UBC Faculty Research and Publications

Impact of flanking chromosomal sequences on localization and silencing by the human non-coding RNA XIST Kelsey, Angela D; Yang, Christine; Leung, Danny; Minks, Jakub; Dixon-McDougall, Thomas; Baldry, Sarah E; Bogutz, Aaron B; Lefebvre, Louis; Brown, Carolyn J Oct 2, 2015

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
52383-13059_2015_Article_774.pdf [ 2.05MB ]
Metadata
JSON: 52383-1.0361961.json
JSON-LD: 52383-1.0361961-ld.json
RDF/XML (Pretty): 52383-1.0361961-rdf.xml
RDF/JSON: 52383-1.0361961-rdf.json
Turtle: 52383-1.0361961-turtle.txt
N-Triples: 52383-1.0361961-rdf-ntriples.txt
Original Record: 52383-1.0361961-source.json
Full Text
52383-1.0361961-fulltext.txt
Citation
52383-1.0361961.ris

Full Text

RESEARCH Open AccessImpact of flanking chromosomal sequenceson localization and silencing by the humannon-coding RNA XISTAngela D. Kelsey1†, Christine Yang1†, Danny Leung2,3, Jakub Minks1, Thomas Dixon-McDougall1, Sarah E.L. Baldry1,Aaron B. Bogutz1, Louis Lefebvre1 and Carolyn J. Brown1*AbstractBackground: X-chromosome inactivation is a striking example of epigenetic silencing in which expression of thelong non-coding RNA XIST initiates the heterochromatinization and silencing of one of the pair of X chromosomesin mammalian females. To understand how the RNA can establish silencing across millions of basepairs of DNA wehave modelled the process by inducing expression of XIST from nine different locations in human HT1080 cells.Results: Localization of XIST, depletion of Cot-1 RNA, perinuclear localization, and ubiquitination of H2A occurs atall sites examined, while recruitment of H3K9me3 was not observed. Recruitment of the heterochromatic featuresSMCHD1, macroH2A, H3K27me3, and H4K20me1 occurs independently of each other in an integration site-dependentmanner. Silencing of flanking reporter genes occurs at all sites, but the spread of silencing to flanking endogenoushuman genes is variable in extent of silencing as well as extent of spread, with silencing able to skip regions. Thespread of H3K27me3 and loss of H3K27ac correlates with the pre-existing levels of the modifications, and overall theextent of silencing correlates with the ability to recruit additional heterochromatic features.Conclusions: The non-coding RNA XIST functions as a cis-acting silencer when expressed from nine different locationsthroughout the genome. A hierarchy among the features of heterochromatin reveals the importance of interactionwith the local chromatin neighborhood for optimal spread of silencing, as well as the independent yet cooperativenature of the establishment of heterochromatin by the non-coding XIST RNA.Keywords: XIST, Long non-coding RNA, Dosage compensation, X-chromosome inactivation, Nucleolar localization,Facultative heterochromatin, SMCHD1, macroH2A, H3K27me3, H4K20me1BackgroundTo avoid a functional gene dosage imbalance between thesexes, one of the two X chromosomes in female placentalmammals is transcriptionally silenced [1]. This process ofX-chromosome inactivation (XCI) occurs early in develop-ment and is generally random in all human tissues witheither the paternal or maternal X chromosome becom-ing the inactive X (Xi). The X-inactivation centre (XIC),which is located at Xq13 in humans, is the region ofthe X that is necessary for the chromosome to be inac-tivated, and contains the XIST gene (X-inactive specifictranscript) [2–4]. Remarkably, the approximately 17 kbspliced and polyadenylated long non-coding XIST RNAuniquely localizes to the chromosome from which it istranscribed [5]. The coating of the Xi by the XIST RNAresults in a substantial epigenetic transformation, losingepigenetic modifications associated with active chromatin(notably histone acetylation) and gaining modificationsassociated with inactive chromatin (including H3K27me3,H3K9me2/3, H4K20me1, and H2AK119u1). The Xi alsobecomes enriched in several other proteins, including thehistone variant macrohistone H2A (macroH2A), the nu-clear matrix protein hnRNPU, and the epigenetic regula-tors SMCHD1 and ASH2L ([6, 7], reviewed in [8]). Inaddition, the Xi is peripherally or perinucleolarly localized[9] with perinucleolar targeting during S phase suggested* Correspondence: Carolyn.Brown@UBC.Ca†Equal contributors1Department of Medical Genetics, Molecular Epigenetics Group, Life SciencesInstitute, University of British Columbia, Vancouver, CanadaFull list of author information is available at the end of the article© 2015 Kelsey et al. Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, andreproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link tothe Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver(http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.Kelsey et al. Genome Biology  (2015) 16:208 DOI 10.1186/s13059-015-0774-2to be important for maintenance of silencing. A furtherfeature of the Xi is silencing of repetitive elements, asvisualized by loss of RNA hybridization with a Cot-1probe for repetitive DNA [10] resulting in what has beentermed a ‘Cot-1 hole’.The timing of acquisition of these features has beenbest studied in mouse, where the differentiation of em-bryonic stem cells (ESCs) provides an in vitro model forthe events of XCI (reviewed in [11]). Early studies in mousesuggested the presence of a developmental window beyondwhich Xist was unable to induce X-chromosome inactiva-tion [12], although macroH2A could be recruited [13]. Afterthis stage, Xist expression was no longer required for main-tenance of silencing [14], consistent with studies in humanshowing maintenance of silencing in the absence of XIST[15, 16]. SATB1 has been suggested to be involved in defin-ing such a window for Xist function [17]; however, Satb1/Satb2-deficient mice are able to undergo X inactivation [18].More recent studies have shown that induction of Xist canrecruit H3K27me3 in mouse somatic cells [19], and XISTinduction recruits multiple features of the Xi in humansomatic cells [10, 20]. In addition, an ongoing role for Xistin stable silencing of the Xi has been shown by loss of Xistresulting in gene reactivation, loss of perinucleolar associ-ation and loss of H3K27me3 [21]. Loss of X-linked gene si-lencing is enhanced by disruption of DNA methylation andother pathways that cooperate with Xist, thus reactivation ofX-linked markers has been used to identify additionalplayers in the pathway [22–24]. Characterization of the on-going role for XIST in somatic cells has important implica-tions for disease, as highlighted by a recent study showingthat deletion of Xist results in hematological malignancies infemale mice due to reactivation of X-linked genes [25].There is substantial evidence for cooperativity of multiplesilencing pathways in the initial silencing of the chromo-some, with XCI able to proceed in the absence of key com-ponents of the silencing machinery such as PRC2 [26–28],PRC1 [29], or macroH2A [30]. In addition to multiple fac-tors cooperating in the process of XCI, different X-linkedgenes may be silenced (or maintained silent) by differentplayers. For example, mutation of the Smchd1 gene resultsin loss of DNA methylation and partial to full reactivationof approximately 20 % of the X-linked genes in mouse [31,32]. Surprisingly, many of the marks of an Xi can be re-cruited by a mouse transgene containing a deletion thatmakes the Xist RNA defective in silencing, although oftenthe recruitment is not as effective as seen with a full-lengthXist (reviewed in [8]). This silencing defective Xist RNA isalso able to form a Cot-1 hole [33], consistent with the Cot-1 hole not reflecting X-linked gene silencing [34], but rathera core of silenced non-coding DNA [35].Spread of silencing to autosomal genes has been observedin unbalanced X/autosome translocations; however, the ex-tent of autosomal silencing is highly variable in both humans(reviewed in [36]) and mice [37]. Silencing of autosomalgenes has also been observed upon integration of Xist/XIST transgenes into autosomes, and localization of theRNA to the autosome is able to induce many features ofthe Xi including nucleolar localization [10, 20, 21, 38]. Re-cently, an XIST transgene was integrated into chromo-some 21 in induced pluripotent stem cells from anindividual with Down’s syndrome and corrected gene ex-pression from chromosome 21 to near normal disomiclevels [39]. Together, these studies demonstrate the XISTRNA is able not only to spread along autosomal materialbut also to recruit some of the heterochromatic featuresassociated with XCI to autosomes. The silencing of the tri-somic chromosome 21 in induced pluripotent stem cellswas proposed as a first step towards ‘chromosome therapy’[39], and for such uses of XIST a better understanding ofthe influence of the chromatin neighborhood is necessary.XIST expression is required to induce the cascade ofchanges that cooperatively silence the X, but relatively lit-tle is known about the process by which the non-codingRNA recruits these changes. In human somatic cells wehave reported the recruitment of several features of XCI,including gene silencing, following XIST expression froman inducible transgene [20, 40]. The separation of XISTexpression from the myriad of changes that occur duringdifferentiation provides an opportunity to dissect the roleof XIST in XCI. By using this inducible XIST transgene toexamine the influence of XIST expression induced fromnine different integration sites we aimed to establish ahierarchy to the features that are recruited by XIST anddetermine if any were influenced by the genomic contextof the XIST integration. We were able to identify features(Cot-1 hole formation, perinucleolar localization, proximalreporter silencing, and H2AK119u1) that are recruitedto all integrations and thus are compatible with any ofthe sites/genomic contexts tested, a feature that is notrecruited to any site examined (H3K9me3) while anotherset of features (macroH2A, H3K27me3, SMCHD1, andH4K20me1 recruitment) are dependent on genomic con-text, but appear independent of each other.ResultsXIST RNA expression results in depletion of Cot-1 RNA andincreased perinucleolar locationIn the HT1080 cells an inducible XIST cDNA has beenpreviously reported to localize in cis to the chromosomefrom which it is transcribed, resulting in the recruitmentof some chromatin modifications and repression of bothflanking reporter genes as well as flanking endogenousgenes [20, 40]. To test whether there are differences inthe functionality of the XIST RNA due to influences oflocal DNA sequences and chromatin environment weintegrated a single copy of the full-length inducible XISTcDNA into nine different FRT sites in HT1080 cellsKelsey et al. Genome Biology  (2015) 16:208 Page 2 of 16containing a constitutive Tet-repressor transgene to allowinduction of XIST by treatment of the cells with the tetra-cycline analog doxycycline (DOX). The sites of the FRT in-tegrations were identified by inverse PCR and sequencingto be 1p, 3q, 4q, 7p, 7q, 8p, 12q, 15q, and an X chromo-some FRT site at Xq23 was previously reported [41]. Theintegration sites (described in Additional file 1) are in bothG-dark and G-light regions (four and five, respectively),and three integrations occurred within a gene.For each clone, the inducible promoter was activatedfor 5 days and the localization of XIST RNA assessed byRNA FISH. At each integration site the XIST RNA wasable to localize and formed an XIST cloud comparableto that observed in normal female cells (Fig. 1). The levelof XIST RNA varied between cell lines and also withindifferent cultures of the same cell line, showing fromfive- to 30-fold induction of XIST after 5 days of DOX(Additional file 1). The integrations into 3q, 7p, and 15q(all G-dark) showed lower expression by q-RT-PCR,which is consistent with a significantly smaller signal forXIST for the 3q integration clone relative to all other in-tegrations except for 15q (P ≤0.01). Co-hybridizationwith fluorescently labelled Cot-1 also showed depletionof Cot-1 hybridization coincident with the XIST cloud ateach integration site, observable in the line diagrams ofsignal intensity across the XIST cloud (Fig. 1). Recentlystable Cot-1 repeat RNA has been shown to be associ-ated with euchromatic chromosomes, yet excludedfrom the Xi resulting in the Cot-1 RNA hole [42]. Wenoted differences in the intensity of the Cot-1 holes;however, in attempting to quantify such differences itbecame apparent that the XIST RNA signal was oftenFig. 1 XIST RNA localizes and forms Cot-1 holes when expressed from nine different integration sites. Shown is XIST RNA FISH (green) upon expressionof an inducible XIST transgene integrated into the indicated chromosomal locations in HT1080 cells. IMR90 cells (a female fibroblast line) are shown asa positive comparator. Cells were counter-stained with DAPI (blue) and co-hybridized with Cot-1 (labelled with spectrum-red, but shown in grayscale).Arrow indicates the location of the XIST signal and reduction in Cot-1 staining. Graphs to the right show the RGB intensities across the lines shown inthe picture inserts drawn through the XIST clouds (XIST (green), Cot-1 (red), and DAPI (blue))Kelsey et al. Genome Biology  (2015) 16:208 Page 3 of 16at the nuclear or nucleolar periphery in the HT1080 cells,and that measuring the intensity of the Cot-1 hole couldbe influenced by nuclear location. The Xi is generally lo-cated at the nuclear or nucleolar periphery; however, asthese were autosomal integration sites, it seemed that ex-pression of XIST might be altering nuclear location.To test whether XIST expression was relocating thechromosome from which it was expressed, the XISTsignal was scored for being in contact with either thenuclear periphery or the nucleolus, prior to and afterXIST induction. Prior to DOX induction there was onlya small focus of XIST expression (see [20]); however thissignal was sufficient to identify the location of the inte-grated XIST. In six of the nine integration sites, induc-tion with DOX resulted in a significant increase inperinucleolar localization (P ≤0.05), a trend shared withthe other integration sites (Table 1). Perinuclear associ-ation, on the other hand, showed no significant differ-ence for five of the integrations, with three integrationsshowing a significant increase and the 8p integrationshowing a significant decrease (P ≤0.01; Table 1). Thefull distribution of localization before and after inductionof XIST is shown in Additional file 2. The 8p integrationsite showed the highest proportion of perinucleolar-associated XIST signals (56 %) and also the greatestincrease in perinucleolar association following XISTinduction (27 %). In mouse, localization of Xist to theperinucleolar compartment was shown to be necessaryfor the silencing activity of Xist [21], leading us to ques-tion whether the differing nucleolar recruitment andCot-1 hole formation that we observed might be reflect-ive of silencing ability.XIST RNA expression silences nearby reporter andendogenous genesWe previously reported silencing of a flanking EGFPreporter gene at the 3q integration site [20, 40]. At theother FRT sites we did not co-integrate a reporterconstruct; however, the integration of XIST into theFRT site results in expression of an upstream Hygromy-cin (Hyg) gene. Robust silencing of Hyg was observed atall integrations after 5 days of XIST expression (Fig. 2a),suggesting that XIST is able to silence a virally-derivedpromoter (SV40), consistent with our previous demon-stration that an EGFP reporter driven by the CMV pro-moter could be silenced [20]. The 7q integration siteshowed significantly less silencing than the 1p, 3q, 7p,8p, 15q, and Xq integration sites (P ≤0.01). As only therepeat A region of XIST is required for silencing of theflanking reporter genes [40], we generated a constructcontaining the XIST repeat A and a DsRed reporterdriven by the mouse Pgk1 gene promoter, which is nor-mally X-linked and subject to XCI (Fig. 2b). We integratedthis construct into six of the integration sites and againobserved consistent silencing of greater than 90 % (Fig. 2c),suggesting both viral and mouse-derived promoters couldbe silenced by XIST in the HT1080 cells. Consistent withprevious results with the EGFP reporter gene at 3q, silen-cing of the dsRed reporter was reversible when inductionof XIST expression was stopped by removal of DOX(Fig. 2d).Given the capacity of the XIST RNA to silence in cis,and the apparent spread of the RNA along the chromo-some based on our RNA FISH data, we questionedwhether there would be silencing of endogenous genesat additional sites adjacent to the XIST transgenes.The HT1080 cells remain diploid although they carryseveral structural rearrangements (46,XY,del (1)(p21),i(3)(p10), i(3)(q10), der(4)t(1;4)(p21;p16), der(5)t(5;5)(p15;?),der(11)t(3;11)(q11;q25) see additional details in methods).We generated allele-discriminating pyrosequencing assaysto examine silencing of candidate genes flanking theintegration sites (Fig. 3). The phase of the polymor-phisms relative to the integration was not known, butthe allelic expression change upon DOX induction ofXIST is presented as a cis-linked loss of expression asTable 1 Increased nucleolar association of chromosomes expressing transgenic XISTIntegration site % perinuclear % perinuclear % perinuclear % perinucleolar % perinucleolar % perinucleolarNo DOX DOX CHANGE No DOX DOX CHANGE1p 27 34 7 10 36 16a3q 78 70 −8 10 18 84q 57 74 17b 12 25 14a7p 32 27 −5 29 47 18a7q 30 53 23a 28 39 11c8p 59 39 −21a 29 56 27a12q 36 39 3 18 35 17a15q 24 36 14c 46 53 7Xq23 42 46 5 31 41 10The numbers shown are based on three independent experiments for the DOX results and one experiment for the No DOX results, with ≥50 cells counted for eachintegration site in each experiment. Chi-squared test (aP ≤0.001; bP ≤0.01; cP ≤0.05)Kelsey et al. Genome Biology  (2015) 16:208 Page 4 of 16previously demonstrated for the 3q integration forwhich we were able to assign the allelic loss to thechromosome bearing the inducible XIST [40]. Individ-ual pyrosequencing results for the locus closest to5 Mb from the integration site are shown in Fig. 3a,with the silencing percentages for all genes examinedshown in Fig. 3b as a function of distance from the inte-gration site (all assays are shown in Additional file 3).While no significant changes were observed for the con-trol integrations (clones with XIST integrated on differentchromosomes), all XIST integration sites except 4q showedat least one gene with significant allelic silencing. Therewas much more variability between integration sites forendogenous gene silencing than was seen for the silencingof Hyg. The 8p-integrated XIST clone displayed the mostsilencing, with four out of the five genes tested showing60–80 % silencing. The 4q clone, in contrast, showed nosignificant silencing for any of the three genes tested. Twodifferent integration sites on chromosome 7 showed quitedifferent results, with only one of seven genes assayedshowing over 20 % silencing for the 7p integration site,while five of the seven genes showed over 20 % silencingwith XIST expressed from the 7q integration site. Therewas also discontinuous spread of silencing. For example, inthe 1p integration site clone, two genes located approxi-mately 200 kb from the XIST transgene failed to silence(1 % silencing), whereas the RHBDL2 gene located ap-proximately 400 kb from the XIST transgene silenced byapproximately 70 %. In addition to variation between theintegration sites in the number of genes that were silenced,there were also significant differences in the extent of genesilencing between genes that showed silencing. More thanhalf of the significant changes demonstrated less than 50 %silencing of one allele, and a significant change as small as6 % for the ZNF710 gene on 15q was observed, indicatingthat XIST can cause a continuum of silencing.To distinguish differences between clones that areattributable to the integration site rather than clone-to-clone variation, we analyzed clones that were inde-pendent integrations of the XIST cDNA into the samegenomic location. Intriguingly, for the 12q integrationsite, we observed a significant correlation between theamount of silencing and the level of XIST expression(Fig. 3c). This correlation was observed for both culturesof the same single-cell clone and additional independentclones at the same integration site. Such a correlation wasnot observed for genes silenced by XIST expression fromthe 8p integration site using multiple cultures and twoindependent clones (Fig. 3d). Thus, while variation in thelevel of XIST expression occurred within a clone, the im-pact of this variation depended on the integration site, andwith similar XIST expression a similar extent of silencingwas observed between independent integrations into thesame FRT site. A second clone from the Xq integrationsite showed very similar silencing, while a second clonefrom the 7q integration site showed lower XIST expres-sion levels and failed to silence (Additional files 1 and 3),suggesting that the silencing ability of 7q integrations, like12q integrations, might be influenced by XIST expressionlevels. Since the clones at the 8p integration site si-lenced across a range of XIST levels, it seemed possiblethat, unlike silencing of the reporter genes, the main-tenance of silencing of endogenous genes might not beXIST-dependent, so we analyzed silencing of endogen-ous flanking genes after removal of DOX for the 8p00.20.40.60.811p* 3q* 7p 7q 12q 8pDsRED expression (5d (12d*) DOX / No DOX) 00.20.40.60.811p 3q 4q 7p 7q 8p 12q 15q Xq23Hygromycin expression  (5d DOX / No DOX) a b c 00.20.40.60.811p 3qDsRED expression (7d DOX5d No DOX/ No DOX)d 1 kb Fig. 2 Silencing of flanking reporter genes upon XIST expressionfrom various integration sites. a Relative level of Hyg expressionafter 5 days of XIST expression induced by DOX treatment comparedwith no DOX levels, measured by q-RT-PCR, for each of nine differentintegration sites as listed. Error bars show the standard deviationof biological triplicates. A one-way ANOVA with Tukey’s MultipleComparison Test gives the following differences: 1p and 7q**, 3qand 7q**, 7p and 7q*, 7q and 8p**, 7q and 15q***, 7q and Xq**(*P ≤0.05; **P ≤0.01; ***P ≤0.001). b Map of transgene containing aninducible construct of the 5’A repeats of XIST as well as a DsRed-Express2reporter gene expressed from a constitutive Pgk1 promoter. c Silencingof DsRed-Express2 relative to no DOX cells, measured by flow cytometry,after 5 or 12 days of DOX induction of 5’A region of XIST (from constructshown in part b) that had been integrated into six different integrationsites as listed. The error bars represent ±1 s.d. of the silencinglevels of individual single-cell clones (N = 8–11). d DsRed-Express2expression, measured by flow cytometry, following induction(7 days DOX) and subsequent repression (5 days no DOX) ofrepeat A of XIST at two integration sites (1p and 3q)Kelsey et al. Genome Biology  (2015) 16:208 Page 5 of 160204060801000204060801000204060801000.00 0.05 0.10 0.15 0.20020406080100XIST/PGK1%silencing0.2 0.4 0.6-20020406080100XIST/PGK1%silencingab3q1p 4q 7p 7q 8p 12q 15q XqcMMACHC - 1p DDX60 - 4q AOAH - 7pp=0.0022ACN9 - 7qRespective No DOXRespective DOXControl No DOXControl DOXNEIL2 - 8p%allele freq p=0.0050 p=0.0022POLR3B - 12qp=0.0152ZNF710 - 15q020406080100%allele freq0204060801000204060801000204060801000204060801000204060801000204060801008pCTSB: nsSTC1: nsd12qOAS3: r=0.9341***POLR3B: r=0.7918***-100 -75 -50020406080100-20 0 20 50 75 100Distance to tg (Mb)%silencinge%allele freqAGPAT5 - 8p CTSB - 8p STC1 - 8pNo DOX DOXControl No DOX Control DOXXq integration sitePLS3 IL13RA1 TMEM1640.00.51.01.52.0fold change (rel. to NoDOX)DOXp=0.0039p=0.0183p<0.0001p=0.0022 p=0.0022 p=0.0022DOX     No DOXDOX     No DOXFig. 3 Allelic silencing of flanking endogenous genes upon XIST induction. a Allele-discriminating RT-PCR pyrosequencing assay for genes closestto 5 Mb of integration site for each integration site, comparing triplicate cDNAs from untreated cells (No DOX) and following 5-day DOXinduction of XIST in duplicate pyrosequencing reactions. cDNA from a different integration was also assessed (additional assays are shown inAdditional file 3). P values of significantly silenced genes are listed. b Summary of the silencing observed for individual genes for each of thenine integration sites (color-coded as shown in the legend); plotted by distance from the integration site on the chromosome (from short tolong arm). The allelic change is shown as percent silencing, which was calculated as: (allele frequency No DOX – allele frequency 5d DOX)/allele frequency No DOX × 100) for the pyrosequencing assays. For the Xq integration the silencing was determined by q-RT-PCR since thechromosome is hemizygous. Phase was determined for the 3q integration but for all integrations the allelic change is shown as silencing.Integrations on other chromosomes showed no silencing upon DOX induction. c Correlation between extent of silencing and level of XISTRNA. Five different clones (symbols) and cultures show variation in the level of XIST RNA after DOX induction (as measured by qRT-PCR forXIST relative to PGK1), and for 12q this correlates well with the extent of silencing of two genes assayed by allelic pyrosequencing after RT-PCR (OAS3,P <0.0001; POLR3B, P = 0.0004). A similar analysis (d) for the chromosome 8p integration site showed a similar variation in XIST levels, but no correlationwith extent of silencing of two loci on 8p. e Removal of XIST expression after 5-day DOX induction resulted in substantial reactivation of endogenousgenes in the 8p and Xq integration sitesKelsey et al. Genome Biology  (2015) 16:208 Page 6 of 16and Xq integration sites. All genes examined showedpartial to complete reactivation (Fig. 3e) 5 days after in-duction of XIST had ceased, indicating that the silen-cing observed requires ongoing XIST expression.RNA-seq confirms differences in silencing capacity of XISTat different integrationsThe significant differences between the integration sitesin their capacity to silence endogenous genes suggestedan important impact of the genomic context of the inte-gration; however, it was possible that by chance thegenes chosen to be tested were non-random in theirability to silence. Therefore, we chose to examine threeclones with XIST at different integration sites usingRNA-seq to generate a detailed view of any variability insilencing capacity. We chose to examine the 8p and 12qintegration sites, which had shown the most silencing,but different sensitivities to XIST levels, as well as the1p integration site, which had shown limited silencing.As our candidate genes examined by pyrosequencinghad shown significant reductions in gene expression thatranged from as little as 6 % to approximately 80 % silen-cing of one allele, we did not expect complete silencingof one allele that would reduce expression levels overallby 50 %. We chose a stringent threshold of 30–60 %total reduction to classify genes as silencing, at whichlevel a significantly greater proportion of genes were ob-served within 30 Mb of the 8p and 12q integration sitescompared to the genome (Chi-square test, P <0.0001).Consistent with our candidate gene analysis, the increasein proportion of genes in this range was not significantfor the 1p integration site. Examination of expressionchanges as an allelic change is more sensitive to the par-tial reduction of expression of one allele, and should alsostill detect changes in allelic expression when the totalexpression level of the gene is regulated in transalthough the number of genes that can be examined isreduced by the requirement for an expressed poly-morphism. We examined the allelic change in expressionon chromosomes 1, 8, and 12 (Fig. 4), and again a sig-nificantly higher proportion of genes showing an allelicchange of greater than 30 % was seen for genes flankingthe 8p integration site. A Chi-square permutation testdemonstrated that the number of genes silenced on 8pfollowing XIST induction decreased with increasing dis-tance (P = 0.008). The biological duplicate for RNA-seqwas highly concordant for the percent allelic gene silen-cing for 8p (Spearman r = 0.5806; P <0.0001 for geneswith FPKM ≥5). The proportion of genes showing allelicsilencing for genes flanking the 1p integration site wassignificant (P = 0.0269); however no significant changewas seen for the 12q integration site. We show only thedistal end of chromosome 1 as the HT1080 cells carry atranslocation of one chromosome 1 to chromosome 4(approximately 55 Mb distal to the integration site).We validated several of the observed changes by py-rosequencing or q-PCR, and the assays are includedin Additional file 3, and highlighted on Fig. 4. Percentallelic silencing was highly concordant between pyro-sequencing and RNA-seq (r = 0.9341; P <0.0001). Ofthe 17 genes showing at least 50 % allelic silencing in both8p RNA-seq replicates, two genes (DLC1 and STC1) didnot show a decrease in total expression, suggesting thatauto-regulation could be another source of discrepancybetween total read and allelic read changes. The high rateof validation of the observed changes by pyrosequencingof biological triplicates substantiated that silencing spreadto endogenous genes. Intriguingly, on chromosome 8there appeared to be domains of silencing separated byareas that were more resistant to the action of XIST. Inorder to explore what features might lead to differentialsusceptibility to silencing we performed chromatin immu-noprecipitation followed by sequencing (ChIP-seq) for in-duced and uninduced cells with XIST integrated at 8p.Chromatin features of regions silenced by XIST insomatic cellsChIP-seq for both the archetypical facultative heterochro-matic mark H3K27me3 and the active mark H3K27acshowed significant, but opposing changes on chromosome8 upon DOX induction of XIST from the 8p integrationsite, and these changes were even more dramatic onthe short arm where XIST is integrated (Fig. 5a, b).Surprisingly, DOX treatment significantly increasedH3K27ac across the genome, with most chromosomesshowing an increase, suggesting a widespread impactof the antibiotic (Fig. 5b). This unexpected change inchromatin accessibility did not extend to any significantchr1 chr8 chr12 chr1 chr8 chr12rep1 rep2chr1 chr8 chr12Fig. 4 Silencing of endogenous genes upon ectopic XIST expression.Allelic inactivation of chromosomes 1, 8, and 12 is shown as a heatmap following DOX induction of XIST in cells containing integrationsinto the chromosome listed in green. Colors denote the allelicexpression change as assessed by RNA-seq (red = downregulation;yellow= no change) for those genes with allelic reads with FPKM ≥5.The integration site is marked as a black arrow; confirmatory pyrose-quencing assays are indicated by blue lines, and the centromere isshown as a black line. For chromosome 1, the chromosome is truncatedat the site of a translocation as it is unknown whether the integration ison the translocated chromosome (boxed region)Kelsey et al. Genome Biology  (2015) 16:208 Page 7 of 16H3K27me3-0.10.00.10.20.30.4chr88p8qchange in levels upon DOX(DOX - NoDOX)0 50 1000.000.020.040.060.08gene density0 50 1000%10%20%30%40%50%0%10%20%30%40%50%LINE Aluadefgh1 Mb binscnormalized ChIP-seq level-10 -5 +10+5kb kbrelative distance across geneTSS TESchr8 p-arm0 50 100-1.0-0.50.00.51.0RNA-seq:  %total silencing0 50 100025050075050000.00.20.40.60.81.0HiC (relative to AGPAT5)RNA-seq:  %allelic changebH3K27ac-2-101chr88p8qchange in levels upon DOX(DOX - NoDOX)50 100-200020040060080010001200normalized levels of histone markH3K27me3 change upon DOXH3K27ac change upon DOXH3K27me3 no DOXH3K27ac no DOXFig. 5 (See legend on next page.)Kelsey et al. Genome Biology  (2015) 16:208 Page 8 of 16change in H3K27me3 across the genome, and the antici-pated enrichments in the imprinted regions of KCNQ1and IGF2R were observed before and after DOX (Add-itional file 4). When examined across all genes onchromosome 8p, the loss of acetylation was most notableat the promoter, the site of most pre-existing acetylation,but loss was seen throughout the upstream and gene bod-ies (Fig. 5c). The genes from 8q showed changes similar tothat of other autosomes, with an increase only detected atthe promoter (Additional file 5). The gain of H3K27me3on chromosome 8p was observed across both genic andintergenic regions (Fig. 5c), with no change observed forgenes on 8q or autosomes (Additional file 5). In order tocorrelate the change in H3K27me3 and H3K27ac with thesilencing observed, as well as other features of thechromosome, we plotted the initial levels (without DOX)as well as the total change in both marks along thechromosome (Fig. 5d). While the loss of H3K27ac couldbe anticipated to occur from locations where there wasacetylation initially, the recruitment of H3K27me3 uponXIST induction also mirrored the pre-existing levels re-markably well (Spearman r = 0.8671; P <0.0001 for 8p). InFig. 5e we show the average silencing determined fromtotal reads, while in Fig. 5f we show allelic gene silencing.We assessed allelic changes in H3K27ac as well, and thelimited informative genes showed a significant correlationbetween allelic silencing and loss of acetylation (correlationfor 29 genes with data for both, r = 0.5904; P = 0.0007).Mouse Xist localizes to sequences that are in contact withthe integration site as determined by chromatin conform-ation capture [43, 44], and therefore we extracted the Hi-Ccontacts anchored at the 1 Mb domain containing the 8pintegration site from the published Hi-C data [45] (Fig. 5f).There are more contacts, as well as stronger silencing,closer to the integration site, confounding the ability toexamine correlations. Interestingly, the Hi-C contacts cor-related with the pre-existing (No DOX) H3K27me3 levelsalong the chromosome 8 short arm (Spearman r = 0.4996;P = 0.0006), although we did not identify an association ofallelic silencing with domains designated as closed noropen [46] (Fisher’s exact test). Allelic silencing corre-lated with both the gain of H3K27me3 (Spearman r =0.4599; P = 0.0003) and loss of acetylation (Spearman r =−0.4557, P = 0.0004). Given the proposed role for repeti-tive elements in the XCI process [47–49], and the ten-dency for mouse Xist to first localize to gene-rich regionsduring early expression in embryonic stem cells [43, 44]we show the LINE and ALU distribution (Fig. 5f) and thegene density (Fig. 5g) along chromosome 8; however, nei-ther feature showed a significant correlation with silen-cing, although as would be expected, gene density is amajor contributor to total levels of H3K27 acetylation. Tofurther explore the relationship of the ability to modifychromatin and the silencing ability of the differentchromosomal integration sites we performed immuno-fluorescence in conjunction with FISH for XIST.Recruitment of heterochromatic modifications to the siteof XIST localizationIn addition to H3K27me3, the Xi is known to be enrichedfor additional histone marks and proteins associated withrepressed chromatin; however, the hierarchy of recruit-ment of these heterochromatic features by XIST has notbeen clarified. Therefore, we examined the co-localizationof these features with XIST expressed from various inte-gration sites (Table 2). The extent of co-localization canvary with the cell cycle, and therefore we categorize theco-localization as positive (consistently observed in greaterthan 25 % of XIST-positive cells), negative (consistentlyobserved in less than 10 % of cells) or +/− (in between10 % and 25 % co-localization, or variable between repli-cates). No recruitment of H3K9me3 was observed at thefour sites examined, while H2AK119u1 was observed atall four examined integration sites. The integrationonto Xq showed recruitment of all marks (with the ex-ception of H3K9me3) and is shown in Fig. 6. Recruit-ment was still less than was seen for a female cell line – inIMR90 we routinely detect over 80 % co-localization.H3K27me3 was not recruited to the 1p or 3q integrationsites, while SMCHD1 was not enriched at 1p, 3q, and also(See figure on previous page.)Fig. 5 Correlation of genomic neighborhood and XIST-induced silencing on chromosome 8. a ChIP-seq changes observed for H3K27me3 on eachchromosome, with chromosome 8 also subdivided into 8p (arm with integration) and 8q. H3K27me3 showed significant (P = 3.3e−243, pairedt-test) changes for chromosome 8 with the genome showing no significant change. b ChIP-seq changes observed for H3K27ac on each chromosome,with chromosome 8 also subdivided into 8p (arm with integration) and 8q. The chromosome 8 decrease (P = 2.2e−40) as well as the genome-wideincrease (P = 2.0e−205) were both highly significant. c Average H3K27ac and H3K27me3 normalized across genes on 8p. Normalized ChIP-seq level areshown across genes and for the 10 kb upstream and downstream of genes before XIST expression (NoDOX) and after XIST expression (DOX) as well asfor input, color-coded as outlined. d Total H3K27me3 (blue line) and H3K27ac (pink line) in No DOX overlaid with change in H3K27me3 (blue),H3K27ac (pink) along chromosome 8 after 5 days of DOX induction of XIST. e Change in total reads for genes with average ≥5 FPKM shown as percentsilencing. f Density of nuclear contacts with the 8p integration site (AGPAT5) as identified in IMR90 fibroblast cells by HiC [45] in 1 Mbbins (grey shading). Allelic silencing of genes is shown superimposed as green dots. g Density of LINE (blue line) and ALU (green line)repetitive elements per 1 Mb bin. Shown on the axes are the average genome (purple) and X-chromosomal (orange) densities of theelements. h Gene density in 1 Mb bins along the chromosomeKelsey et al. Genome Biology  (2015) 16:208 Page 9 of 164q integration site clones. The integration clone at 3q didhowever recruit macroH2A, which was not seen to beenriched at the 1p or 7q integration clones. Overall the in-tegration at 1p showed the least enrichment of marks withonly variable recruitment of H4K20me1 in addition toH2AK119u1. To further examine dependence on integra-tion site, independent clones integrated into 7q, Xq, and8p were compared by IF-FISH for H3K27me3 and/orH4K20me1. Of the six side-by-side comparisons of inde-pendent clones at the same integration site, enrichment ofmarks were very comparable showing an average of 6 %difference in enrichment and all falling within the samecategory (except for 8p with H4K20me1 which spannedthe +/− and +).The only mark examined that was not detected at anysite was the H3K9me3 mark, which is often associatedwith constitutive heterochromatin. H2AK119u1 wasseen at all of the four integrations examined, suggestingit is less dependent on local chromatin structure thanthe other marks which were heavily dependent uponTable 2 Co-localization of histone modifications with induced XIST RNA signalClone H3K27me3 H4K20me1 macroH2A SMCHD1 H2AK119u1 H3K9me31p − +/− −a − + −3q − +a −a4q +/− − +/− −a7p +/− +/− + +/−a7q +/−b −b −a +/−8p +b +/−b + a +a + −12q +/−b −a + +/− + −15q +/− +/− +/− +/−Xq + +b +a + + −aIn situ hybridization performed at least in duplicatebIn situ hybridization performed on two independent clonesCo-localization of the histone modification with the XIST signal was counted in at least 30 cells. No co-localization (−) was <10 % co-localization; +/− was between10 % and 25 % co-localization and + is assigned to integration sites with over 25 % co-localizationH3K27me3 XIST Merge macroH2A XIST Merge SMCHD1 XIST Merge uH2A Merge H3K9me3 XIST Merge H4K20me1 XIST Merge XIST Fig. 6 Immunofluorescence shows features enriched at site of XIST RNA. The XIST RNA is identified by RNA FISH (green, except for H2AK119u1(uH2A) photos, shown in greyscale) after 5 days DOX induction for the Xq integration clone. Co-IF with the antibodies listed was performed(shown in grayscale), with the merged image in color. Bars to the right show the proportion of XIST-positive HT1080 cells (integration into Xq)displaying enrichment of the indicated chromatin modification/proteinKelsey et al. Genome Biology  (2015) 16:208 Page 10 of 16the integration site and surprisingly independent ofeach other. The X chromosome showed the best abilityto recruit all features examined, with the next best inte-gration site also being the one that showed the greatestgene silencing (8p).DiscussionWe have examined the impact of genomic location ofXIST on its ability to alter nuclear ultrastructure,chromatin state, and gene expression in a somatic cellline. We observe considerable heterogeneity in the ex-tent of silencing and the recruitment of chromatinmarks, depending upon the integration site, providingus with an opportunity to dissect the interactions be-tween features in a system in which complete recruit-ment and silencing does not occur. The ability ofXIST to localize to all integration sites supports thatthere is limited sequence specificity for the RNA withthe chromosome, in agreement with the spread of in-activation reported in X/autosome translocations, andvarious mouse Xist transgenes (reviewed in [36, 50]).The lack of sequence specificity is also in line withthe recently described proximity-transfer model whichsuggests that Xist first associates with sequences inphysical proximity to the Xist locus and then transfersto gene-rich regions which are topologically associatedwith the integration site [43].Spread of silencing to autosomes in these HT1080cells is not seen to the extent that is observed for X/autosome translocations [48, 49] or an autosomal XISTtransgene in iPS cells [39], suggesting that bypassingdifferentiation reduces the ability to inactivate a chromo-some. While early mouse studies suggested the presenceof a limited developmental window during which silen-cing could be induced [12] we see XIST-dependentsilencing of endogenous genes in these HT1080 fibro-sarcoma cells up to almost 50 Mb from the integrationsite, although endogenous gene silencing is most prom-inent closer to the integration site. It is possible thatthe cancerous origin of these cells has reactivated crit-ical developmental gatekeepers, as specific mouse can-cers have been shown to allow Xist-induced silencing[17, 25]. The ability of XIST to induce gene silencing insomatic cells has important implications for cancercells where rearrangements or reactivation of XISTmay bring previously active genes under the influenceof XIST.We observe a Cot-1 hole and increased perinucleolarassociation at all integration sites, consistent with previ-ous reports that XIST/Xist expression from autosomesincreases perinucleolar association [10, 21]. In our as-sessment of multiple XIST integrations, we observedheterogeneity in the extent of perinucleolar association,and the correlation with silencing was limited. The threeintegration sites for which the increase in perinucleolarassociation was not significant (15q, 3q, and Xq) were allG-dark integration sites; however, the 7p integration wasalso in a G-dark band yet demonstrated a significantincrease in perinucleolar localization, suggesting an in-complete association between perinucleolar associationand G-dark integrations with lower XIST levels (Fig. 7).The other feature that we observed at all integrationsexamined was H2AK119u1, which is established by thePRC1 complex. We observed H2AK119u1 in the ab-sence of SMCHD1, macroH2A, and H3K27me3 in the1p integration cells, suggesting that XIST may be able todirectly recruit the PRC1 complex, in agreement withthe PRC2-independent PRC1 recruitment previouslysuggested in mice [27]. As the 1p integration site demon-strated limited silencing in the presence of H2AK119u1recruitment, we conclude that H2AK119u1 is insufficientfor the spread of gene silencing. No integration site exam-ined showed recruitment of H3K9me3, although the anti-body clearly hybridized to the native Xi in female cells.Therefore it appears that facultative heterochromaticmarks can be recruited by XIST in these cells, but theestablishment of a ‘locked-in’ silent state required add-itional layers of developmentally-regulated chromatin con-densation. Without such locks on silencing, the continuedsilencing of both reporter genes and endogenous geneswas XIST-dependent, undergoing reactivation upon re-moval of XIST induction.Interestingly, no single factor was seen to associatewith more robust silencing (Fig. 7), demonstrating con-siderable redundancy in XIST-inducible silencing path-ways. Consistent with previous studies in mice, silencingof endogenous genes in these human cells was observedin the absence of recruitment of macroH2A (7p) [30],SMCHD1 (3q) [51], or H3K27me3 (12q) [27]. MacroH2Ahas been reported to be recruited by induction of mouseXist expression [13], and is also lost upon deletion of Xist[14]; yet intriguingly we observe macroH2A recruitmentto be site-dependent, rather than solely XIST-dependent.We observed that macroH2A is not required for XIST-induced silencing, which is consistent with the ability ofmice that are knocked-out for both macroH2A1 andmacroH2A2 to continue to undergo XCI [30]. The re-cruitment of SMCHD1 and macroH2A seems to be inde-pendent of each other as, for example, the 3q integrationsite recruits macroH2A but not SMCHD1, while the 7qintegration shows some SMCHD1 recruitment, but nomacroH2A recruitment. There was a positive correlationbetween the amount of SMCHD1 recruitment andthe final perinucleolar localization (Spearman r = 0.8;P = 0.01), with the proportion of cells positive forSMCHD1 always equal or lower to the percent seento be perinucleolar. Thus perinucleolar localizationmay be required but not sufficient to recruit SMCHD1.Kelsey et al. Genome Biology  (2015) 16:208 Page 11 of 16The strong ability of the X integration to recruit allfeatures confounded the ability to detect correlations;however, it appears that the majority of the features are in-dependently recruited to the chromosome that expressesXIST. For several of the integration sites two or more in-dependent clones with similar XIST levels behaved simi-larly supporting that the variation we observe is due to theintegration site. In addition, the 4q and 12q clones werederived from a different subclone of HT1080 from theother autosomal integration sites; however, they did notappear to be more similar to each other, arguing againstthe variation arising during subclone generation.The differential recruitment of chromatin marksdepending on integration site suggests that multiple si-lencing pathways not only work in parallel to promotesilencing, but that their recruitment is favored by differentunderlying DNA sequences, consistent with previousstudies using genes that escape from inactivation to iden-tify 12 features of the DNA sequence of the X chromo-some that may be involved in the spread or maintenanceof XCI [52].ConclusionsOverall, we saw variability in the recruitment of chroma-tin marks between the integration sites, highlighting theimportance of the integration site in modulating XISTfunction. Localization of XIST with a concomitant deple-tion of Cot-1 RNA, recruitment of H2AK119u1, and a shiftto perinucleolar location was seen at all integrations, andthus reflect features established by XIST independent ofthe local chromatin environment. In contrast, recruitmentof SMCHD1, macroH2A, H3K27me3, and H4K20me1appeared to be strongly influenced by the site of XISTexpression. As some silencing was observed at allintegrations, this study demonstrates that silencing inhuman somatic cells can occur in the absence ofmacroH2A, SMCHD1, and H3K27me3/H4K20me1 re-cruitment, underscoring the independent but coopera-tive nature of the X-chromosome inactivation process.The X chromosome demonstrated the most consistentability to recruit the heterochromatic marks of XCI,consistent with an evolutionary accumulation of DNAfeatures enabling the recruitment of heterochromaticmarks to the X chromosome.MethodsGeneration and culture of cell lines and identification ofthe transgene integration siteHT1080 HH1 cells were transfected with pcDNA6/TR,and two subclones (HT1080HH1-2-3 or 2–12) expressingFig. 7 Schematic of features examined at the site of XIST RNA induction. Nine different integration sites of XIST were examined, and these werein both G-light (pale gray) and G-dark genomic locations. Upon DOX induction XIST was expressed (intensity of green oval reflects averageamount of XIST expression) and increased perinucleolar localization was observed (blue oval intensity reflects increase, with significant changesencircled in black). H2AK119u1 was enriched at all four integration sites examined. The enrichment of chromatin marks or proteins that werevariably recruited (see Table 2) is shown as solid (enrichment >25 %), dotted (enrichment between 10 % and 25 %) or unfilled (enrichment≤10 %). The integration sites are ordered by ranking of gene silencing observed by pyrosequencing (fill of red rectangle reflecting proportionof silenced genes, see Additional file 3)Kelsey et al. Genome Biology  (2015) 16:208 Page 12 of 16high RNA levels of the Tet-repressor were subsequentlytransfected with pFRT/lacZeo (Life Technologies) at lowconcentrations. The full-length inducible XIST cDNA con-struct [20] was co-transfected with the pOG44 plasmidexpressing Flp recombinase for site-specific recombin-ation into the FRT site followed by Hygromycin selectionand confirmation of loss of Zeomycin resistance. Cellswere grown at 37 °C with 5 % CO2 in DMEM supple-mented with penicillin/streptomycin, non-essential aminoacids, and 10 % V/V fetal bovine serum. XIST expressionwas induced with the addition of 1 μg/mL doxycycline tothe culture medium.Southern blotting identified eight FRT integrationsthat appeared single copy. Inverse PCR utilizing primerscomplementary to a sequence within the integratedpFRT plasmid was used to identify the precise integra-tion site of the transgenes in the HT1080 cell lines. Theends of linearized plasmids are subject to exonucleaseactivity, thus the actual integrated transgene often lacksseveral hundred of base pairs on each end which wasfirst identified by a series of PCR assays prior to restric-tion endonuclease digestion with a frequently-cuttingrestriction endonucleases identified to cut in theremaining plasmid, followed by ligation with T4 DNAligase (Invitrogen) to create circular DNA molecules.The captured genomic DNA was amplified by nestedPCR with primers facing outward from the plasmidfragment for sequencing and the genomic location wasidentified using the BLAT algorithm [53]. No X-linkedintegration was identified, so the F55 HT1080 clonefrom Yan and Boyd was used [41].While cancer derived, the HT1080 cells remain dip-loid with four structural rearrangements detected byspectral karyotyping (46,XY,del (1)(p21), i(3)(p10),i(3)(q10), der(4)t(1;4)(p21;p16), der(5)t(5;5)(p15;?),der(11)t(3;11)(q11;q25). Using our allelic pyrosequenc-ing assays we observed instability of chromosome 3 in twoof nine clones (see also [40]) and homozygosity for assayson 4q in one clone, while assays on chromosomes 1, 7, 8,and 15 remained diploid in the clones tested, suggestingthat the unbalanced rearranged chromosomes were theleast stable in these cells. We analyzed two independentclones at the Xq, 7q, and 8p integration site and five inde-pendent clones of the 8p integration site. These cloneswere individual single-cell clones following Flp-mediatedrecombination into the FRT site. Each clone was selectedfor Hyg-resistance and assessed for loss of Zeomycinsensitivity.As fluorescent reporters allow for efficient screening,we created a plasmid that carries both the inducible re-peat A and a DsRed-Express2, driven by the mouse Pgk1promoter (Fig. 2b). To test whether the ability of repeatA to silence the reporter depends on the genomic inte-gration site, we inserted the repeat A – DsRed-Express2transgene into six of the HT1080 cell lines with a knownchromosomal location of the FRT integration site for as-sessment of silencing by flow cytometry as previouslyperformed [40].RNA FISH and immunofluorescenceCells were grown on glass coverslips. Upon removalfrom cell culture the coverslips were first rinsed in ice-cold CSK buffer (0.3 M sucrose, 100 μM NaCl, 10 μMPIPES, 3 μM MgCl2), then permeabilized with 0.5 %Triton-X 100 in CSK for 8 min on ice and then fixed in4 % paraformaldehyde for 8 min at room temperature.Coverslips were stored at 4 °C in 70 % ethanol. Just priorto RNA FISH, the coverslips were immersed in 100 %ethanol for 5 min and left to air dry. FISH was per-formed with two probes: an XIST probe and a Cot-1(Invitrogen) probe. Both probes had been directly fluo-rescently labeled using the Nick Translation Reagent Kit(Abbott Molecular, Inc.) with Spectrum red-UTP (Vysis)for Cot-1 DNA probes and Spectrum green-UTP (Vysis)for XIST probes. Approximately 150 ng of each probewas mixed together along with 20 μg salmon testesDNA then air dried in a speed vacuum, resuspended in10 μL deionized formamide, denatured at 80 °C for10 min, and then mixed with 10 μL hybridization buffer(20 mg/mL BSA, 4XSSC, 20 %). This was pipetted ontoa small square of Parafilm and the coverslip was placedon top of the probe mixture. Another piece of Parafilmwas then placed on top and the edges were sealed toprevent the drying out of the coverslip. Hybridizationtook place overnight in a humidified chamber at 37 °C.The next day the coverslips were rinsed as follows:20 min in 50 % formamide/50 % 4XSSC at 37 °C,20 min in 2X SSC at 37 °C, and 20 min in 1X SSC atroom temperature. Coverslips were then stained withDAPI and mounted onto microscope slides with Vecta-shield (Vector Laboratories). Cells were observed on aLeica inverted microscope (DMI 6000B) at 100X mag-nification and images were obtained using a Retiga4000R (Q-Imaging) camera with Openlab software (Perki-nElmer). Images were processed using Adobe PhotoshopCS4 to reduce background and correct for variation inFISH efficiency between different images. A one-wayANOVA test in GraphPad was used to determine signifi-cantly different signal sizes. Line scans were generatedusing Image J software (NIH) by drawing a line throughthe area of interest and plotting the RGB intensities acrossthe line.To determine the nuclear location, XIST signals werescored visually in Photoshop as being either ‘perinuclearonly’ (for example, adjacent to and in contact with thenuclear periphery), ‘perinucleolar only’ (that is, adjacentto and in contact with a Cot-1 negative nucleolus), ‘both’(that is, adjacent to and in contact with both the nuclearKelsey et al. Genome Biology  (2015) 16:208 Page 13 of 16periphery and a Cot-1 negative nucleolus), or ‘neither’.XIST signals scored as ‘both’ are included in the ‘peri-nuclear’/‘perinucleolar’ percentages in Table 1. For the‘5d DOX’ counts, results for each integration site are theaverage of three independent experiments performed ondifferent coverslips, by at least two independent ob-servers, with a minimum of 50 cells counted each time.The ‘No DOX’ counts were done once with a minimumof 60 cells counted per integration site.For combined RNA FISH and immunofluorescence,coverslips, which had been stored at 4 °C in 70 % etha-nol, were first rinsed in PBS then placed onto a smallamount of PBT (PBS with 1 % BSA and 0.1 % Tween20) containing 0.4 U/μL Ribolock RNase inhibitor. Cov-erslips were sealed between two layers of Parafilm andleft in the blocking buffer at room temperature for20 min, then transferred from blocking buffer to PBTcontaining 1:100 primary antibody and 0.4 U/μLRibolock, sealed between two layers of Parafilm and leftat room temperature for 4–6 h. Coverslips were thenwashed three times, for 5 min each time, at roomtemperature in PBS containing 0.1 % Tween 20, thenput onto a small amount of PBT containing 1:250 fluo-rescently labeled secondary antibody and 0.4 U/μLRibolock, sealed between two layers of Parafilm and leftat room temperature in the dark for 45 min. Coverslipswere then washed three times, for 5 min each time, atroom temperature in the dark in PBS containing 0.1 %Tween 20. Coverslips were then fixed in 4 % PFA in PBSfor 10 min at room temperature in the dark, and washedfor 5 min in PBS before continuing on to RNA FISH,making sure that the coverslips remained in the darkthroughout the RNA FISH procedure. Antibodies usedin immunofluorescence include: anti-H3K27me3 (07–449from Millipore); anti-macroH2A (07–219 from Millipore);anti-SMCHD1 (ab31865 from Abcam); anti-H2AK119u1(05–678 from Millipore); anti-H3K9me3 (07–442 fromMillipore); anti-H4K20me1 (07–440 from Millipore).RNA isolation, reverse transcription Q-PCR and allelicdiscrimination by pyrosequencingRNA was isolated from cell pellets stored at −70 °C usingTRIZOL (Invitrogen) according to the manufacturer’s in-structions and then treated with DNase1. cDNA was gen-erated in the range of 0.5–2.5 μg RNA using M-MLVreverse transcriptase for qPCR on a StepOnePlusTMReal-Time PCR System (Applied Biosystems, Darmstadt,Germany), using Maxima Hot Start Taq (Thermo Sci-entific) and EvaGreen dye (Biotium). The followingconditions were used: 95° for 5 min, followed by 40 -cycles of (95° for 15 s, 60° for 30 s, 72° for 1 min), anda melt curve stage of (95° for 15 s, 60° for 1 min, in-crease of 0.3° until 95°). The expression levels of genesof interest were normalized to the expression of ACTB orPGK1. Primer sequences are found in Additional file 6.Pyrosequencing of cDNA before and after DOX induc-tion of DNA, and of clones containing integrations on alter-nate chromosomes were examined. Each 25 μL PCR wasperformed with 1x PCR Buffer (Invitrogen), 0.2 mMdNTPs, 0.625 U Taq DNA polymerase (Invitrogen), 0.5 μMforward and reverse primers, and 50–100 ng of cDNA for(94C for 30 s, 58.3C for 30 s, 72C for 1 min) × 35 cycles,and 72 °C for 10 min for final extension. One of the for-ward and reverse primers was biotinylated for template iso-lation during pyrosequencing preparation. Universal M13primer was also used for some assays (see Additional file 6),where the primer to be biotinylated instead contained theM13 sequence at the 5′end (5′CGC CAG GGT TTT CCCAGT CAC GAC3′). Nested PCR was run for the assaysthat utilized the universal primers: the first round ofPCR was performed with the same cycling conditionsbut only for 15x cycles, with the M13-tagged primer(without the biotin) and its paired forward or reverseprimer; the second round of PCR was run with 1 μL ofPCR product from the first round of PCR as the tem-plate, as well as the biotinylated M13 primer and thenon-M13-tagged primer that was used in the firstround, under the same cycling condition but for 20 cy-cles. Pyrosequencing was performed on the PyroMarkMD machine (Qiagen). Template preparation for pyro-sequencing was done according to manufacturer’s proto-col. For each assay, 10–15 μL of PCR products was usedas template and CDT tips were used to dispense thedNTPs.Sequencing analysisFor both ChIP-seq and RNA-seq, library preparationand sequencing was done according to Illumina proto-cols, and reads were aligned to hg18 reference genomeusing Tophat. RNA-seq was performed as previouslydescribed [54], first on two 8p clones following 5 daysDOX and the 8p clone with No DOX producing 50 bppaired-end reads, with a Pearson correlation betweenreplicate DOX FPKM values of r = 0.9934 (P <0.0001). Asecond RNA-seq of the 12q and 1p integration sites withNo DOX and following 5 days DOX produced 76 bppaired-end reads. Total RNA expression was quantifiedin FPKM using Cufflinks. ChIP of H3K27me3 (ActiveMotif antibody 61017) and H3K27ac (Active Motif anti-body 39133) was performed on the 8p clone with andwithout DOX treatment, as previously done [55] using20 μg of chromatin and 3 μg of antibody per IP. ChIP-seq produced 36 bp single-end reads, and the ChIP-seqdata were analyzed using SeqMonk to identify enrichedH3K27ac peaks (MACS, P value = 1 × 10−3) and to quan-tify the level of H3K27me3 in 2 kb and 1 Mb non-overlapping windows.Kelsey et al. Genome Biology  (2015) 16:208 Page 14 of 16Allelic information for both RNA-seq and ChIP-seq wasobtained with a custom workflow in Galaxy [56–58].Briefly, mapped reads were run through Samtools mpileupto obtain SNPs relative to the reference genome in thecontrol and experimental (integration of interest) samples,and allelic ratio of reads was calculated at each variant. Inaddition, only the variants with a biallelic ratio of 0.3–0.7in the No DOX (control) sample were further examined.In order to phase the variants, the allele with lower readsfor each variant site in the DOX (experimental) samplewas considered to be silenced in cis with the XIST trans-gene. Variant sites were next combined based on regionsof interest, either gene location or genomic regions forRNA-seq and ChIP-seq, respectively. Variant sites thathad an allelic ratio 0.15 greater or less than the overallgene allelic ratio for the sample were removed and anew overall gene ratio was calculated. For ChIP-seq,variants with reads >2-fold different from the averagenumber of reads in the region of interest were alsoexcluded. Percent allelic silencing for a given gene or re-gion of interest was defined as (control allelic ratio –experimental allelic ratio)/control allelic ratio. The calcu-lated ratio was the frequency of the allele with lower readsin the experimental sample for each gene or region ofinterest. We also required that there were at least 4 readsand 5 reads with allelic information per gene (RNA-seq)and region of interest (ChIP-seq), respectively, in both thecontrol and experimental samples to be considered foranalysis.Hi-C analysisWe obtained normalized Hi-C data for human femalefibroblast line IMR90 [45] and the Hi-C analysis wasdone as previous [43]. We calculated the 1 Mb Hi-Ccounts by summing the counts in all 40 kb bins withineach 1 Mb bins across the chromosome, with the an-chor being the 1 Mb bin containing the XIST transgene(AGPAT5 for 8p clone).Data availabilityThe sequence data described are available in GSE68109.Additional microscopy images are available in Figshare(http://dx.doi.org/10.6084/m9.figshare.1529822).Additional filesAdditional file 1: Genomic location of FRT integration sites in theHT1080 cell lines. (DOCX 129 kb)Additional file 2: Location of XIST RNA signal for each integrationsite with and without DOX treatment. The numbers shown are basedon three independent experiments for the DOX results and oneexperiment for the No DOX results, with ≥50 cells counted for eachintegration site in each experiment. (PDF 57 kb)Additional file 3: Candidate gene silencing assays for eachintegration site. One assay for each integration site was included inFig. 2. a Allelic pyrosequencing of three cDNAs from No DOX andDOX were compared in duplicate pyrosequencing reactions. As acontrol cDNA from a different integration was also assessed. b Forthe X-chromosome integration site q-RT-PCR was used to determinesilencing as the cells are hemizygous. c Q-RT-PCR assays were alsoperformed to validate some autosomal silencing and compare withpyrosequencing and RNA-seq. (PDF 933 kb)Additional file 4: Imprinted regions showed broad enrichment ofH3K27me3 and punctate peaks of H3K27ac. Probes with values in theextreme 5 % of inputs were removed. (PDF 175 kb)Additional file 5: Average H3K27ac and H3K27me3 for genes on 8qand chromosome 1. The normalized ChIP-seq level are shown across anaggregate of genes and for the 10 kb upstream and downstream before(NoDOX) and after XIST expression (DOX). (PDF 246 kb)Additional file 6: Table of primers. (XLSX 16 kb)Competing interestsThe authors declare that they have no competing interests.Authors’ contributionsADK performed the RNA FISH and IF experiments and drafted the manuscript,CY performed the RNA and ChIP-seq analysis and allelic and q-RT-PCRexperiments, DCYL performed the ChIP and RNA-seq experiments, TDMDperformed additional RNA FISH experiments, JM created the DS-REDconstruct and performed flow cytometric analyses, SELB generated andcultured the cell lines, ABB established the allelic workflow, LL and CJBprovided guidance and interpretation of results and all authors contributed tothe final manuscript.AcknowledgementsThis work was supported by CIHR grants to CJB (MOP-13680) and LL(MOP-119357). The authors thank Jen Chow for developing the inducibleXIST clones, and Jordan Henriksen, Irene Qi and Allison M. Cotton forassistance and advice on FISH.Author details1Department of Medical Genetics, Molecular Epigenetics Group, Life SciencesInstitute, University of British Columbia, Vancouver, Canada. 2Ludwig Institutefor Cancer Research, University of California at San Diego School of Medicine,La Jolla, CA, USA. 3Division of Life Science, The Hong Kong University ofScience and Technology, Clear Water Bay, Hong Kong, China.Received: 17 April 2015 Accepted: 10 September 2015References1. Lyon MF. Gene action in the X-chromosome of the mouse (Mus musculusL.). Nature. 1961;190:372–3.2. Brown CJ, Lafreniere RG, Powers VE, Sebastio G, Ballabio A, Pettigrew AL, etal. Localization of the X inactivation centre on the human X chromosome inXq13. Nature. 1991;349:82–4.3. Brown CJ, Ballabio A, Rupert JL, Lafreniere RG, Grompe M, Tonlorenzi R, etal. A gene from the region of the human X inactivation centre is expressedexclusively from the inactive X chromosome. Nature. 1991;349:38–44.4. Brown CJ, Hendrich BD, Rupert JL, Lafreniere RG, Xing Y, Lawrence J, et al.The human XIST gene: analysis of a 17 kb inactive X-specific RNA thatcontains conserved repeats and is highly localized within the nucleus. Cell.1992;71:527–42.5. Clemson CM, McNeil JA, Willard HF, Lawrence JB. XIST RNA paints theinactive X chromosome at interphase: evidence for a novel RNA involved innuclear/chromosome structure. J Cell Biol. 1996;132:259–75.6. Blewitt ME, Vickaryous NK, Hemley SJ, Ashe A, Bruxner TJ, Preis JI, et al. AnN-ethyl-N-nitrosourea screen for genes involved in variegation in themouse. Proc Natl Acad Sci U S A. 2005;102:7629–34.7. Pullirsch D, Hartel R, Kishimoto H, Leeb M, Steiner G, Wutz A. The Trithoraxgroup protein Ash2l and Saf-A are recruited to the inactive X chromosomeat the onset of stable X inactivation. Development. 2010;137:935–43.Kelsey et al. Genome Biology  (2015) 16:208 Page 15 of 168. Sado T, Brockdorff N. Advances in understanding chromosome silencing bythe long non-coding RNA Xist. Phil Trans Royal Society London Series B.2013;368:20110325.9. Barr ML, Carr DH. Correlations between sex chromatin and sexchromosomes. Acta Cytol. 1962;6:34–45.10. Hall LL, Byron M, Sakai K, Carrel L, Willard HF, Lawrence JB. An ectopic humanXIST gene can induce chromosome inactivation in postdifferentiation humanHT-1080 cells. Proc Natl Acad Sci U S A. 2002;99:8677–82.11. Chow J, Heard E. X inactivation and the complexities of silencing a sexchromosome. Curr Op Cell biol. 2009;21:359–66.12. Wutz A, Jaenisch R. A shift from reversible to irreversible X inactivation istriggered during ES cell differentiation. Mol Cell. 2000;5:695–705.13. Rasmussen TP, Wutz A, Pehrson JR, Jaenisch R. Expression of Xist RNA is sufficientto initiate macrochromatin body formation. Chromosoma. 2001;110:411–20.14. Csankovszki G, Panning B, Bates B, Pehrson JR, Jaenisch R. Conditionaldeletion of Xist disrupts histone macroH2A localization but notmaintenance of X inactivation. Nat Genet. 1999;22:323–4.15. Rack KA, Chelly J, Gibbons RJ, Rider S, Benjamin D, Lafreniere RG, et al.Absence of the XIST gene from late-replicating isodicentric X chromosomesin leukemia. Hum Mol Genet. 1994;3:1053–9.16. Brown CJ, Willard HF. The human X inactivation center is not required formaintenance of X inactivation. Nature. 1994;368:154–6.17. Agrelo R, Souabni A, Novatchkova M, Haslinger C, Leeb M, Komnenovic V,et al. SATB1 defines the developmental context for gene silencing by Xist inlymphoma and embryonic cells. Dev Cell. 2009;16:507–16.18. Nechanitzky R, Davila A, Savarese F, Fietze S, Grosschedl R. Satb1 and Satb2 aredispensable for X chromosome inactivation in mice. Dev Cell. 2012;23:866–71.19. Jeon Y, Lee JT. YY1 tethers Xist RNA to the inactive X nucleation center.Cell. 2011;146:119–33.20. Chow JC, Hall LL, Baldry SE, Thorogood NP, Lawrence JB, Brown CJ.Inducible XIST-dependent X-chromosome inactivation in human somaticcells is reversible. Proc Natl Acad Sci U S A. 2007;104:10104–9.21. Zhang LF, Huynh KD, Lee JT. Perinucleolar targeting of the inactive X duringS phase: evidence for a role in the maintenance of silencing. Cell.2007;129:693–706.22. Csankovszki G, Nagy A, Jaenisch R. Synergism of Xist RNA, DNA methylation,and histone hypoacetylation in maintaining X chromosome inactivation.J Cell Biol. 2001;153:773–83.23. Bhatnagar S, Zhu X, Ou J, Lin L, Chamberlain L, Zhu LJ, et al. Genetic andpharmacological reactivation of the mammalian inactive X chromosome.Proc Natl Acad Sci U S A. 2014;111:12591–8.24. Minkovsky A, Sahakyan A, Rankin-Gee E, Bonora G, Patel S, Plath K. TheMbd1-Atf7ip-Setdb1 pathway contributes to the maintenance of Xchromosome inactivation. Epigenetics Chromatin. 2014;7:12.25. Yildirim E, Kirby JE, Brown DE, Mercier FE, Sadreyev RI, Scadden DT, et al. Xist RNAis a potent suppressor of hematologic cancer in mice. Cell. 2013;152:727–42.26. Silva J, Mak W, Zvetkova I, Appanah R, Nesterova TB, Webster Z, et al.Establishment of histone h3 methylation on the inactive X chromosomerequires transient recruitment of Eed-Enx1 polycomb group complexes. DevCell. 2003;4:481–95.27. Schoeftner S, Sengupta AK, Kubicek S, Mechtler K, Spahn L, Koseki H, et al.Recruitment of PRC1 function at the initiation of X inactivation independentof PRC2 and silencing. EMBO J. 2006;25:3110–22.28. Kalantry S, Magnuson T. The Polycomb Group Protein EED Is Dispensable forthe Initiation of Random X-Chromosome Inactivation. PLoS Genet. 2006;2, e66.29. Leeb M, Wutz A. Ring1B is crucial for the regulation of developmentalcontrol genes and PRC1 proteins but not X inactivation in embryonic cells.J Cell Biol. 2007;178:219–29.30. Tanasijevic B, Rasmussen TP. X chromosome inactivation and differentiationoccur readily in ES cells doubly-deficient for macroH2A1 and macroH2A2.PLoS One. 2011;6, e21512.31. Blewitt ME, Gendrel AV, Pang Z, Sparrow DB, Whitelaw N, Craig JM, et al.SmcHD1, containing a structural-maintenance-of-chromosomes hingedomain, has a critical role in X inactivation. Nat Genet. 2008;40:663–9.32. Mould AW, Pang Z, Pakusch M, Tonks ID, Stark M, Carrie D, et al. Smchd1regulates a subset of autosomal genes subject to monoallelic expression inaddition to being critical for X inactivation. Epigenetics Chromatin. 2013;6:19.33. Chaumeil J, Le Baccon P, Wutz A, Heard E. A novel role for Xist RNA in theformation of a repressive nuclear compartment into which genes arerecruited when silenced. Genes Dev. 2006;20:2223–37.34. Thorogood NP, Brown CJ. Active chromatin marks are retained on Xchromosomes lacking gene or repeat silencing despite XIST/Xist expressionin somatic cell hybrids. PLoS One. 2010;5, e10787.35. Clemson CM, Hall LL, Byron M, McNeil J, Lawrence JB. The X chromosome isorganized into a gene-rich outer rim and an internal core containingsilenced nongenic sequences. Proc Natl Acad Sci U S A. 2006;103:7688–93.36. Yang C, Chapman AG, Kelsey AD, Minks J, Cotton AM, Brown CJ. X-chromosomeinactivation: molecular mechanisms from the human perspective. Hum Genet.2011;130:175–85.37. Duthie SM, Nesterova TB, Formstone EJ, Keohane AM, Turner BM, ZakianSM, et al. Xist RNA exhibits a banded localization on the inactive Xchromosome and is excluded from autosomal material in cis. Hum MolGenet. 1999;8:195–204.38. Lee JT, Strauss WM, Dausman JA, Jaenisch R. A 450 kb transgene displaysproperties of the mammalian X-inactivation center. Cell. 1996;86:83–94.39. Jiang J, Jing Y, Cost GJ, Chiang JC, Kolpa HJ, Cotton AM, et al. Translatingdosage compensation to trisomy 21. Nature. 2013;500:296–300.40. Minks J, Baldry SE, Yang C, Cotton AM, Brown CJ. XIST-induced silencing offlanking genes is achieved by additive action of repeat a monomers inhuman somatic cells. Epigenetics Chromatin. 2013;6:23.41. Yan C, Boyd DD. Histone H3 acetylation and H3 K4 methylation definedistinct chromatin regions permissive for transgene expression. Mol CellBiol. 2006;26:6357–71.42. Hall LL, Carone DM, Gomez AV, Kolpa HJ, Byron M, Mehta N, et al. StableC0T-1 repeat RNA is abundant and is associated with euchromaticinterphase chromosomes. Cell. 2014;156:907–19.43. Engreitz JM, Pandya-Jones A, McDonel P, Shishkin A, Sirokman K, Surka C,et al. The Xist lncRNA exploits three-dimensional genome architecture tospread across the X chromosome. Science. 2013;341:1237973.44. Simon MD, Pinter SF, Fang R, Sarma K, Rutenberg-Schoenberg M, BowmanSK, et al. High-resolution Xist binding maps reveal two-step spreadingduring X-chromosome inactivation. Nature. 2013;504:465–9.45. Dixon JR, Selvaraj S, Yue F, Kim A, Li Y, Shen Y, et al. Topological domains inmammalian genomes identified by analysis of chromatin interactions.Nature. 2012;485:376–80.46. Nora EP, Dekker J, Heard E. Segmental folding of chromosomes: a basis forstructural and regulatory chromosomal neighborhoods? BioEssays. 2013;35:818–28.47. Lyon MF. X-chromosome inactivation: a repeat hypothesis. Cytogenet CellGenet. 1998;80:133–7.48. Cotton AM, Chen CY, Lam LL, Wasserman WW, Kobor MS, Brown CJ. Spreadof X-chromosome inactivation into autosomal sequences: role for DNAelements, chromatin features and chromosomal domains. Hum Mol Genet.2014;23:1211–23.49. Bala Tannan N, Brahmachary M, Garg P, Borel C, Alnefaie R, Watson CT, et al.DNA methylation profiling in X;autosome translocations supports a role forL1 repeats in the spread of X chromosome inactivation. Hum Mol Genet.2014;23:1224–36.50. Gendrel AV, Heard E. Noncoding RNAs and epigenetic mechanisms duringX-chromosome inactivation. Annu Rev Cell Dev Biol. 2014;30:561–80.51. Gendrel AV, Apedaile A, Coker H, Termanis A, Zvetkova I, Godwin J, et al.Smchd1-dependent and -independent pathways determine developmentaldynamics of CpG island methylation on the inactive X chromosome. DevCell. 2012;23:265–79.52. Wang Z, Willard HF, Mukherjee S, Furey TS. Evidence of influence ofgenomic DNA sequence on human X chromosome inactivation. PLoSComput Biol. 2006;2, e113.53. UCSC Genome Bioinformatics. Available at: http://genome.ucsc.edu/cgi-bin/hgBlat.54. Xie W, Schultz Matthew D, Lister R, Hou Z, Rajagopal N, Ray P, et al.Epigenomic analysis of multilineage differentiation of human embryonicstem cells. Cell. 2013;153:1134–48.55. Hawkins RD, Hon GC, Lee LK, Ngo Q, Lister R, Pelizzola M, et al. Distinctepigenomic landscapes of pluripotent and lineage-committed human cells.Cell Stem Cell. 2010;6:479–91.56. Goecks J, Nekrutenko A, Taylor J. The Galaxy Team. Galaxy: a comprehensiveapproach for supporting accessible, reproducible, and transparentcomputational research in the life sciences. Genome Biol. 2010;11:R86.57. Blankenberg D, Von Kuster G, Coraor N, Ananda G, Lazarus R, Mangan M,et al. Galaxy: a web-based genome analysis tool for experimentalists. CurrProtoc Mol Biol. 2010; Chapter 19:Unit 19.10.1-21.58. Giardine B, Riemer C, Hardison RC, Burhans R, Elnitski L, Shah P, et al. Galaxy: aplatform for interactive large-scale genome analysis. Genome Res. 2005;15:1451–5.Kelsey et al. Genome Biology  (2015) 16:208 Page 16 of 16

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.52383.1-0361961/manifest

Comment

Related Items