This document is downloaded from DR‑NTU (https://dr.ntu.edu.sg) Nanyang Technological University, Singapore.

Genome‑wide RNAi screen identify melanoma‑associated antigen Mageb3 involved in X chromosome inactivation

Li, Wei; Hong, Ru; Lai, Lan‑Tian; Dong, Qiman; Ni, Peiling; Chelliah, Rosi; Huq, Mehnaz; Siti Nadirah Ismail; Chandola, Udita; Ang, Zhiwei; Lin, Bingqing; Chen, Xin; Chen, Lingyi; Zhang, Li‑Feng

2018

Li, W., Hong, R., Lai, L.‑T., Dong, Q., Ni, P., Chelliah, R., . . . Zhang, L.‑F. (2018). Genome‑wide RNAi screen identify melanoma‑associated antigen Mageb3 involved in X chromosome inactivation. Journal of Molecular Biology, 430(17), 2734‑2746. doi:10.1016/j.jmb.2018.05.031 https://hdl.handle.net/10356/139663 https://doi.org/10.1016/j.jmb.2018.05.031

© 2018 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY‑NC‑ND license (http://creativecommons.org/licenses/by‑nc‑nd/4.0/).

Downloaded on 01 Oct 2021 00:29:58 SGT Article

Genome-Wide RNAi Screen Identify Melanoma-Associated Antigen Mageb3 Involved in X Chromosome Inactivation

Wei Li 1,†, Ru Hong 1,†, Lan-Tian Lai 1, Qiman Dong 3, Peiling Ni 3, Rosi Chelliah 1, Mehnaz Huq 1, Siti Nadirah Binte Ismail 1, Udita Chandola 1, Zhiwei Ang 1, Bingqing Lin 1, Xin Chen 2, Lingyi Chen 3 and Li-Feng Zhang 1

1 - School of Biological Sciences, Nanyang Technological University, 60 Nanyang Drive, Singapore 637551 2 - State Key Laboratory of Medicinal Chemical Biology, Key Laboratory of Bioactive Materials, Ministry of Education, Tianjin Key Laboratory of Protein Sciences and College of Life Sciences, Nankai University, Tianjin 300071, China 3 - Division of Mathematical Sciences, School of Physical and Mathematical Sciences, Nanyang Technological University, 21 Nanyang Link, Singapore 637371

Correspondence to Lingyi Chen and Li-Feng Zhang: [email protected]; [email protected] https://doi.org/10.1016/j.jmb.2018.05.031 Edited by M Yaniv

Abstract

Xist (inactivated X chromosome specific transcript) is a prototype long noncoding RNA in charge of epigenetic silencing of one X chromosome in each female cell in mammals. In a genetic screen, we identify Mageb3 and its homologs Mageb1 and Mageb2 as genes functionally required for Xist-mediated gene silencing. Mageb1–3 are previously uncharacterized genes belonging to the MAGE (melanoma-associated antigen) gene family. Mageb1–3 are expressed in undifferentiated ES cells and early stages of in vitro differentiation, a critical time window of X chromosome inactivation. Mageb3 showed both cytoplasmic and nuclear localization without enrichment on the inactive X (Xi). Mageb3 interacted with Polycomb group ring finger 3 (Pcgf3), a RING finger protein involved in recruiting Polycomb activities onto Xi. Mageb3 overexpression stabilized Pcgf3 protein. Mageb1–3 gene knockout affected H3K27me3 enrichment and the spreading of gene silencing along Xi. These data suggested that Mageb3 might regulate the recruitment of the Polycomb complex onto Xi and subsequent H3K27me3 modification through Pcgf3. Moreover, the nucleolar enrichment of Mageb3 was diminished when nuclear matrix factor hnRNP U is overexpressed, implying the interaction between Mageb3 and nuclear matrix, which is another possible mechanism for Mageb3 to regulate X chromosome inactivation. © 2018 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

Introduction However, acting as a recruiter, Xist must work together with other factors to achieve XCI. Currently, a few Xist (inactivated X chromosome specific transcript) epigenetic modifications along the inactive X chromo- RNA is a well-known long noncoding RNA (lncRNA) some (Xi) are known to be associated with Xist, involved in X chromosome inactivation (XCI), a dosage including histone modification such as H3K27me3 compensation mechanism of mammals, whereby one and histone variant such as macroH2A. However, X chromosome in each female cell is epigenetically genetic studies show that none of them are essential for silenced to balance the X-linked gene dosage between XCI [2].WhenXist is conditionally deleted from somatic males and females [1]. At the onset of XCI (~E5.5 cells in which XCI has been fully established, all the mouse embryos), the X-linked Xist gene becomes known Xist-associated epigenetic modifications are allele specifically expressed from one X and the RNA removed together with Xist, but XCI remains largely transcripts coat chromosome territory in cis recruiting intact [2]. These results indicate the existence of other epigenetic factors to establish the chromosome- unknown epigenetic factors or mechanisms, which wide gene silencing [1]. Xist expression triggers XCI. are recruited or established by Xist RNA during early

0022-2836/© 2018 The Author. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). J Mol Biol (2018) 430, 2734–2746 Mageb3 is involved in X chromosome inactivation 2735 embryonic development and functionally essential for common backbone sequence of the lentiviral vector to XCI. Intensive effort of isolating Xist-binding proteins amplify the shRNA sequences. Each PCR product was identified Spen [3–5], which brings histone deacetylase sequenced independently. Thirteen candidate genes activity onto the Xi. Experimental evidence shows Spen werecommonlyfoundinall6PCRsamples(Fig. 1f) is required in XCI during in vitro differentiation of including tumor suppressor P53, which is consistent embryonic stem cells (ES cells) [4–7]. However, Spen with the design of the genetic screen. One hundred is not specifically expressed in early embryonic tissues thirty-four shRNAs can be commonly found in all three [8]. The essential role of Spen in XCI is not fully rounds of genetic screen (Supplementary Table 1). supported by the mutant phenotype of Spen knockout Unfortunately, we did not identify any known XCI factor mouse embryos [8]. in the screening results, which may relate with the quality of individual shRNAs in the library or the sensitivity of cells to the dosage change of an individual Results gene (shRNAs are not able to completely knockout a gene). From these data and three rounds of pilot screen To search for novel genes involved in XCI, we carried (data not shown), we selected, validated using the cell out a genetic screen using a genome-wide short hairpin survival assay and ruled out 52 individual shRNAs RNA (shRNA) library. A male ES cell line with an (Supplementary Table 2). Furthermore, we validated inducible Xist transgene (iXist) was established for the using the cell survival assay and ruled out 6 candidate genetic screen. iXist was established from Ainv15 cells, genes using the CRISPR/Cas9 approach (Supplemen- a male mouse ES cell line carrying an engineered tary Table 3, Supplementary Fig. 2). From the 13 cassette upstream of the X-linked hypoxanthine candidate genes, which were commonly found in all 6 guanine phosphoribosyl transferase (Hprt) gene PCR samples (Fig. 1f), Mageb3 was successfully (Fig. 1a). Through Cre-mediated gene targeting, a validated as a gene functionally important for XCI full-length Xist transgene including its polyA signal was (Figs. 1fand2). inserted downstream of the tetracycline response To validate Mageb3, we knocked out the gene in iXist element of Ainv15 cells, restoring the neomycin (iXist-KO1) using CRISPR/Cas9 with a nucleotide oligo resistance (Fig. 1a). Moreover, a red fluorescent protein as the repair template. The mutated Mageb3 allele (tdTomato) was included as a reporter gene (Fig. 1a). contains a premature stop codon 3 bp downstream of Both neomycin resistance and tdTmomato could be the gene's start codon (Fig. 2a). A HindIII restriction site used as reporters to indicate status of XCI. The cell was created for the genotyping purpose (Fig. 2a). line's genotype was confirmed by polymerase chain Moreover, a reading-frame shift was left in the rest of reaction (PCR; Supplementary Fig. 1a) and DNA thecodingregion(Fig. 2a). The Mageb3 (−/−) fluorescence in situ hybridization (FISH) (Fig. 1b). Its genotype of iXist-KO1 cells was confirmed by HindIII karyotype was confirmed as 40XY (Supplementary digestion and DNA sequencing (Supplementary Fig. 1c). Induced Xist RNA transcripts coated (Fig. 1c) Fig. 3). Mageb3 is highly homologous with two other thesingleXchromosomeinthe male cells and silenced genes (Mageb1 and Mageb2) in the gene family its genes (Supplementary Fig. 1b) causing cell growth (Supplementary Fig. 5). In the mouse genome, retardation and cell death (Fig. 1d). Although induced Mageb1 and Mageb2 are X-linked genes encoding Xist expression resulted in XCI in both undifferentiated identical proteins [9] (Supplementary Fig. 5) and and differentiating cells, the killing effect of induced XCI Mageb3 is located on chromosome 2 (Fig. 2b). The is significantly stronger in differentiating cells in the sequence of the shRNA clone (TRCN0000112714) presence of geneticin (G418) (Supplementary Fig. 1b). cannot distinguish the three genes, although the target We established a protocol (Fig. 1e) whereby more than oftheshRNAcloneislistedasMageb3 in the shRNA 99% of the iXist cells were killed by induced XCI in the library design. In this study, for genotyping PCR and presence of G418 (Fig. 1d) and applied this protocol to allele-specific reverse transcription PCR (RT-PCR), three independent rounds of genetic screen using a allele-specific primers could only be designed to pooled lentiviral shRNA library. The genomic DNA distinguish Mageb3 from Mageb1 and Mageb2. We isolated from the surviving cells of each round of screen coin the term Mageb1–3 referring to all three genes and was subjected to PCR reactions in duplicate using the the term “Mageb1&2” to discuss the experimental

Fig. 1. The genetic screen. (a) iXist, the transgenic male mouse ES cell line engineered for the genetic screen. (b) DNA FISH on a metaphase chromosome spread of an iXist cell. DNA was stained by DAPI (blue). Chromosome X was detected by a Cy3-labeled chromosome paint (red). The two copies of the Xist gene were detected by a FITC-labeled nucleotide probe (green). The signals correspond precisely to the positions of the transgenic (white arrow) and endogenous (yellow arrow) copies of Xist. (c) Xist RNA FISH on iXist cells. Xist RNA was detected by a Cy3-labeled nucleotide probe (red). The cells were cultured as undifferentiated ES cells without G418. (d) Induced XCI along the single X chromosome in iXist cells (male) causes massive cell death in the differentiating cells and in the presence of G418. We named this experiment as the “cell survival assay” and used it for functional validation of induced XCI. (e) The experimental design of the genetic screen on iXist cells using a genome-wide shRNA library. (f) A shortlist of candidate genes identified in the screen. 2736 Mageb3 is involved in X chromosome inactivation

(a) Hprt tetOP ∆NEO Ainv15 (ChrX)

Hprt tetOP pPGK-ATG Xist NEO iXist (ChrX) loxP tdTomato Xist polyA signal

(b) DAPI Xist Gene Chr X Merge

iXist

(c) iXist Cells (undifferentiated) DAPI Xist RNA Merge

(e) iXist cells Control Lentiviral shRNA Library Transduction

-Lif +G418 +Dox dox 48hrs for 8 days Genomic DNA Extraction from Surviving Cells

(d) Cell Survival Assay Day 0 Day 8 PCR -Lif, +G418, -Dox 500,000 Cells 5,800,000 Cells -Lif, +G418, +Dox 500,000 Cells 7,000 Cells Solexa Sequencing (f) shRNA ID Read Count Gene Name Gene ID Gene Description TRCN0000110606 6773 Myo10 17909 myosin X TRCN0000095612 5395 Lbh 77889 limb-bud and heart TRCN0000101245 4793 Nktr 18087 natural killer tumor recognition sequence TRCN0000012359 3703 p53 22059 transformation related protein 53 TRCN0000112714 3572 Mageb3 17147 melanoma antigen, family B, 3 TRCN0000174709 3372 Fastkd2 75619 FAST kinase domains 2 TRCN0000032413 3212 Prss51 1E+08 protease, serine 51 TRCN0000097365 2982 Amt 434437 aminomethyltransferase TRCN0000054378 2541 Gnb2l1 14694 guanine nucleotide binding protein, beta polypeptide 2 like 1 TRCN0000100580 2257 Sec24c 218811 SEC24 related gene family, member C (S. cerevisiae) TRCN0000033070 2142 Ela2a 13706 elastase 2A TRCN0000042159 866 Akr1c18 105349 aldo-keto reductase family 1, member C18 TRCN0000034470 398 Ccl2 20296 chemokine (C-C motif) ligand 2

Fig. 1 (legend on previous page) results in which the allele-specific primers cannot repair template is allele-specific to Mageb3, the guide distinguish Mageb1 from Mageb2. For CRISPR- RNA (gRNA) cannot distinguish the three genes. Cas9-mediated genome editing work, although the Therefore, Mageb1&2 can be disrupted by non- Mageb3 is involved in X chromosome inactivation 2737

0.5 kb (a) (b) ATG TAG Mageb1&2 Mageb3 Mageb3 (Chr X) (Chr 2) iXist-KO1 -,- -/- +,- ATG-CCT-AGG-GGT-CAA-AAG-AGT-AAG-AGC iXist-KO2 -/- Wild Type Met Pro Arg Gly Gln Lys Ser Lys Ser

iXist-KO1 ATG-CCT-TAA-GCT-TTC-CCG-TGC-TAA-ACG

(c) (d) iXist iXist-KO1 iXist-KO1-RiXist-KO2 iXist-KO2-R 8% * 8% *

6% 6% * *

4% 4% no dox

2% 2% Cell Surviving Rate 0 0 Differentiation Day 8 iXist iXist In vitro with dox iXist-KO1 iXist-KO2 iXist-KO2-R iXist-KO1-R

(e) iXistiXist-KO1 iXist-KO2

no dox 78.5% 73.3% 67.3% Differentiation Day 8

9.63% 45.2% 47.7% with dox In vitro Far Red

0 TomatoRed

(f) 4 Nanog 5 Gata4 In vitro Differentiation day 0 3 2 day 4 day 10

0 0 I II III IV V I II III IV V

10 Fgf5 9 Mageb 1- 3 I. Male mES Wild Type II. iXist III. iXist-KO1 5 5 IV. Female mES Wild Type V. Primary MEF (Male Left, Female Right)

0 0 I II III IV V I II III IV V

Fig. 2 (legend on next page) homologous end joining in this experiment. The Supplementary Fig. 6). To rule out the off-target effect genotyping results confirmed the genotype of iXist- of the CRISPR/Cas9 system, we established iXist-KO2 KO1 as (Mageb1&2 −,−; Mageb3−/−)(Fig. 2b, using a different gRNA. The genotype of iXist-KO2 was 2738 Mageb3 is involved in X chromosome inactivation confirmed as (Mageb1&2 +,−; Mageb3−/−)(Fig. 2b, the last exon. Therefore, the mutant RNA transcripts of Supplementary Figs. 4 and 6). Consistent with the Mageb1–3 are not sensitive to non-sense mediated genetic screen results, iXist-KO1 and iXist-KO2 cells decay [11]. The mutated RNA transcripts were showed significantly higher survival rates than iXist detected in iXist-KO1 at a level similar to the level in cells upon induced XCI (Fig. 2cand d). When Mageb3 wild-type cells (Fig. 2f). was introduced back into the two knockout cell lines as To validate and characterize the functional roles of a stable transgene (iXist-KO1-R and iXist-KO2-R), it Mageb1–3 in the endogenous XCI, we knocked out the successfully restored the killing effect of induced XCI genes in 3F1 cells [12]. 3F1 is a female mouse ES cell on the cells (Fig. 2c). Fluorescence-activated cell line carrying X chromosomes from two different genetic sorting (FACS) analysis on the expression of the backgrounds, the 129 mouse strain (X129) and the Mus TomatoRed reporter gene confirmed the disrupted XCI musculus castaneus (CAST/Ei) mouse strain (XCast). status of mutant cell lines (Fig. 2e). These results show The “preemptive choice” mutant phenotype of 3F1 cells that Mageb3 is involved in Xist-mediated gene causes non-random inactivation of the X129 allele [12]. silencing in iXist cells. Due to the extremely high Therefore, XCI status of X-linked genes can be sequence homology among the three genes, accessed by RNA allelotyping of X129 and XCast. Mageb1&2 are also likely involved in the same After a few rounds of genome editing with the gRNA process. The functional roles of the three proteins are and the repair template used to create iXist-KO1, redundant, as over-expression of Mageb3 alone in the we only obtained a heterozygous mutant clone mutant cell lines was sufficient to restore the killing (Supplementary Fig. 8). From the heterozygous effect of induced XCI. It also should be noticed that, clone, we obtained homozygous mutants using the although the rescue effect of gene knockout is same gRNA and a different repair-template plasmid significant and obvious (the cell surviving rates of carrying a hygromycin selection cassette. Two mutant mutants are more than 10 times of the control's clones (F-M3-KO1 and F-M3-KO2) were established surviving rate), knocking out Mageb1–3 (iXist-KO1) (Supplementary Fig. 8). The genotype of F-M3-KO1 was did not fully protect the cells (cell surviving rate 4%–6%) confirmed as (XX, Mageb1&2 −,−/−,−; Mageb3−/−) from the induced XCI (Fig. 2c and d). The cell surviving (Supplementary Fig. 6). Interestingly F-M3-KO2 is an rate of mutant cells suggests that Mageb1–3 are likely XO cell line (Supplementary Fig. 9). Its genotype was involved in one functional layer of Xist-mediated gene confirmed as (XO, Mageb1&2 −,−; Mageb3−/−) (Sup- silencing, a complex process orchestrated by many plementary Fig. 6). proteins. In corroboration with this notion, shRNA XCI can be recapitulated in female mouse ES cells by knockdown of polycomb group ring finger 3 (Pcgf3), a in vitro differentiation [1]. Defects of dosage compensa- known XCI factor [10], resulted in a comparable cell tion generate difficulties for cells during in vitro differen- surviving rate (Supplementary Fig. 7). tiation. Indeed, the cell count of F-M3-KO1 on day 6 of in Gene expression profiling shows that Mageb1–3 are vitro differentiation was only ~50% of the cell count of the expressed in undifferentiated ES cells and early stages control 3F1 cells (Fig. 3a). Interestingly, this mutant of in vitro differentiation but not in fibroblasts (Fig. 2f). phenotype is linked with dosage compensation, as F- The expression profile of Mageb1–3 is consistent with M3-KO2 (XO, Mageb1&2 −,−; Mageb3−/−) did not their functional roles in Xist-mediated gene silencing. encounter the same difficulty during in vitro differentiation iXist-KO1 cells showed normal colony morphology and (Fig. 3a). It is noticeable that the XO cell line showed a normal expression profile of pluripotency-related genes growth rate faster than the “wild type” 3F1 cells, which (Fig. 2f). iXist-KO1 cells also showed normal gene indicates that XX dosage compensation is not a expression profile during in vitro differentiation (Fig. 2f). troublesome issue for the XO cell line to deal with during Therefore, knocking out Mageb1–3 did not disrupt in vitro differentiation (Fig. 3a). pluripotency and in vitro differentiation of the cells. We carried out padlock single-nucleotide polymor- Mageb3 is a single-exon gene. Mageb1&2 are two phism (SNP) capture to comprehensively profile the genes in which the entire coding region is included in XCI status of X-linked genes in the mutant cells.

Fig. 2. Functional validation of Mageb3. (a) Knocking out Mageb3 in iXist using the CRISPR/Cas9 system with a repair template. The start codon of Mageb3 is shown in red. The selected PAM site is shown in green. The stop codon introduced 3 nt downstream of the start codon on the mutant allele is shown in blue. The HindIII restriction site created by the repair template for genotyping is underscored. (b) The genotypes of Mageb1, Mageb2 and Mageb3 in the two knockout cell lines (iXist-KO1 and iXist-KO2). (c) Functional validation of Mageb3 by deletion of the gene in iXist. Two knockout cell lines were established using two different PAM sites (iXist-KO1 and iXist-KO2). For further validation, a transgenic copy of Mageb3 was introduced back into the knockout cell lines to “rescue” the phenotype (iXist-KO1-R and iXist-KO2-R). Cells were subjected to the cell survival assay. The cell surviving rates are shown. (d) Images of EBs of in vitro differentiation day 8. Cells were cultured in the presence of G418. Scale bar represents 200 μm. (e) FACS analysis on the expression of the TomatoRed reporter gene. (f) Quantitative RT-PCR to show the gene expression profile of Mageb1–3 and multiple pluripotency-related genes in different types of cells. Data are shown in relative fold expression. Normalization was performed using Actb. Error bars indicate SEM (n = 3). Mageb3 is involved in X chromosome inactivation 2739

Padlock capture is a high-resolution RNA allelotyping day 0 (undifferentiated) and day 6 differentiating cell method [13,14].SincetheX129 allele is non-randomly populations of both 3F1 and F-M3-KO1. Padlock SNP selected as the Xi in 3F1 cells (X129 XCast) [12],RNA capture was carried out on each biological sample in allelotyping of X-linked genes reveals the XCI status duplicate (Fig. 3b). Hierarchical clustering the RNA chromosome-wide. The padlock probe library was allelotyping results yielded two “clades” (Fig. 3b) corre- designed to target 2969 SNPs covering 1110 (~55%) of sponding to day 0 and day 6 samples. Within each theX-linkedgenes[14]. RNA samples were isolated from “clades,” the two samples from the same cell line were

129 Cast (a) (b) Allelotype of X-linked Genes: (Reads + 10) / (Reads + 10) 1 million (1M) cells day 0 3F1-day 0-Exp1 3F1-day 0-Exp2 in vitro differentiation 6 days F-M3-KO1-day 0-Exp1 -4.00 -2.67 F-M3-KO1-day 0-Exp2 -1.33 * 0.00 4M * 3F1-day 6-Exp1 1.33 2.67 3F1-day 6-Exp2 4.00 3M F-M3-KO1-day 6-Exp1 F-M3-KO1-day 6-Exp2 2M XCI status of X-linked Genes: Allelotype (Day6) / Allelotype (Day0) * Day 6 Cell Count 1M -2.00 3F1 -1.33 -0.67 0.00 F-M3-KO1 0.67 3F1 1.33 2.00 F-M3-KO1 F-M3-KO2

(c) Centromere Xist

Genes with XCI defeciency in F-M3-KO1

Ch X X-linked Genes Genes silenced along the “Xi” in 3F1 day 0 cells, but not silenced in F-M3-KO1 day 0 cells

(d) Gene Position along Chr X 5.0E+07 1.0E+08 1.5E+08 Allelotype of an X-linked gene: 4 (Reads 129 +10) / (Reads cast +10) 3

2 Comparison of XCI status of a gene: (Allelotype of F-M3-KO1) / (Allelotype of 3F1) 1

0 log10 ((Allelotype of X-linked genes of 3F1 day0 cells) / 0.2)

-3 log10(Comparison of XCI status of day0 cells) -2

-1

(e) Dox (8 hours) Dox (24 hours) Dox (48 hours) 5 16 18 iXist 16 Dox iXist-KO1 14 14 iXist-KO2 4 12 12 No Dox iXist 10 *** 2 * * *** ** *** ** ** ** ** ** * *** * * *** ** 2 ** 4 ** ** 1 *** 2 *** * *

(Normalized to beta-actin) 0 0 0 Relative Expression Level vs iXist Xist Hprt Neo Xist Hprt Neo Xist Hprt Neo Gpc4 Gpc4 Gpc4 Mecp2 Mecp2 Mecp2

TdtomatoRed TdtomatoRed TdtomatoRed

Fig. 3 (legend on next page) 2740 Mageb3 is involved in X chromosome inactivation clustered together. These results confirm the consistency and its working mechanism may relate to chromosome of the RNA allelotyping experiments (Fig. 3b). The results territory or chromatin architecture. also clearly reveal the XCI effect on X129 allele. Although Xist RNA FISH did not reveal any difference of Xist the 3F1 samples and the F-M3-KO1 samples were cloud formation between 3F1 and F-M3-KO1 (Fig. 4a distinguished by hierarchical clustering, the mutant cell and data not shown). We further studied the epigenetic line and the control cell line showed similar levels of modifications of Xi in the mutant cells. H3K27me3 is a chromosome-wide gene silencing. We further compared histone modification enriched on the chromosome the XCI status of the two cell lines (Fig. 3b). The result territory of Xi [16]. The methylase in charge of this shows the difference between the two cell lines lies in the modification is Ezh2 [17], a Polycomb group protein. In gene-silencing pattern along the X chromosomes. Some the day 6 differentiating cell population, only 52% (n = genes showed better XCI status in F-M3-KO1, and some 111) of the Xist clouds in F-M3-KO1 co-localized with other genes showed better XCI status in 3F1 cells (Fig. the enrichment of H3K27me3 compared to 73% (n = 3b). Interestingly, the genes, which showed XCI deficien- 104) of the Xist clouds in 3F1 (Fig. 4a). Similar cy in the mutant cells, are exclusively located along the observation was made on macroH2A (mH2A), a two thirds of mouse X chromosome downstream of the histone variant known to be enriched on Xi [18].In Xist gene (Fig. 3c). 3F1 cells carry a deletion of Tsix gene the day 6 differentiating cell population, 42% (n = 110) (antisense transcript of Xist) along its X129 allele [12].Itis of the Xist clouds in F-M3-KO1 co-localized with the known that Tsix deletion not only results in preemptive enrichment of mH2A compared to 50% (n = 104) of the choice but also causes a slight Xist up-regulation along Xist clouds in 3F1 (Fig. 4a). the X129 allele in undifferentiated cells [12,15]. Interest- To further investigate the roles of Mageb3 in XCI, we ingly, our chromosome-wide gene allelotyping analysis performed co-immunoprecipitation (co-IP) of Mageb3 reveals a special group of genes, which are already and Pcgf3 (Fig. 4b). MAGE proteins are known to silenced along X129 in undifferentiated 3F1 cells (Fig. interact with RING finger proteins, especially the E3 3c and d). We name this phenomenon as “preemptive RING ubiquitin ligases [19]. Pcgf3 is a RING finger XCI.” Intriguingly, the preemptively silenced genes are protein involved in XCI as a critical recruiter of all clustered in a ~30-Mb chromosomal region Polycomb group proteins [10]. The co-IP result upstream of the Xist gene (Fig. 3c and d). Furthermore, indicates the interaction between Mageb3 and Pcgf3 the preemptive XCI status of undifferentiated 3F1 is (Fig. 4b). Moreover, overexpression of exogenous exclusively disrupted in undifferentiated F-M3-KO1 Mageb3 increased the expression level of Pcgf3 cells, although the XCI status of these genes is largely protein (Fig. 4c). These results suggest that Mageb3 normal in the mutant cells during in vitro differentiation binds with Pcgf3 and protects the protein from (Fig. 3c and d). To further study the role of Mageb3 degradation, which is consistent with the attenuated during the early onset of XCI, we induced Xist H3K27me3 enrichment observed in the female mutant expression in undifferentiated ES cells and performed cell lines of Mageb3 knockout. quantitative RT-PCR on individual X-linked genes (Fig. To study the subcellular localization of Mageb3, we 3e). The deficiency of Xist-mediated gene silencing established Mageb3–GFP fusion plasmid construct. A was more clearly observed in mutant cells treated with human cell line, 293T cells, was chosen for the Dox for 24 h, but less so in the mutant cells treated with experiment because of the high transfection efficiency Dox for 8 or 48 h (Fig. 3e). These results suggest that of the cells. Mageb3–GFP showed both nuclear and Mageb3 may play facilitative roles in Xist-mediated cytoplasmic localization (Fig. 4d). Moreover, Mageb3– gene silencing especially during the early onset of XCI, GFP was also clearly enriched in nucleoli (Fig. 4d),

Fig. 3. Functional characterization of Mageb1–3 during the endogenous XCI in female cells. (a) Wild-type and mutant ES cells were subjected to in vitro differentiation and the cell count of the day 6 differentiating cell population is shown. Data are shown as mean ± S.D. of experimental triplicate. The statistical analysis used is the Student's t-test. One asterisk indicates p values smaller than 0.01. (b) Chromosome-wide RNA allelotyping results are plotted in hierarchically clustered heatmaps. The average was taken from the duplicates of each biological sample. The chromosome-wide XCI status of F-M3-KO1 and 3F1 cells is shown in a heatmap. The green-dotted rectangle indicates the genes with better XCI status in 3F1 cells. The orange- dotted rectangle indicates the genes with better XCI status in F-M3-KO1. Note: Data are shown in log scale. (c) The distribution pattern of X-linked genes, the genes showing XCI deficiency in F-M3-KO1 differentiating cells and the special genes, which were silenced along the future Xi in undifferentiated 3F1 cells. (d) The allele-specific gene expression status of undifferentiated 3F1 cells and the comparison of the “XCI status” of undifferentiated 3F1 cells with undifferentiated F-M3-KO1 cells. Note: An allelotype score less than 0.2 is chosen as the threshold to call XCI on the X129 allele. (e) Quantitative RT-PCR results to show the Xist-mediated silencing on X-linked genes. ES cells were subjected to in vitro differentiation without G418 at the time of Dox treatment. The results are shown in relative fold expression. Normalization was performed using Actb. The expression level of each gene in undifferentiated cells is set as 1. Error bars indicate S.D. (n = 3). The statistical analysis used is the Student's t-test. p Value was calculated between the pair of data sets of each gene. *p b 0.05; **p b 0.01; ***p b 0.001. Mageb3 is involved in X chromosome inactivation 2741

(a) DAPI H3K27me3 Xist Merge

73% (n = 104) 3F1

10 µm

52% (n = 111) F-M3-KO1

DAPI mH2A Xist Merge

50% (n = 104) 3F1

10 µm

F-M3-KO1 42% (n = 110)

(b) Input IP (d) Mageb3-GFP Merge + + + + Pcgf3-3 x HA - + - + Mageb3-3 x Flag HA 95% (n = 20) Flag 10 M mH2A1.2-mCherry 293T (c) + + Pcgf3-3 x HA - + Mageb3-3 x Flag 15% Pcgf3-3 x HA (n = 20)

Mageb3-3 x Flag hnRNPU-mCherry

Tubulin

Fig. 4. The interactions between Mageb3 and other XCI proteins. (a) Immuno-RNA FISH to detect H3K27me3 (green), mH2A (green) and the Xist RNA (red). DNA was counterstained with DAPI (blue). The ES cells were subjected to in vitro differentiation for 6 days. The percentage of Xist clouds overlapping with an enrichment of the corresponding immunostaining is shown on the right. (b) Mageb3-Pcgf3 co-IP: Empty vector or Mageb3-Flag overexpressing plasmid, together with Pcgf3-HA overexpressing plasmid, was transfected into HEK293T cells. Forty-eight hours after transfection, cell lysates were prepared for co-IP experiment with anti-Flag M2 beads. (c) Mageb3 stabilizes Pcgf3: Empty vector or Mageb3-Flag overexpressing plasmid, together with Pcgf3-HA overexpressing plasmid, was transfected into HEK293T cells. Cells were harvested for Western blot at 48 h after transfection. (d) Live-cell images of Mageb3–GFP (green) co- expressed in 293T cells together with mH2A1.2–mCherry (red) or hnRNPU–mCherry (red). The white arrow indicates a nucleolus enrichment of Mageb3. The yellow arrow indicates an mH2A1.2 enrichment on Xi. The cells were transiently transfected with two plasmids. Within the cell population positive for both the green and the red fluorescent signals, the percentage of cells showing nucleolus enrichment of Mageb3 is shown on the right. 2742 Mageb3 is involved in X chromosome inactivation although the enrichment could be due to over- comprehensive RNA allelotyping along mouse X expression of Mageb3. mH2A1.2 is a histone variant chromosome and found that Mageb1–2 were not known to be enriched on Xi [18] (Fig. 4d). When 293T included in the data because no suitable SNPs were cells were co-transfected with Mageb3–GFP and found [14]. Therefore, there is still no direct evidence mH2A1.2–mCherry constructs, the results confirm showing the XCI status of Mageb1–2. It should also be that Mageb3 is not enriched on Xi (Fig. 4d). A previous noticed that the gene structure of Mageb2 is still study shows that Necdin, a MAGE family member, is awaiting for further elucidation, as the gene may have enriched in nucleoli and associates with heteroge- alternative first exons regulated by different promoters neous nuclear ribonucleoprotein U (hnRNP U) [20]. [9]. Unfortunately, we could not perform Western blot to hnRNP U is a nuclear matrix protein involved confirm the gene knockout status as there are no in XCI [21,22]. Interestingly, when Mageb3–GFP and available antibodies. It is possible that the mild mutant hnRNPU–mCherry were co-expressed, over- phenotypes observed in mutant cells are due to expression of hnRNP U disrupted the nucleolus incomplete gene knockout. enrichment of Mageb3 (Fig. 4d). To further confirm Mageb1–3 are likely involved in XCI possibly this result, we overexpressed Mageb3–GFP and through a mechanism related to nuclear matrix or hnRNPU–mCherry in mouse ES cells and obtained chromosomal architecture. Mouse chromosomes similar results (Supplementary Fig. 10). These results are telocentric, which means that the centromere is imply the interaction between Mageb3 and hnRNP U located at one end of a chromosome. The two ends and provide insights into the functional mechanism of of each mouse chromosome can be distinguished as Mageb3 in XCI. the proximal end and the distal end, based on their distance to the centromere. Interestingly, the distal telomere of mouse X chromosome can be distin- Discussion guished from the proximal telomere and all the autosomal telomeres by its association with a large Taken together, our study identifies Mageb1–3 as signal of telomeric repeat-containing RNA (Terra) genes functionally important for XCI. Poorly charac- [27]. Interestingly, the genes, which showed pre- terized in previous studies, Mageb1–3 belong to emptive XCI status in undifferentiated 3F1 cells, are melanoma-associated antigen (MAGE) gene family. clustered in a ~30-Mb chromosomal region close to The first member of the gene family, MAGEA1,was the distal end of mouse X chromosome. identified as a tumor antigen from a melanoma patient Many MAGE genes are X-linked [23]. MAGEB1–3 who had an unusually favorable clinical course [23]. are all X-linked in the human genome. In the mouse, Thereafter, a large gene family (N40 genes in the Mageb1–2 are X-linked and Mageb3 is located on human genome) was identified [24]. The region of chromosome 2. Mageb3 is a single-exon gene and homology of the gene family is the MAGE domain. Like likely a retrogene originated from its progenitor genes MAGEA1, many genes of the family are completely along X chromosome [28]. It is interesting to study the silenced in normal adult tissues except male germ roles of MAGEB1–3 during XCI in human. It is also cells, occasionally placenta and ovary, but ectopically interesting to study whether MAGE proteins are expressed in various types of tumors. Since many involved in meiotic silencing of the sex chromosomes. MAGE proteins are tumor antigens, significant effort has been concentrated on exploring their potentials in immunotherapy, although little success has been Materials and Methods achieved [24]. The molecular function of some of the MAGE genes was only identified recently [24]. Previous studies show Mageb family genes express Cell lines and culture in migrating PGCs [23] and during spermatogenesis [25]. Interestingly, spermatogenesis, especially the Mouse fibroblast cells and HEK293T cells were pachytene stage, is special time window in the life cultured in DMEM medium with 10% FBS at 37 °C in cycle of a male during which the X chromosome is a5%CO2 incubator. Mouse ES cells were cultured silenced [26]. Here, we show that Mageb1–3 are in 2i medium with Lif [29]. LS2 is a wild-type female expressed in ES cells and involved in XCI. Our data ES cell line derived in-house using 2i medium with show that Mageb1–3 are expressed at the highest Lif. J1 is a wild-type male ES cell line (ATCC SCRC- level in undifferentiated ES cells and the expression 1010). Ainv15 was obtained from Dr. Chen Lingyi as level decreases during in vitro differentiation (Fig. 2f). a gift. Primary mouse embryonic fibroblasts were The significantly steeper decrease of Mageb1–3 isolated from d13.5 mouse embryos. expression in differentiating female ES cells is possibly duetotheXCIstatusoftheX-linkedMageb1–2.We Genetic screen using lentiviral shRNA library tried to design allele-specific primers to distinguish Mageb1–2 from Mageb3, but failed to obtain conclu- A pooled lentiviral shRNA library was purchased sive results. We also checked our previous data of from Sigma-Aldrich (Cat# SHPM01). The library Mageb3 is involved in X chromosome inactivation 2743 contains 10 sub-pools. Two million iXist cells were TCCAGAAGTCATTCTTGTTGTCACTGCCATCAAG transduced by each sub-pool separately at a MOI of CTTAGGCAATACAGGTGGATGTGGTGTTCCTGG 0.5. Two million transduced cells were used as the GATCCTGT-3′. starting cell population (day 0) for in vitro differenti- ation during which cells were cultured using differ- Validation of individual shRNA entiation medium containing 1 μg/ml Dox, 400 μg/ ml G418 and 50 μg/ml L-ascorbic acid. For the first An shRNA system (OligoEngine, pSUPER RNAi 4 days of in vitro differentiation, embryoid bodies System) was used. shRNA sequences were de- (EBs) were culture in suspension. On day 4, EBs signed according to shRNA clones of the lentiviral were transferred to a gelatinized T75 tissue culture shRNA library (Sigma-Aldrich, Cat# SHPM01). flask. On day 5, EBs failed to attach to the tissue culture flask were washed away. On day 8, surviving Flow cytometry/FACS cells from 10 flasks transduced by different sub- pools of the shRNA library were harvested and At day 8 post-differentiation, cells were trypsinized combined. Genomic DNA was isolated and subject- and suspended in PBS. FACS was carried out on a ed to PCR amplification using the following two PCR BD LSR II flow cytometer (BD Biosciences, USA). primers: Backbone-F (5′-TACAAAATACGTGACG- Data acquisition and analysis were carried out with TAGAAA-3′ ); Backbone-R (5′ -TTTG BD FACSDiva™ software (BD Biosciences, USA). TTTTTGTAATTCTTTA-3′). To avoid over amplifica- At least 10,000 events were acquired and recorded tion, the PCR reaction was terminated within the for each sample. exponential range monitored by a real-time PCR system (CFX Connect, Bio-Rad). The sequencing Immuno-RNA FISH and DNA FISH primer used is shRNA-Read (5′-ACACCTTGTG GAAAGGACGAAACACCGG-3′). The sequencing In vitro differentiating ESCs were transferred onto was carried out on an Illumina MiSeq system using coverslips on day 4. On day 6, cells were washed the MiSeq Nano flow cell. About 1.1 million reads in with PBS and fixed with 4% paraformaldehyde in total and about 181k reads of each sample were PBS for 5 min. Prior to immunofluorescence, cells obtained from the sequencing run. on coverslips were permeabilized with 0.5% Triton X-100 in PBS for 30 min. Immuno-RNA FISH and Genome editing using the CRISPR/Cas9 system DNA FISH were carried out as previously described [30]. Immunofluorescence was performed using gRNA was cloned into px330 (Addgene plasmid mouse monoclonal antibodies against macroH2A.1 #42230). The following nucleotide oligo was used as (Abcam, Cat# ab183041) and H3K27me3 (Merck, the repair template. iXist-KO1: 5′-CTCTGCAGTGG Cat# 07-449) with a secondary antibody conjugated GCTGAACTACTTGAACCTCCCTGCGT with Alexa Fluor 488 (green) (Abcam, Cat# GACTGCTGTCGTTTAGCACGGGAAAGCTTAAGG ab150077). Immunostaining was followed by RNA CATGTTGACTGTGCTCAGGACTGAAGTAGAA FISH. The Xist RNA was detected with Sx9 probe, a GAGCTGACAGGCAGGAATCAGGACT-3′. Itm2b P1 DNA construct containing a 40-kb genomic KO: 5′-ATCCGGAGGGACGATGAGCGCCTCCTCG fragment covering the Xist gene. Nucleotide ana- CTGCTCTTGGGCTCGTCCTTCTTGGCCTCCTTA logs used in probe labeling were Cy3-dUTP AGCTTACGCCGAGTTGAACGTCACCTTCACC (Amersham, Cat# PA53022). For DNA FISH, X ATGGTGCGGCCTGGGGCGCCCTAGGCGGC and Y chromosomes were detected using mouse GACGGC-3′.Apex1KO:5′-AAGAAGAGTAAGGGGG chromosome X paint (MetaSystems, Cat# D-1420- CAGCAAAGAAAACCGAGAAGGAGGCCGCGGGA 050-OR) and mouse chromosome Y paint (Cambio, GAGGGCCCTGTCTAAGCTTCTGTACGAG Cat# 1189-YMF-02), respectively. Microscopy work GACCCTCCAGATCAGAAAACCTCACCCAGTGG was carried out on an Eclipse Ti microscope (Nikon) CAAATCTGCCACACTCAAG-3′. Kdm5c KO: 5′-GGC with a digital camera (Clara Series model C01, TACCCGGATCCACCCACAAACTCGCCGCTGGC Andor). GACCATCAGCTCCCAGCCTTACCGCAAGCTT AGGGTGGGCGGATCTTGCAAATGCCG Quantitative RT-PCR GACTTCTCGGCGATGGGTCTGATTTTCGC GATGTA-3′.Mul1KO:5′-ACCTTCGATGACAGCATA Total RNA was isolated by TRIzol (Life technologies). AGGCACACACTTCCCCGGTGCTTCTGAAA cDNA was synthesized using iScript reverse transcrip- GAATGCCCTTTAAGCTTAATGAATCTTCT tion kit (170-8840, Bio-Rad). The real-time PCR was TAGCTCCCTAGATGCAAAATAAATAGATAGA carried out on the CFX Connect real-time PCR system TATTCAGTTGGCCTCTGA-3′.Phc3KO:5′-AGAAA (Bio-Rad) using the SsoAdvanced Universal STBY TACCTGAACAGCATGTCGGTCTGACCCGCTG Green Supermix (Bio-Rad). The following PCR primers TAAAGCTTATGACACGGATGTTCCAGA were used: Beta-actin (Actb)-F (5′-ACTGCCGCATC GCTCGGCTCACTATCCATA-3′.Scmh1KO:5′-GTC CTCTTCCTC-3′); Actb-R (5′-CCGCTCGTTGCCAAT 2744 Mageb3 is involved in X chromosome inactivation

AGTGA-3′); Nanog-F (5′-TCGAATTCTGGGAAC TAGTGT-3′); CA-2-FA.Indx93Sol (5′-CAAGCAGAA GCCTC-3′); Nanog-R (5′-GTCTTCAGAGGAAGGGC GACGGCATACGAGATATTGGCCGGTCTGCCAT GAG-3′); Gata4-F (5′-CCTGGAAGACACCCCA CCGACGGTAGTGT-3′). The following sequencing ATCTC-3′); Gata4-R (5′-GGTAGTGTCCCGTCCCAT primers were used: Read1.Miseq (5′-ATCGGCTA C-3′); Fgf5-F (5′-CCTTGCGACCCAGGAGCTTA-3′); CACGCCTATCGGGAAGCTGAAG-3′); IndexRead Fgf5-R (5′-CCGTCTGTGGTTTCTGTTGAGG-3′); (5′-ACACTACCGTCGGATGGCAGACCG-3′). Se- M123-F (5′-CCCAGCACTCATTCCTATTTGC-3′); quencing was carried out on a NextSeq system. M123-R (5′-GGAGACCTGTCCTAGGCAAC-3′); mHprt_qF (5′-GATTAGCGATGATGAACCAGGTT- shRNA knockdown of Pcgf3 3′); mHprt_qR (CCTCCCATCTCCTTCATGACA-3′); tdTomato_qF (5′-CCGACATCCCCGATTACAAGAA An shRNA system (OligoEngine, pSUPER RNAi GC-3′); tdTomato_qR (5′-TTGTAGATCAGCGTG System) was used to knockdown mouse polycomb CCGTC-3′); mGpc4_qF (5′-GGCAGCTGGCACTAG group ring finger 3 (Pcgf3, gene ID: 69587). Three TTTG-3′); mGpc4_qR (5′-AACGGTGCTTGGGAGAG pairs of DNA oligos containing the desired shRNA AG-3′); G418_qF (5′-GGCTATTCGGCTATGACT sequences were used as a mixture: 95495F (5′-GATC GGGC-3′); G418_qR (5′-GCAGTTCATTCAGGGCA CCCGCTACACTACAGACCCAAGATTTCAAGA CCG-3′); Mecp2_qF (5′-CAGGGAGGAAAAGTCAGA GAATCTTGGGTCTGTAGTGTAGCTTTTTA-3′); AGACC-3′); Mecp2_qR (5′-AATGGTGGGCTGAAGG 95495R (5′-AGCTTAAAAAGCTACACTACAGACCC TTGTA-3′); qXist E1_F (5′-CGGCCTCTAGTTTGTCC AAGATTCTCTTGAAATCTTGGGTCTGTAGTG ATT-3′); qXist E1_R (5′-GATGGCATGATGGAATT TAGCGGG-3′); 95496F (5′-GATCCCCCCATTACAG GAG-3′). TATATCGGTCATTTCAAGAGAATGACCGATA TACTGTAATGGTTTTTA-3′); 95496R (5′-AGCTTAA Padlock probe SNP capture AAACCATTACAGTATATCGGTCATTCTCTTG AAATGACCGATATACTGTAATGGGGG-3′); 95497F The padlock probe library design and padlock SNP (5′-GATCCCCCCTCCTTTAATGAGCTGGACATT capture procedure were carried out as previously CAAGAGATGTCCAGCTCATTAAAGGAGG described [14]. In brief, each reaction was carried out in TTTTTA-3′); 95497R (5′-AGCTTAAAAACCTCCTTTA 20 μl volume containing 1 unit Ampligase (A3210K, ATGAGCTGGACATCTCTTGAATGTCCAGCTCAT Epicenter), 1 unit Phusion High-Fidelity DNA Polymer- TAAAGGAGGGGG-3′). ase (M0530, New England BioLabs), 1× Phusion High-Fidelity DNA Polymerase buffer, 10 nM dNTP. Immunoprecipitation cDNA was generated using SuperScript III first-strand synthesis system (18080-051, Life Technologies). Plasmids expressing Mageb3–3×Flag and/or Two hundred nanograms of single-stranded cDNA Pcgf3–3×HA were transfected into HEK293T cells. and 2 pmol padlock probe were used in each reaction. Forty-eight hours after transfection, cells were har- Nicotinamide adenine dinucleotide (NAD+) was pro- vested, and cell extracts were prepared in lysis buffer vided in each reaction at a final concentration of [20 mM Tris–HCl (pH 8.0), 137 mM NaCl, 10% 0.5 mM. glycerol, 1% NP-40 and 2 mM EDTA] with protease The multiplexed sequencing libraries were PCR inhibitor (Roche), on ice for 30 min. After centrifugation amplified in a real-time PCR system (CFX Connect, at 12,000g for 20 min, the supernatant was collected Bio-Rad) using the following primers: CA-2-RA. and incubated with anti-Flag M2 magnetic beads Miseq (5′-AATGATACGGCGACCACCGAGATCTA (Sigma) at 4 °C overnight. The beads were washed TCGGCTACACGCCTATCGGGAAGCTGAAG-3′); three times with lysis buffer, and the bound proteins CA-2-FA.Indx4Sol (5′-CAAGCAGAAGACGGCA were released from the beads by boiling the beads in TACGAGATTGGTCACGGTCTGCCATCCGACGG 2× SDS loading buffer for 5 min. Western blot TAGTGT-3′);CA-2-FA.Indx7Sol (5′-CAAGCAGAAG was performed to detect the proteins present in IP ACGGCATACGAGATGATCTGCGGTCTGCC samples. ATCCGACGGTAGTGT-3′); CA-2-FA.Indx18Sol (5′- CAAGCAGAAGACGGCATACGAGATGCG Western blot GACCGGTCTGCCATCCGACGGTAGTGT-3′);CA- 2-FA.Indx45Sol (5′-CAAGCAGAAGACGGCATACG Cells were lysed, and total protein concentration AGATCGTAGTCGGTCTGCCATCCGACGG was measured using BCA Protein Assay Kit (Beyo- TAGTGT-3′); CA-2-FA.Indx76Sol (5′-CAAGCAGAA time) to ensure equal loading. Samples were GACGGCATACGAGATAATAGGCGGTCTGC resolved by SDS-PAGE followed by transferring CATCCGACGGTAGTGT-3′); CA-2-FA.Indx91Sol onto a PVDF membrane (Millipore). Membranes (5′ -CAAGCAGAAGACGGCATACGAGATA were probed with primary antibodies. Bound primary CATCGCGGTCTGCCATCCGACGGTAGTGT-3′); antibodies were recognized by HRP-linked second- CA-2-FA.Indx92Sol (5′-CAAGCAGAAGACGGCAT ary antibodies (GE Health-care). HRP activity was ACGAGATTCAAGTCGGTCTGCCATCCGACGG detected by ECL Plus (Beyotime). Digital images Mageb3 is involved in X chromosome inactivation 2745 were taken by the automatic chemiluminescence References imaging analysis system (Tanon).

[1] B. Payer, J.T. Lee, X chromosome dosage compensation: how mammals keep the balance, Annu. Rev. Genet. 42 (2008) 733–772. Acknowledgments [2] R. Basu, L.F. Zhang, X chromosome inactivation: a silence that needs to be broken, Genesis 49 (2011) 821–834. This work was supported by Singapore Ministry [3] A. Minajigi, J.E. Froberg, C. Wei, H. Sunwoo, B. Kesner, D. of Education Academic Research Fund Colognori, et al., Chromosomes. A comprehensive Xist (MOE2015-T2-1-093) and by the Singapore interactome reveals cohesin repulsion and an RNA-directed National Research Foundation under its Cooper- chromosome conformation, Science 349 (2015). ative Basic Research Grant administered by the [4] C.A. Mchugh, C.K. Chen, A. Chow, C.F. Surka, C. Tran, P. Singapore Ministry of Health's National Medical Mcdonel, et al., The Xist lncRNA interacts directly with SHARP to Research Council (NMRC/CBRG/0006/2012). silence transcription through HDAC3, Nature 521 (2015) 232–236. The authors declare no competing financial [5] C. Chu, Q.C. Zhang, S.T. da Rocha, R.A. Flynn, M. Bharadwaj, interests. J.M. Calabrese, et al., Systematic discovery of Xist RNA binding proteins, Cell 161 (2015) 404–416. [6] A. Monfort, G. Di Minin, A. Postlmayr, R. Freimann, F. Arieti, Data Availability S. Thore, et al., Identification of Spen as a crucial factor for Xist function through forward genetic screening in haploid The sequencing data of this study are available in embryonic stem cells, Cell Rep. 12 (2015) 554–561. sequence read archive (accession number [7] B. Moindrot, A. Cerase, H. Coker, O. Masui, A. Grijzenhout, SRP140616). G. Pintacuda, et al., A pooled shRNA screen identifies Rbm15, Spen, and Wtap as factors required for Xist RNA- mediated silencing, Cell Rep. 12 (2015) 562–572. [8] K. Kuroda, H. Han, S. Tani, K. Tanigaki, T. Tun, T. Furukawa, Appendix A. Supplementary Data et al., Regulation of marginal zone B cell development by MINT, a suppressor of Notch/RBP-J signaling pathway, Supplementary data to this article can be found Immunity 18 (2003) 301–312. online at https://doi.org/10.1016/j.jmb.2018.05.031. [9] O. De Backer, A.M. Verheyden, B. Martin, D. Godelaine, E. De Plaen, R. Brasseur, et al., Structure, chromosomal location, Received 11 January 2018; and expression pattern of three mouse genes homologous to the human MAGE genes, Genomics 28 (1995) 74–83. Received in revised form 17 May 2018; [10] M. Almeida, G. Pintacuda, O. Masui, Y. Koseki, M. Gdula, A. Accepted 17 May 2018 Cerase, et al., PCGF3/5-PRC1 initiates Polycomb recruit- Available online 23 May 2018 ment in X chromosome inactivation, Science 356 (2017) 1081–1084. Keywords: [11] L. Cartegni, S.L. Chew, A.R. Krainer, Listening to silence and Xist; understanding nonsense: exonic mutations that affect Mageb3; splicing, Nat. Rev. Genet. 3 (2002) 285–298. hnRNP U; [12] J.T. Lee, N. Lu, Targeted mutagenesis of Tsix leads to chromatin architecture; nonrandom X inactivation, Cell 99 (1999) 47. nuclear matrix [13] K. Zhang, J.B. Li, Y. Gao, D. Egli, B. Xie, J. Deng, et al., Digital RNA allelotyping reveals tissue-specific and allele- † specific gene expression in human, Nat. Methods 6 (2009) These authors contributed equally to this work. 613–618. [14] R. Hong, B. Lin, X. Lu, L.T. Lai, X. Chen, A. Sanyal, et al., Abbreviations used: High-resolution RNA allelotyping along the inactive X MAGE, melanoma-associated antigen; Xist, inactivated X chromosome: evidence of RNA polymerase III in regulating chromosome specific transcript; ES cells, embryonic stem chromatin configuration, Sci. Rep. 7 (2017), 45460. cells; Xi, inactive X chromosome; Pcgf3, Polycomb group ring [15] B.K. Sun, A.M. Deaton, J.T. Lee, A transient heterochromatic finger 3; hnRNP U, heterogeneous nuclear ribonucleoprotein state in Xist preempts X inactivation choice without RNA U;lncRNA,longnoncodingRNA;XCI,Xchromosome stabilization, Mol. Cell 21 (2006) 617–628. inactivation; Hprt, hypoxanthine guanine phosphoribosyl [16] K. Plath, J. Fang, S.K. Mlynarczyk-Evans, R. Cao, K.A. transferase; iXist, A male ES cell line with an inducible Xist Worringer, H. Wang, et al., Role of histone H3 lysine 27 methylation in X inactivation, Science 300 (2003) 131–135. transgene; FISH, fluorescence in situ hybridization; G418, [17] R. Cao, L. Wang, H. Wang, L. Xia, H. Erdjument-Bromage, P. geneticin; shRNA, short hairpin RNA; RT-PCR, reverse Tempst, et al., Role of histone H3 lysine 27 methylation in transcription polymerase chain reaction; gRNA, guide RNA; Polycomb-group silencing, Science 298 (2002) 1039–1043. FACS, fluorescence-activated cell sorting; SNP, single- [18] C. Costanzi, J.R. Pehrson, Histone macroH2A1 is concentrat- nucleotide polymorphism; co-IP, co-immunoprecipitation; ed in the inactive X chromosome of female mammals, Nature EBs, embryoid bodies; Actb, beta-actin. 393 (1998) 599–601. 2746 Mageb3 is involved in X chromosome inactivation

[19] J.M. Doyle, J. Gao, J. Wang, M. Yang, P.R. Potts, MAGE– [25] F. Clotman, O. De Backer, E. De Plaen, T. Boon, J. Picard, Cell- RING protein complexes comprise a family of E3 ubiquitin and stage-specific expression of mage genes during mouse ligases, Mol. Cell 39 (2010) 963–974. spermatogenesis, Mamm. Genome 11 (2000) 696–699. [20] H. Taniura, K. Yoshikawa, Necdin interacts with the ribonucleo- [26] J.M. Turner, Meiotic silencing in mammals, Annu. Rev. protein hnRNP U in the nuclear matrix, J. Cell. Biochem. 84 Genet. 49 (2015) 395–412. (2002) 545–555. [27] L.F. Zhang, Y. Ogawa, J.Y. Ahn, S.H. Namekawa, S.S. Silva, [21] Y. Hasegawa, N. Brockdorff, S. Kawano, K. Tsutui, S. J.T. Lee, Telomeric RNAs mark sex chromosomes in stem Nakagawa, The matrix protein hnRNP U is required for cells, Genetics 182 (2009) 685–698. chromosomal localization of Xist RNA, Dev. Cell 19 (2010) [28] P.J. Wang, X chromosomes, retrogenes and their role in male 469–476. reproduction, Trends Endocrinol. Metab. 15 (2004) 79–83. [22] R. Helbig, F.O. Fackelmayer, Scaffold attachment factor A [29] Q.L. Ying, J. Wray, J. Nichols, L. Batlle-Morera, B. Doble, J. (SAF-A) is concentrated in inactive X chromosome territories Woodgett, et al., The ground state of embryonic stem cell through its RGG domain, Chromosoma 112 (2003) 173–182. self-renewal, Nature 453 (2008) 519–523. [23] A.J. Simpson, O.L. Caballero, A. Jungbluth, Y.T. Chen, L.J. [30] L.F. Zhang, K.D. Huynh, J.T. Lee, Perinucleolar targeting of the Old, Cancer/testis antigens, gametogenesis and cancer, Nat. inactive X during S phase: evidence for a role in the maintenance Rev. Cancer 5 (2005) 615–625. of silencing, Cell 129 (2007) 693–706. [24] J.L. Weon, P.R. Potts, The MAGE protein family and cancer, Curr. Opin. Cell Biol. 37 (2015) 1–8.