro invasion assay for shCtrl and shYTHDF2 GSC11 cells. Representative invasion images of the cells (left panel) and quantitation of relative cell invasion (right panel). Scale bar = 50 μm. d Nude mice intracranial tumor assay using shCtrl, shYTHDF2, and shYTHDF2 plus shRNA-resistant-YTHDF2 GSC11 cells. Brain sections stained with hematoxylin and eosin (H&E) show representative tumor xenografts. Tumor volumes were calculated. e, f Quantification of Ki-67 (e) or cleaved Caspase-3 (f) positive cells in brain tumor sections of shCtrl, shYTHDF2, and shYTHDF2 plus shRNA-resistant-YTHDF2 GSC11 cells. g In vivo invasion assay for shCtrl, shYTHDF2, and shYTHDF2 plus shRNA-resistant-YTHDF2 GSC11 cells. Representative H&E staining showing edges of the mice brain tumors (left panel), and the relative invasive fingers per tumor were counted (right panel). Scale bar = 200 μm. h Overall survival of mice injected with the indicated GSC11 cells. i Nude mice intracranial tumor assay using LN229 cells with vector only or with YTHDF2 overexpression. Brain sections stained with H&E show representative tumor xenografts. Tumor volumes were calculated. j In vivo invasion assay for LN229 cells with vector only or with YTHDF2 overexpression. Representative H&E staining showing edges of the mice brain tumors (left panel), and the relative invasive fingers per tumor were summarized (right panel). Scale bar = 200 μm. k Overall survival of mice injected with the indicated LN229 cells. For a-c, data are mean ± S.E.M., n = 3 biologically independent experiments (one-way ANOVA Tukey's post hoc test). For d-k, n = 10 mice per group examined over two independent experiments. For d-g, data are mean ± S.D. (one-way ANOVA Tukey's post hoc test). For h and k, twosided log-rank test is conducted. For i and j, data are mean ± S.D. (unpaired two-sided t test). Source data are provided as a Source data file. treatment, expression of wild-type YTHDF2 inhibited the mRNA expression of LXRA and HIVEP2, but m 6 A-recognition defective YTHDF2 did not have the effect ( Supplementary Fig. 6g). Thus, the above results indicate that YTHDF2 can mediate m 6 Adependent mRNA decay to restrain LXRA and HIVEP2 expression under the activation of EGFR/SRC/ERK signaling in GBM cells.
LXRα and HIVEP2 are functionally essential targets of YTHDF2 in cell proliferation, invasion, and cholesterol dysregulation of GBM cells. In order to determine the role of LXRα and HIVEP2 in mediating YTHDF2 biological effects on GBM proliferation and invasion, we knocked down LXRα and HIVEP2 individually or simultaneously in YTHDF2-depleted GSC11 cells. A single knockdown of LXRα or HIVEP2 partially rescued the proliferation, and the double knockdown completely restored cell viability (Fig. 7a, b, Supplementary Fig. 7a-d). In contrast, the invasion could only be rescued by knockdown of LXRα but not HIVEP2 and the double knockdown showed comparable effect as LXRα knockdown, suggesting that YTHDF2 promotes invasion through the downregulation of LXRα (Fig. 7c, Supplementary  Fig. 7e).
LXRα regulates cholesterol homeostasis through regulating uptake and efflux of cholesterol 29 . Cancer cells depend on cholesterol for their growth but activation of LXR blocks GBM growth by reducing intracellular cholesterol 12 Fig. 6a). Thus, we tested whether YTHDF2 could affect cholesterol homeostasis through the increased LXRα expression. First, we measured several cholesterol metabolism-related genes regulated by LXRα. Knockdown of YTHDF2 significantly upregulated the expressions of downstream targets of LXRα, including ABCA1, ABCG1, and APOE (Fig. 7d, Supplementary Fig. 7f), whereas overexpression of YTHDF2 decreased the expression of these genes (Fig. 7e). Consequently, YTHDF2 depletion decreased the level of cellular cholesterol, which could be rescued by knockdown of LXRα (Fig. 7f). YTHDF2 affects cellular cholesterol in an m 6 Adependent manner because overexpression of the wild-type but not m 6 A-recognition defective YTHDF2 increased cellular cholesterol (Fig. 7g). Moreover, cholesterol supplement could partially rescue cell viability inhibited by YTHDF2 knockdown (Supplementary Fig. 7g). LXR regulates cholesterol homeostasis by suppressing low-density lipoprotein (LDL, a carrier for cholesterol) uptake and promoting cholesterol efflux 21 . Thus, we examined LDL uptake in GSCs. YTHDF2 knockdown inhibited the uptake of LDL (Fig. 7h), and LXRα knockdown rescued LDL uptake (Fig. 7h, Supplementary Fig. 7h). Conversely, overexpression of wild-type YTHDF2, but not the m 6 Arecognition defective YTHDF2, increased LDL uptake (Fig. 7i, Supplementary Fig. 7i). Given that LXRα can also increase the efflux of cholesterol 22 , we thus examined the role of YTHDF2 on cholesterol efflux. Knockdown of YTHDF2 increased ApoA1 dependent cholesterol efflux and knockdown of LXRα could reverse the effect ( Supplementary Fig. 7j). These results suggest that YTHDF2 regulates cholesterol homeostasis through LXRα. It has been reported that LXR regulates cell invasion through APOE 18 . To examine whether APOE plays a role in YTHDF2mediated cell invasion, we depleted APOE gene from GSC11 YTHDF2-knockdown cells. We found that conditioned media from YTHDF2-depleted GSC11 cells reduced cell invasion, but the reduction was rescued by APOE depletion (Fig. 7j, Supplementary Fig. 7k), suggesting that YTHDF2 regulates cell invasion partially through the inhibition of APOE. Also, SSTR2, which is reported to have an anti-proliferative effect in cancer cells, is a downstream target of HIVEP2 23 . We thus examined the effect of SSTR2 knockdown in YTHDF2-mediated cell viability. We found that YTHDF2 regulated the expression of SSTR2 in GSC11, GSC7-2, and LN229 cells (Fig. 7k, l, Supplementary Fig. 7l). Moreover, SSTR2 knockdown partially rescued the reduction of cell viability caused by YTHDF2 depletion ( Supplementary  Fig. 7m), suggesting that YTHDF2 regulates cell viability partially through inhibiting SSTR2. Taken together, the above results indicate that YTHDF2 regulates GBM cell proliferation and invasion through LXRα and HIVEP2.
Suppression of LXRα and HIVEP2 by YTHDF2 is essential for YTHDF2's function in GBM tumorigenesis. To ascertain that suppression of LXRα and HIVEP2 by YTHDF2 is responsible for YTHDF2's function in GBM tumorigenesis, we asked whether knocking down LXRα and HIVEP2 could reverse the effects of YTHDF2 inhibition in vivo. The mice injected with GSC11-or GSC7-2-control cells all developed brain tumors resembling human GBM, whereas depletion of YTHDF2 substantially inhibited brain tumor formation (Fig. 7m). Depletion of LXRα and HIVEP2 largely abolished tumor growth inhibition by YTHDF2 knockdown (Fig. 7m) and drove the brain tumors more invasive as compared to those of YTHDF2 knockdown (Fig. 7n). Moreover, depletion of YTHDF2 prolonged the survival of tumor-bearing mice as compared with the control, whereas LXRα and HIVEP2 knockdown reversed the effect of YTHDF2 depletion on the survival of tumor-bearing mice (Fig. 7o). Furthermore, we confirmed that the control tumor xenografts have very low expression of LXRα and HIVEP2 proteins (Supplementary Fig. 7n) but depletion of YTHDF2 increased the expression of LXRα and HIVEP2 ( Supplementary Fig. 7n). Taken together, these results suggest that YTHDF2 and YTHDF2-mediated regulation of LXRα and HIVEP2 play important roles in GBM tumorigenesis.

Discussion
Ythdf2 knockout in zebrafish affects maternal-to-zygotic transition 30 , and Ythdf2 mutation in mice caused female infertility 11 . Ythdf2 knockout mice also displayed early brain developmental failure due to decreased neural stem/progenitor cell proliferation and differentiation 10 . In human, YTHDF2 has been shown to affect hematopoietic stem cell amplification and leukemia survival 31,32 , but our understanding of disease-associated expression and function of YTHDF2 remain limited. In this study, we identified YTHDF2 as a prognostic factor for poor glioma patient survival and found that YTHDF2 overexpression in GBM cells is required for GBM cell growth, invasion, and tumor formation in vivo. YTHDF2 knockdown inhibits GBM proliferation, invasion, and tumorigenicity largely through restraining LXRα and HIVEP2 expression which directly impacts patient survival. Furthermore, our data connected EGFR signaling, a critical driver of GBM development and therapy resistance, with the m 6 A-dependent mRNA clearance. EGFR-YTHDF2 b Western blotting of YTHDF2 in U87/EGFR cells cultured with or without EGF. c Western blotting of YTHDF2 in LN229/EGFRvIII tet-on cells treated with increasing concentration of doxycycline (Dox, 0, 1, 10 μg/mL) for 48 h. d Stability of YTHDF2 was measured by cycloheximide (CHX, 50 μg/mL) chase assay in GSC11 cells with vehicle or Gefitinib treatment (left panel), and analysis of YTHDF2 protein lifetime (right panel). Data are mean ± S.D. of three independent experiments (unpaired two-sided t test). e Stability of YTHDF2 was measured by CHX (50 μg/mL) chase assay in LN229/EGFRvIII tet-on cells cultured with or without Dox (10 μg/mL, left panel), and analysis of YTHDF2 protein lifetime (right panel). Data are mean ± S.D. of three independent experiments (unpaired two-sided t test). f Western blotting of YTHDF2 in GSC11 cells cultured with or without EGF and treated with inhibitor of SRC (SU6656) or ERK1/2 (SCH772984). g Stability of YTHDF2 was measured by CHX chase assay in LN229/EGFRvIII tet-on cells cultured with Dox plus vehicle or SCH772984 treatment (left panel), and analysis of YTHDF2 protein lifetime (right panel). Data are mean ± S.D. of three independent experiments (unpaired two-sided t test). h Co-immunoprecipitation (co-IP) of YTHDF2 with phospho-ERK1/2 (Thr202/Tyr204) in whole-cell extracts from GSC11 cells cultured with or without EGF. i Co-IP of phospho-ERK1/2 (Thr202/Tyr204) with YTHDF2 in whole-cell extracts from GSC11 cells cultured with or without EGF. j Proximity ligation assay (PLA) of phospho-ERK1/2 and YTHDF2 in LN229/EGFRvIII tet-on or GSC11 cells (left panel) and quantification of average PLA puncta per cell (right panel). Scale bar = 20 μm. Data are mean ± S.D., n = 6 field pictures quantified over three independent experiments (unpaired two-sided t test). For a-c, f, h, and i, representative blots of three independent experiments are shown. Source data are provided as a Source data file.  Fig. 8). GBM cells are highly dependent on cholesterol for survival 29 , and therefore activation of LXR subjected GBM cells to death 12,14 . LXRA expression is significantly lower than LXRB in GBM 12 . We found that YTHDF2 was highly expressed in GBM and mediated mRNA clearance of LXRA but did not affect the mRNA of LXRB. In fact, in the TCGA glioma dataset, the levels of YTHDF2 is inversely correlated with the levels of LXRA. Thus, the YTHDF2-mediated LXRA downregulation might explain the difference between expressions of LXRA and LXRB in GBM. HIVEP2 transcriptionally regulates class I MHC, interleukin-2 receptor, somatostatin receptor II, and interferon-beta gene expressions, and regulates MYC, NF-κB, or TGF-β signaling 33 . Exogenous expression of HIVEP2 inhibited cell growth, induced differentiation, and blocked the cell cycle of glioma cells in vitro 26 . We demonstrated that YTHDF2-mediated HIVEP2 downregulation promoted GBM progression (Fig. 7). HIVEP2 is frequently downregulated in human breast cancer 34 , but the mechanism is unclear. Our finding provides a possibility that YTHDF2-mediated HIVEP2 mRNA clearance accounts for inhibition of HIVEP2 expression.
Mutations of YTHDF2 might also affect the stability of YTHDF2 protein. Thus, we queried TCGA data for mutations in YTHDF2. We found that among 507 lower-grade glioma patients and 378 GBM patients, only two patients were identified by YTHDF2 mutation (one V505I mutation and one R527W mutation). These suggest that the amino acids of YTHDF2 are rarely mutated in GBM, which reinforces the importance of our findings on the regulation of YTHDF2 protein by EGFR/SRC/ ERK activation.
In summary, we show that EGFR/SRC/ERK phosphorylates and stabilizes YTHDF2 protein, leading to repression of YTHDF2 target gene expression. These findings reveal the roles of EGFR signaling in tumorigenesis, as well as a mechanism responsible for YTHDF2 overexpression in GBM. We also present compelling evidence that the YTHDF2-mediated, m 6 A-dependent mRNA clearance of LXRA and HIVEP2 is required for cholesterol dysregulation, cell proliferation, invasion, and tumorigenesis of GBM. Our data on GBM inhibition by ablation of YTHDF2 demonstrated the potential of targeting YTHDF2 for therapeutic intervention of the life-threatening brain tumor.

Methods
Cell lines and primary cell cultures. Human glioma Hs683 and SW1783 cell lines and GBM T98G, U87 MG, LN229 cell line were purchased from the American Type Culture Collection (ATCC). U251 MG cell line was purchased from Sigma-Aldrich (the European Collection of Authenticated Cell Cultures). U87/EGFR cells were a gift from Dr. Frank B. Furnari. These cell lines were cultured in Dulbecco's modified Eagle's medium (DMEM) with 10% fetal bovine serum. GSC11, GSC17, GSC20, GSC23, GSC7-2, and GSC6-27 were obtained from fresh surgical specimens of human primary and recurrent GBMs under IRB protocol LAB04-0001 approved by the Institutional Review Board of the University of Texas MD Anderson Cancer Center with the written informed consents from the patients. The GSCs were cultured as tumorspheres in DMEM/F12 medium supplemented with B27 supplement (Life Technologies), bFGF and EGF (20 ng/mL each). Only early-passage GSCs were used for the study. The characteristics of the GSCs were presented previously 5 . Normal human astrocytes were purchased from LONZA (CAT#CC-2565) and cultured in AGM Astrocyte Growth Medium (CAT#CC-3186, LONZA) following the manufacturer's protocol. For Tet-inducible EGFRvIII expression, LN229/EGFRvIII cells (Tet-on) were a gift from Dr. Paul S. Mischel, and were cultured in DMEM supplemented with 10% Tet-free FBS (Clontech). Cells were harvested and analyzed 24-48 h depending on the experimental conditions after the addition of indicated concentration of doxycycline (10 μg/mL as default concentration). For GSC differentiation, GSC11 or GSC7-2 cells were plated on tissue culture dishes in DMEM with 10% FBS for 1 week.
Mice and animal housing. Male and female athymic nude mice at 6-8 weeks age were used in the experiments. Mice were grouped by 5 animals in large plastic cages and were maintained under pathogen-free, ambient temperature 25°C, humidity 48-60%, and a 12-h dark/light cycle conditions according to the NIH Guide for the Care and Use of Laboratory Animals. All mouse experiments were reviewed and approved by Institutional Animal Care and Use Committees of the University of Texas MD Anderson Cancer Center and the Virginia Commonwealth University.
Intracranial tumor assay. All animal studies were performed with athymic nude mice at 6-8 weeks age that were randomly allocated to each group. For the studies of investigating mice survival, mice were intracranially injected with 10,000 GSC11, 10,000 GSC7-2, or 500,000 LN229 cells. When reaching moribund condition or 90 days endpoint, mice were euthanized, and the data were analyzed by Kaplan-Meier plot. For analysis of tumor growth and invasion, mice intracranially injected with 50,000 GSC11, 50,000 GSC7-2, or 500,000 LN229 cells were euthanized 30 days after injection and brains were collected, formaldehyde-fixed and paraffin-embedded. Brain sections were processed with hematoxylin and eosin (H&E) staining to examine tumor growth and invasion. The formula of tumor volume (V) = L × W 2 /2, where L is the length and W is the width of the tumor, was used to calculate the volume of each brain tumor. Relative invasive fingers per tumor were estimated microscopically by counting protruded tumor tissue figures and disseminated areas as previously described 35 .
Analysis of RIP-sequencing data. All samples of RNA-binding protein immunoprecipitation sequencing (RIP-seq) studies were performed by Illumina Hiseq 2500 with single-end 50-bp read length. The deep sequencing data were aligned to human reference genome version 37 (GRCh37) using hisat2 36 with default setting. After reads were mapped with hisat2, Stringtie 37 was used to calculate the read counts of each gene which represent their transcriptional expression level. Then we performed DEseq2 38 to calculate normalized reads counts in each gene and detect enriched transcripts with a double threshold on the log2-fold change (>1) and the correspondent statistical significance (P value <0.05). Analyzed data are available in Supplementary Data 1. Raw data are available under accession number GSE142828 on the NCBI GEO database.
RNA sequencing. Total RNA was isolated from YTHDF2 knockdown (siYTHDF2) or control (siControl) GSC11 cells using Trizol reagent (Thermo Fisher). The RNA-Seq library was generated using TruSeq Stranded mRNA Library Prep Kit (Illumina) and sequenced on a NovaSeq 6000 sequencer. Data were mapped using TopHat (TopHat v2.0.14) onto the human reference genome version 37 (GRCh37), and quantified using Cufflinks (Cufflink v2.2.1) and the GENCODE gene model. Analyzed data are available in Supplementary Data 3. Raw data are available on GEO database under the accession number GSE142828. mRNA stability assay. GSCs were plated in a poly-lysine coated 6-cm dish and incubated with actinomycin D (Santa Cruz) at 5 µM for the indicated time. The first time point (t = 0 h) was taken as after 20 min, then 2, 4, and 6 h. Total RNA extracted from each sample was used for qRT-PCR analysis. See Supplementary  Table 1 for quantitative PCR primers. According to previous studies 39 , the Fig. 5 Identification of YTHDF2 targets by high-throughput RNA-seq and RIP-seq. a Bland-Altman plot of YTHDF2 RIP-seq in GSC11 cells. For each transcript, the average signal (measured as log 2 normalized reads counts) against the RIP-seq log 2 fold enrichment (RIP versus INPUT) was plotted. Significantly enriched targets are highlighted in red. b Volcano plot of RNA-seq results in siControl and siYTHDF2 GSC11 cells. Red dots indicate fold change >1.5 and p value <0.05. c Heatmap of expression and enrichment of 19 overlapped genes in RNA-seq, RIP-seq, and TCGA datasets results. d Schematic workflow of YTHDF2 downstream targets analysis. e, f Plots showing YTHDF2 binding to individual mRNAs of LXRA (e) and HIVEP2 (f) genes in GSC11 cells, as measured by RIP-seq. The normalized reads distribution for input (blue) and YTHDF2 RIP along mRNAs (red). g CLIP-qPCR showing the association of LXRA and HIVEP2 transcripts with YTHDF2 in GSC11 cells. Data are mean ± S.E.M., n = 4 biologically independent experiments (unpaired two-sided t test). h Analysis of mRNA levels of YTHDF2, LXRA, HIVEP2, and LXRB in GSC11 cells stably expressing control or YTHDF2 shRNA, using qPCR. Data are mean ± S.E.M., n = 3 biologically independent experiments (unpaired two-sided t test). i Analysis of mRNA levels of LXRA, HIVEP2, and LXRB in LN229 cells transfected with vector or YTHDF2 expression constructs, using qPCR. Data are mean ± S.E.M., n = 5 biologically independent experiments (unpaired two-sided t test). j Kaplan-Meier overall survival plot showing survival rates for lower-grade glioma and GBM patients having both low LXRA expression (<median) and low HIVEP2 expression (<median) (blue), and both high LXRA expression (>median) and high HIVEP2 expression (>median) (red) in the TCGA dataset (two-sided log-rank test). k Percentage of low LXRA/HIVEP2 expression and high LXRA/HIVEP2 expression in YTHDF2 low expression and high expression group. Chi-squared test, χ 2 = 69.63, p < 0.0001. Source data are provided as a Source data file. equation below was used to calculate the mRNA half-life (t 1/2 ):  Immunofluorescence (IF) staining. For IF analysis of cultured cells, cells were fixed with 4% formaldehyde (Thermo Fisher) for 15 min followed by methanol permeabilization. For IF analysis of specimen, deparaffinized, rehydrated through an ethanol series followed by antigen retrieval with sodium citrate or tris-EDTA buffer according to antibody manufacturer's protocol. And then, blocked with 5% normal goat serum (Genescript) with 0.1% Triton X-100 in TBS for 60 min at room temperature. Immunostaining was performed using the appropriate primary and appropriate Alexa Fluor ® 488 or Alexa Fluor ® 594 secondary antibodies (Invitrogen, dilution 1:1000). Nuclei were counterstained with DAPI. Images were taken with a ZEISS Axio Scope.A1 Upright Microscope, or with a confocal imaging system (Olympus FluoView FV1000).
Western blotting. For evaluating protein expression in multiple cell lines, or analyzing protein expression in the cells under different conditions, protein extracts were obtained by lysing normal human astrocytes, lower-grade glioma, glioblastoma, and GSC cells, or indicated U87, LN229, or GSC cells with Laemmli sample buffer. Proteins extracted from the cells were applied equally to 10% SDS-PAGE for separation. Proteins were then transferred onto a PVDF membrane (GE), and the resulted membrane was washed and incubated with 5% Blotting-Grade Blocker (Bio-Rad) or BSA in TBST buffer for 1 h at room temperature. The membrane was washed twice with TBST and incubated with primary antibody at 4°C overnight with gentle rotation. Then, appropriate secondary antibody conjugated with horseradish peroxidase was applied for 1 h. The chemiluminescence signals were detected with enhanced chemiluminescence (ECL) and quantified by densitometry using the ImageJ software (NIH, Bethesda, USA). At least three independent experiments have been carried out and representative results are shown.
Cell invasion assay. Cell invasion was measured in Matrigel-coated transwell inserts containing polyethylene terephthalate filters with 8-μm pores (Millipore). Specifically, for LN229 cells, 100 μL of transiently or stably transfected 2 × 10 5 LN229 cells in serum-free DMEM or 2 × 10 4 GSC11 cells in growth factor-free DMEM/F12 with 0.05% bovine serum albumin were plated in the upper chamber, whereas 500 μL of DMEM (or DMEM/F12) supplemented with 10% fetal bovine serum was added to the lower well as chemoattractant. For invasion assay using conditioned culture medium, GSC11 cells were first transfected with siCtrl, siYTHDF2, siAPOE or siYTHDF2 plus siAPOE for 24 h. Then, the cells were replenished with growth factor-free DMEM/F12 medium and cultured for additional 24 h before harvesting the culture medium. In total, 2 × 10 4 GSC11 cells were suspended with conditioned medium and plated in the upper chamber, whereas 500 μL of the conditioned medium supplemented with 10% fetal bovine serum was added to the lower well as attractant medium. The indicated cells were allowed to invade for 22 h. Non-invading cells were carefully removed with wet cotton swabs from the top of the membranes. The invading cells of the lower surface were fixed with methanol for 10 min, washed with PBS, and stained with 0.1% crystal violet for 10 min. The invading cells were either counted in at least five random fields per insert from three independent experiments, or quantified dissolved crystal violet with 33% acetic acid under 570 nm from three independent experiments. Images were taken using a Leica microscope connected to a digital camera (Leica DFC500).
BrdU incorporation assay. For the BrdU (5-bromo-2′-deoxyuridine) incorporation assay, cells were cultured with a BrdU-labeling reagent (Life Technologies), performed DNA hydrolysis with 1 M HCl, and stained with an anti-BrdU primary antibody (Cell Signaling) and Alexa Fluor ® 594 secondary antibody according to the manufacturer's instructions.
LDL uptake analysis. To analyze LDL uptake in the indicated LN229 or GSC11 cells, 2 mL of cell suspension containing 6 × 10 4 cells was plated in 6-well plate with cover slide. For GSC11 cells, the cells were attached to 0.01% poly-Llysine coated cover slide. After attachment, cells were incubated with 5 μg/mL  LXRα and HIVEP2 are functionally essential targets of YTHDF2 in cell proliferation, invasion, cholesterol dysregulation, and tumorigenesis of GBM cells. a Western blotting of YTHDF2, LXRα, and HIVEP2 in GSC11 cells stably expressing shCtrl, shYTHDF2, or shYTHDF2 plus shLXRA or/and shHIVEP2. Representative blot of three independent experiments is shown. b Proliferation of the cells in (a) was measured. Data are mean ± S.E.M., n = 6 biologically independent experiments. c In vitro invasion assay for GSC11 cells expressing shCtrl, shYTHDF2, or shYTHDF2 plus shLXRA or/and shHIVEP2. Data are mean ± S.E.M., n = 4 wells examined over three independent experiments. d, e mRNA levels of LXRA and its downstream targets in shCtrl and shYTHDF2 GSC11 cells (d) or in LN229 cells expressing vector or YTHDF2 plasmid (e). Data are mean ± S.E.M., n = 3 biologically independent experiments (unpaired two-sided t test). f, g Relative cellular cholesterol of GSC11 cells stably expressing shCtrl, shYTHDF2, shLXRA, or shYTHDF2 plus shLXRA (f) or LN229 cells expressing vector or YTHDF2 plasmid (g). Data are mean ± S.E.M., n = 6 (f) or n = 3 (g) biologically independent experiments. h, i Quantification of LDL uptake in GSC11 cells expressing shCtrl, shYTHDF2, or shYTHDF2 plus shLXRA (h) or in LN229 cells expressing wild-type (WT) or m 6 A recognition defective (MUT) YTHDF2 (i). Data are mean ± S.E.M., n = 3 biologically independent experiments. j In vitro invasion assay for GSC11 cells treated with conditioned culture medium from GSC11 cells transfected with siCtrl, siYTHDF2, siAPOE, or siYTHDF2 plus siAPOE. Data are mean ± S.E.M., n = 3 biologically independent experiments. k, l mRNA levels of HIVEP2 and its downstream target SSTR2 in shCtrl and shYTHDF2 GSC11 cells (k) or in LN229 cells expressing vector or YTHDF2 plasmid (l). Data are mean ± S.E.M., n = 3 (k) or n = 4 (l) biologically independent experiments (unpaired twosided t test). m Nude mice intracranial tumor assay using shCtrl, shYTHDF2, and shYTHDF2 plus shLXRA and shHIVEP2 GSC11 cells. Brain sections stained with H&E show representative tumor xenografts. Tumor volumes were calculated. Data are mean ± S.D. n In vivo invasion assay for shCtrl, shYTHDF2, and shYTHDF2 plus shLXRA and shHIVEP2 GSC11 cells. Representative H&E staining showing edges of the mice brain tumors (left), and the relative invasive fingers per tumor were counted (right). Scale bar = 200 μm. Data are mean ± S.D. o Overall survival of mice injected with shCtrl, shYTHDF2, and shYTHDF2 plus shLXRA and shHIVEP2 GSC11 cells (two-sided log-rank test). For b, c, f, g, h, i, m, and n, data were analyzed by one-way ANOVA Tukey's post hoc test. For m-o, n = 10 mice per group examined over two independent experiments. Source data are provided as a Source data file.
−80°C overnight. Precipitated RNA was reverse-transcribed and quantified by qPCR for m 6 A enrichment.
Immunohistochemical (IHC) staining. Tissue microarray (BS17017b and GL808) was purchased from US Biomax (Derwood, MD, USA). Anonymous archived human glioblastoma tissue slides were obtained from The University of Texas MD Anderson Cancer Center under a protocol approved by the institutional review board with informed consent. For immunohistochemical (IHC) staining, GBM xenografts, human surgical specimens tissue slides, or TMAs were deparaffinized, rehydrated through an ethanol series followed by antigen retrieval with sodium citrate or tris-EDTA buffer according to antibody manufacturer's instruction. Sections were blocked with 5% normal goat serum (Genescript) with 3% BSA in TBS for 60 min at room temperature and were incubated with 3% H 2 O 2 in PBS for 15 min at room temperature to block endogenous peroxidase and then incubated with appropriate primary antibodies at 4°C overnight. IHC staining was performed with horseradish peroxidase (HRP) conjugates using DAB detection. The immunoreactive score ( The immunoreactive score IRS ð Þ¼ A B ð3Þ Score 0-3 = low (1st quartile), 4-8 = medium (2nd and 3rd quartiles), 9-12 = high (4th quartile).
RNA isolation and quantitative real-time PCR. For RNA isolation, TRIzol reagent (Life Technologies) was added to the indicated GSC or LN229 cells, and total RNA was extracted from the cells. mRNA was purified from the total RNA by poly-A selection with GenElute™ mRNA Miniprep Kit (Sigma). cDNAs were synthesized from the purified mRNA by using iScript TM Reverse Transcription Supermix (Bio-Rad). For quantitative real-time PCR, the PCR reactions were set with PowerUp TM SYBR ® Green Master Mix (Life Technologies) on a 7500 Fast Real-time PCR System (Applied Biosystems) following the manufacturer's instruction. Primers used in this study were summarized in Supplementary Table 1.
Luciferase reporter assay. The 3′UTR of LXRA (ENST00000467728.5, https://uswest.ensembl.org/Homo_sapiens/Transcript/Summary? g=ENSG00000025434;r=11:47257979-47268845;t=ENST00000467728) or HIVEP2 (ENST00000012134.7, https://uswest.ensembl.org/Homo_sapiens/ Transcript/Summary?g=ENSG00000010818;r=6:142751469-142945031; t=ENST00000012134) was amplified by PCR from the cDNA of GSC11 cells using specific primers and was inserted into the NheI and XhoI sites in the downstream of the luciferase gene in the pmirGLO vector. The primers used are listed in Supplementary Table 1. For luciferase reporter assay, LN229/EGFRvIII tet-on cells (2 × 10 4 cells per well in 24-well plates) were transfected with control or YTHDF2 siRNAs, and with pmirGLO luciferase reporter plasmids containing 3′UTR of LXRA or HIVEP2 using X-tremeGENE HP DNA Transfection Reagent (Roche). Luciferase activities were measured with the Dual-Luciferase Reporter Assay System (Promega) 24 h after transfection. Data were first normalized firefly luciferase activity to Renilla luciferase activity for each set of triplicate samples, and then compared to vectoronly group. All experiments were performed at least three times independently.
CLIP-qPCR. Cross-linking and RNA immunoprecipitation (CLIP) RNA immunoprecipitation was performed as follows, following ultraviolet cross-linking, GSC11 or LN229 cells were harvested and nuclear extracts were isolated and sonicated. One microgram of YTHDF2 (Proteintech) antibody or normal rabbit IgG (Millipore) was conjugated to Protein A/G Magnetic Beads (Thermo Fisher) by incubation at 4°C for 4 h, followed by 3× wash and incubation with pre-cleared nuclear extraction in RIP buffer (150 mM KCl, 25 mM Tris pH 7.4, 5 mM EDTA, 0.5 mM DTT, 0.5% IGEPAL ® CA-630, 1× protease inhibitor) at 4°C overnight. After washing with RIP buffer for three times, beads were resuspended in 80 µL PBS, followed by DNA digestion at 37°C for 15 min and incubation with 50 µg of Proteinase K (Sigma-Aldrich) at 37°C for 15 min. Input and co-immunoprecipitated RNAs were recovered by TRIzol (Invitrogen) extraction and used for qPCR analyses using primers listed in Supplementary Table 1.