Abstract
The EWS-FLI1 fusion protein drives oncogenesis in the Ewing sarcoma family of tumors (ESFT) in humans, but its toxicity in normal cells requires additional cellular events for oncogenesis. We show that the lncRNA HOTAIR maintains cell viability in the presence of EWS-FLI1 and redirects epigenetic regulation in ESFT. HOTAIR is consistently overexpressed in ESFTs and is not driven by EWS-FLI1. Repression of HOTAIR in ESFT cell lines significantly reduces anchorage-independent colony formation in vitro and impairs tumor xenograft growth in vivo. Overexpression of HOTAIR in human mesenchymal stem cells (hMSCs), a putative cell of origin of ESFT, and IMR90 cells induces colony formation. Critically, HOTAIR-expressing hMSCs and IMR90 cells remain viable with subsequent EWS-FLI1 expression. HOTAIR induces histone modifications and gene repression through interaction with the epigenetic modifier LSD1 in ESFT cell lines and hTERT-hMSCs. Our findings suggest that HOTAIR maintains ESFT viability through epigenetic dysregulation.
Significance While the EWS-FLI1 fusion gene was determined to be the oncogenic driver in the overwhelming majority of ESFT, it is toxic to cell physiology and requires one or more additional molecular events to maintain cell viability. As these tumors have surprisingly few genetic mutations at diagnosis, epigenetic changes have been considered to be such an event, but the mechanism by which these changes are driven remains unclear. Our work shows that HOTAIR is consistently expressed among ESFT and induces epigenetic and gene expression changes that cooperate in tumorigenesis. Furthermore, expression of HOTAIR allows for cell viability in the setting of subsequent EWS-FLI1 expression. Our findings elucidate new steps of malignant transformation in this cancer and identify novel therapeutic targets.
INTRODUCTION
The Ewing sarcoma family of tumors (ESFT) consists of primitive cancers of the bone and soft tissues that arise in children and young adults. These tumors harbor chromosomal translocations that result in the fusion of the 5’ portion of the EWSR1 gene to the 3’ end of an ETS family member, with >85% resulting in the EWS-FLI1 fusion gene(1). The resultant oncoprotein alters gene expression and alternative splicing(2, 3) and is necessary for tumor viability in laboratory models of ESFT(4). However, exogenous expression of EWS-FLI1 is toxic to normal cells and other cancer cell types, inducing rapid apoptosis or senescence(5). This toxicity suggests that additional cellular events must allow tolerance of the oncoprotein. Three large genomic sequencing studies identified some recurrent genetic mutations in ESFT, including loss-of-function mutations in TP53, CDKN2A, and STAG2(6–8), but these mutations were found in only 25–30% of all samples. Thus, additional molecular changes, such as transcriptional or epigenetic events, must occur to allow tumor cell survival.
Long noncoding RNAs (lncRNAs) have significant roles in the regulation of gene expression, either directly or epigenetically. The lncRNA HOTAIR was specifically shown to direct epigenetic repression in trans across the genome, in part through recruitment of the LSD1/REST/CoREST complex at its 3’ end(9). This RNA-protein complex alters histone methylation at histone 3 lysine 4 (H3K4), demethylated by LSD1 from a dimethylated state (Me2) to mono- (Me) or unmethylated. This histone modification represses gene expression and maintains an embryonic state in tissues where HOTAIR is expressed. HOTAIR is abnormally overexpressed in numerous cancers (reviewed in (10)), and HOTAIR has been shown to epigenetically modify gene expression in these cancers (11–13).
In this study, we evaluated the function of HOTAIR in ESFT. We confirmed that HOTAIR is overexpressed in ESFT cell lines and primary tumors, as compared to normal tissues and to human mesenchymal stem cells (hMSCs), a putative cell of origin of these tumors(14, 15). We demonstrated that HOTAIR is necessary for the formation and viability of ESFT cell line-derived anchorage-independent colonies. We also showed that repression of HOTAIR by shRNA reduced tumor xenograft formation from ESFT cell lines in immunodeficient mice. In contrast, overexpression of HOTAIR in primary and hTERT-immortalized hMSCs induces anchorage-independent colony formation. HOTAIR expression in hMSCs and IMR90 fibroblasts also allows for subsequent viable expression of EWS-FLI1. We verified that HOTAIR associates with LSD1 in ESFT cell lines, and interaction with LSD1 is necessary for colony formation in hMSCs. We modulated HOTAIR expression in ESFT cell lines and hMSCs, alone and with concomitant EWS-FLI1 expression. This change in HOTAIR expression was associated with significant gene expression changes across the transcriptome, including a set of genes modulated across all models, as determined by next-generation RNA-Sequencing (RNA-Seq). We further demonstrated by Chromatin Immunoprecipitation and Sequencing (ChIP-Seq) that HOTAIR expression induced H3K4 demethylation across the genome including in a significant number of genes with corresponding repression of expression.
RESULTS
HOTAIR is overexpressed, independently of EWS-FLI1, in ESFT cell lines and primary tumors as compared to normal tissues and mesenchymal stem cells
We first assessed the expression of HOTAIR in ESFT as compared to normal tissues, utilizing the OncoGenomics DB of the National Cancer Institute (https://pob.abcc.ncifcrf.gov/cgi-bin/JK). We examined next-generation RNA-sequencing (RNA-Seq) data annotated there for ESFT, including 50 cell lines and 72 primary tumor RNA samples(6), with expression normalized to that of a set of samples of normal adult tissues (Figure 1A). HOTAIR was expressed in all samples tested, with over a log-fold higher expression in >90% of cell lines and tumors as compared to normal tissues.
We next examined expression in ESFT as compared to other cancer types. Using the R2 Genomics Analysis and Visualization platform, (http://R2.amc.nl), we compared HOTAIR expression among cancer gene expression datasets that were analyzed using Affymetrix u133 microarrays (Figure 1B). The three datasets with the highest average HOTAIR expression were comprised of ESFT samples (16–18), as compared to 311 other datasets including other cancer cell lines, non-ESFT primary tumor sets, and sets of mixed normal and cancerous samples.
We validated these findings directly by analysis of 13 ESFT cell lines, 22 primary tumor RNA samples, and 2 primary hMSC samples by RT-qPCR. Using a threshold of two-fold expression as compared to hMSC, 11/13 cell lines and 17/22 tumor samples had high HOTAIR expression (Figure 1C). This consistent overexpression of HOTAIR in ESFT supports a functional role for the lncRNA in this cancer.
We hypothesized that HOTAIR expression in ESFT is an EWS-FLI1-independent event. We confirmed this by knocking down EWS-FLI1 expression in 4 ESFT cell lines by siRNA (Figure 1D). In all four cell lines, knockdown of EWS-FLI1 did not result in loss of expression of HOTAIR, and in two lines this knockdown actually led to a modest upregulation of HOTAIR expression. Thus, HOTAIR expression in ESFT cell lines is not driven by EWS-FLI1 and represents an independent biological pathway.
HOTAIR expression correlates with anchorage independent colony formation in ESFT cell lines and hMSCs while maintaining MSC characteristics
We examined the phenotypic effects of HOTAIR in ESFT cell lines by first knocking down expression using shmiRNA in the ES2, TC32, and SK-ES cell lines. We repressed expression to 20–50% of baseline HOTAIR expression (Figure 2A), with some variation among the cell lines. We were unable to maintain viable cells with expression below this level for each cell line, supporting a role for HOTAIR in maintaining cell viability. We also confirmed that loss of HOTAIR expression had no significant effect on EWS-FLI1 protein expression (SI 1A).
We next examined the effects of HOTAIR repression on the growth of ESFT cells. Proliferation of these cells in two-dimensional tissue culture did not show any significant difference at this level of repression (SI 1B). However, prior work in Ewing sarcoma biology demonstrated that inhibition of LSD1, shown in other cancers to interact with HOTAIR, had no significant effect on proliferation but did alter 3D tumorsphere growth(19). Accordingly, we evaluated growth of tumorspheres in soft agar. We found in all three cell line models that repression of HOTAIR resulted in a significant decrease in anchorage-independent colony formation in soft agar (Figure 2B).
hMSCs are the presumptive cell of origin for Ewing sarcoma but they poorly form colonies in 3D culture in vitro and lack tumorigenicity in vivo(20, 21). We evaluated how exogenous HOTAIR expression affected the growth of these cells. We initially used hTERT-immortalized hMSCs for this assay for ease of use in in vitro culture. These cells phenocopy primary hMSCs except for avoiding senescence, and they poorly form colonies in 3D culture. We overexpressed HOTAIR or a control vector in these cells, at levels comparable to ESFT cell lines (Figure 2C). HOTAIR induced colony formation in soft agar, whereas expression of the empty vector did not (Figure 2D). This ability to enable anchorage-independent growth supports a role for HOTAIR not only in cell proliferation and viability but also in tumorigenesis. We repeated this experiment with early passage primary mesenchymal stem cells and found that overexpression of HOTAIR induced colony formation in soft agar (Figure 2C and D). We also used the wildtype IMR90 lung fibroblast cell line, which has been shown to be useful as a model of EWS-FLI1 expression(22) and does not form colonies in 3D culture. Again, overexpression of HOTAIR induced anchorage-independent colony formation in these cells (Figure 2C and D).
MSCs are defined by characteristics that include the ability to grow on plastic, expression of the surface markers CD73, CD90, and CD105, and maintenance of differentiation capacity along mesenchymal lineages(23). ESFT also have most of these properties, distinct from other sarcomas(24, 25). We confirmed that HOTAIR-expressing hTERT-hMSCs maintained these properties, as did the ESFT cell lines with knockdown of HOTAIR expression. Analysis by flow cytometry showed that control cells and cells with modulated HOTAIR expression had strong surface expression of CD73, CD90, and CD105, and had no significant alteration of other lineage markers (SI 2). The cells with modulated HOTAIR expression were also able to be differentiated into osteoblasts or adipocytes in patterns similar to their GFP-expressing controls (SI 3).
HOTAIR interacts with LSD1 in ESFT and requires its 3’ interaction domain for tumorsphere formation capacity in hMSCs
We performed formaldehyde-crosslinked RNA immunoprecipitation and confirmed that, in ESFT cell lines, HOTAIR interacts with LSD1 (SI 4A). We wanted to evaluate if this interaction is necessary for the anchorage-independent colony formation seen in the hMSC models. The interaction domain of HOTAIR with LSD1 has been previously defined(9). We generated a HOTAIR deletion construct that lacked the 3' LSD1 interaction domain (Figure 3A). We then expressed this construct or wild-type HOTAIR in hTERT-hMSCs. We confirmed that expression of each of the constructs was not significantly different amongst the pools of cells (SI 4B), then repeated the anchorage-dependent colony formation assay. As compared to the cells with overexpression of wild-type HOTAIR, those cells with expression of the mutant HOTAIR had a markedly diminished or absent colony formation capacity in soft agar (Figure 3B). This loss of function suggests that HOTAIR must interact with LSD1 to support tumor formation in hMSCs and, analogously, in ESFT.
HOTAIR primes hMSCs for tolerance of EWS-FLI1 and maintains cell viability
As previously noted, exogenous expression of EWS-FLI1 in most normal and malignant cell lines rapidly causes cell death. We confirmed this phenotype by expression of either GFP alone or EWS-FLI1 and GFP in hTERT-hMSCs. Within 48 hours, the EWS-FLI1-overexpressing hTERT-hMSCs morphologically changed with cell shrinkage and nuclear collapse, and no viable cells could be seen at 72 hours (Figure 3C), whereas the GFP-expressing cells remained unchanged. We created a vector for simultaneous co-expression of HOTAIR and EWS-FLI1 and transfected hTERT-hMSCs with this vector. The hTERT-hMSCs again underwent rapid apoptosis. We next used the hTERT-hMSCs that stably expressed HOTAIR and overexpressed EWS-FLI1 in those cells. In marked contrast, these cells stably expressed both EWS-FLI1 and HOTAIR, as confirmed by western blot and RT-qPCR respectively (SI 5A and B), and remained viable without morphologic evidence of differentiation (Figure 3C). These cells also formed tumorspheres in soft agar, at an increased rate as compared to hTERT-hMSC-HOTAIR cells. We were similarly able to stably express EWS-FLI1 in primary hMSCs and IMR90 cells with stable HOTAIR expression (SI 5A-C). These data support the hypothesis that HOTAIR expression in the ESFT precursor cell primes the cell to tolerate the subsequent chromosomal translocation that results in EWS-FLI1, driving malignant transformation to ESFT.
HOTAIR is necessary but not sufficient for tumor xenograft growth in immunodeficient mice
We next evaluated how HOTAIR affects the tumor-initiating capacity of ESFT cells in vivo. We implanted the ESFT cell lines above, with shRNA-mediated knockdown of HOTAIR or nonsilencing control, into the flanks of SCID mice. The tumor xenografts of cells with repressed HOTAIR expression had significantly slower tumor growth across all three cell lines as compared to control (p<0.001 for all three cell lines, Figure 4). We also implanted into the SCID mice the primary hMSC, hTERT-hMSC, and IMR90-GFP, -HOTAIR, and -HOTAIR-EWS-FLI1 cells, at cell numbers up to 1×107cells/implant, but no tumors grew in any of the mice after 12 weeks. These data suggest that HOTAIR is necessary but not sufficient for tumor growth in cells in the context of concomitant EWS-FLI1 expression.
HOTAIR expression in hMSCs and ESFT cells is associated with gene expression changes affecting cell adhesion, extracellular matrix, and embryonic stemness
Given the ubiquitous expression of HOTAIR in ESFT cell lines and tumors and its tumorigenic effect in hMSCs, we aimed to characterize the effects of HOTAIR on gene expression in these models. We used the hTERT-hMSC models as described above, with expression of HOTAIR alone or with subsequent expression of EWS-FLI1; these models expressed HOTAIR 5–8 fold over baseline (SI 5A), in a range observed in ESFT tumors. In ESFT cell lines, we knocked down HOTAIR expression by use of GapmeR antisense oligonucleotides. Treatment of cells with GapmeRs for 48 hours repressed expression to <30% of baseline in 3 ESFT cell lines, ES2, A673, and SK-ES (SI 6A).
We examined mRNA expression in these models by RNA-Seq, identifying transcripts with significantly different expression as compared to the control (adjusted p-value<0.05). In the hTERT-hMSC cells, we identified 2781 transcripts upregulated and 2169 transcripts downregulated in the HOTAIR-overexpressing cells, and 3888 transcripts upregulated and 2435 transcripts downregulated in the HOTAIR-EWS-FLI1 overexpressing cells, as compared to the GFP control (SI 6B). We identified the top 100 differentially-expressed transcripts between the HOTAIR-EWS-FLI1 expressing cells and GFP control and examined their expression in all three cell types (Figure 5A, SI data). Among these transcripts, we found a set of genes that are downregulated in cells with increased HOTAIR expression independently of EWS-FLI1. We also defined a set of transcripts whose expression is upregulated in cells with increased HOTAIR expression independently of EWS-FLI1, as well as a larger set of transcripts specifically upregulated by EWS-FLI1. These patterns are similarly seen in the complete dataset of all differentially expressed genes (SI Data).
In the ESFT cells, a different pattern emerged. Treatment of ES2, A673, and SK-ES cells with the GapmeR repressing HOTAIR caused more genes to be upregulated than downregulated as compared to control (972 upregulated vs. 472 downregulated for ES2, 717 upregulated vs. 506 downregulated for A673, 1080 upregulated vs 815 downregulated for SK-ES, SI 3C). Across all three cell lines there was an intersection set of 81 differentially expressed genes; of these, 49 were upregulated when cells were treated with GapmeRs as compared to control (Figure 5B, SI data). This suggests that HOTAIR expression plays a greater role in directing gene repression in ESFT cells than in driving gene expression.
We evaluated which biological pathways were affected by HOTAIR expression in these cell line model sets. We noted significant heterogeneity of the specific genes affected within each cell line, both in terms of basal expression and change in expression relative to HOTAIR. As such, we identified those differentially-expressed genes in the hTERT-hMSC cells and any of the ESFT cell lines and defined common biological pathways affected by those genes (Table 1, SI Data). HOTAIR expression correlated with repression of genes involved in cell differentiation, extracellular matrix organization, and cell-cell adhesion, and with upregulation of genes affecting angiogenesis, cell motility, and biological adhesion. These functions correlate with tumorigenesis, metastasis, and inhibition of differentiation.
HOTAIR expression induces H3K4 demethylation at the promoters of HOTAIR-repressed genes
As previously mentioned, RNA immunoprecipitation experiments in ESFT cell lines confirmed that HOTAIR associates with LSD1 (SI 4A). LSD1 has been shown in other cancers to repress gene expression through this association with HOTAIR by demethylation of histone 3 at lysine 4 (H3K4), particularly from a dimethylated (Me2) to a monomethylated (Me) state(9). We hypothesized that differential HOTAIR expression in our model systems would alter LSD1-directed histone methylation at gene promoters, repressing gene expression. As such, we performed Chromatin Immunoprecipitation-Sequencing (ChIP-Seq) for H3K4Me1 and H3K4Me2, using ES2 and SK-ES cells with shRNA-mediated knockdown of HOTAIR or nonsilencing control as described above, and the hTERT-hMSC-GFP and hTERT-hMSC-HOTAIR cells. In the ES2 cell lines, we examined those genes that had increased expression when HOTAIR expression was repressed (Figure 5A and B). We found that the loss of HOTAIR also resulted in decreased H3K4Me1 and increased H3K4Me2 around the transcriptional start sites (TSS) of these genes (Figure 6, SI data). In the hTERT-hMSC cells, HOTAIR expression similarly induces increased H3K4Me1 around the TSS of HOTAIR-repressed genes, with a less-pronounced but still significant decrease in H3K4Me2 (SI 7, SI data). In the SK-ES cells, loss of HOTAIR is again associated with significantly decreased H3K4Me broadly adjacent to the TSS, though not as markedly as in the ES2 or hTERT-hMSC cells, and with a particularly increased amount of H3K4Me2 immediately adjacent to the TSS (SI 8, SI data). These data demonstrate that genes repressed in the context of HOTAIR expression consistently have histone methylation changes at their promoters associated with LSD1-mediated epigenetic repression.
DISCUSSION
The Ewing sarcoma family of tumors is characterized by the EWS-FLI1 fusion gene or similar fusion genes. The resultant fusion proteins are the drivers of oncogenesis in these tumors, but these proteins are toxic to most studied normal cells and even other types of cancer cells. The paucity of additional mutations in these tumors suggests that specific epigenetic profiles may allow viability and tumorigenesis of the EWS-FLI1-expressing cells. Our studies suggest that the expression of the lncRNA HOTAIR redirects epigenetic regulation to induce just such a permissive state and allow formation of Ewing sarcoma tumors. HOTAIR is consistently overexpressed in virtually all ESFT primary tumors and cell lines examined by either microarray(18), RNA-Seq(26), or RT-qPCR. This is in stark contrast to the minority of ESFT tumors with genomic mutations, as assayed in three large whole genome sequencing studies(6-8). This consistency of expression supports a function for HOTAIR within this cancer. We additionally confirmed that HOTAIR expression is not driven by EWS-FLI1, further supporting an independent function in ESFT tumorigenesis and viability.
Phenotypically, HOTAIR directly affects growth of ESFT. In the hTERT-hMSC model, overexpression impacted both two-dimensional cell growth and anchorage-independent tumorsphere formation. In contrast, in the ESFT cell lines, knockdown of HOTAIR expression repressed only tumorsphere formation. This difference may be explained by the necessity of HOTAIR in ESFT. Using shmiRNA methods and strict selection conditions, knockdown below 25% was impossible. We additionally tried to knockout the HOTAIR alleles in ESFT cell lines by adapting an RNA-destabilizing element (RDE) (27) that also integrated green or red fluorescent proteins into the locus. While we could successfully insert one RDE into cells, we could not subsequently generate viable cells that lost expression of the second allele (data not shown). Our inability to completely silence HOTAIR expression suggests that a single copy of HOTAIR is both necessary and sufficient for viability of the ESFT cell lines we used. As such, repression of HOTAIR expression in our ESFT cell lines to a level that inhibits proliferation may simply make them nonviable. A conditional expression system may allow for titratable HOTAIR expression while genomic copies of the gene are knocked out, followed by phenotypic evaluation as HOTAIR expression is reduced. Nonetheless, the effects of HOTAIR loss in ESFT 3D colony and tumor formation phenocopy those effects found with inhibition of LSD1 in ESFT(19), supporting a pathophysiologic role of HOTAIR in ESFT tumor viability.
The gene expression analysis of our cell line models further defines how HOTAIR affects ESFT formation and viability. For the RNA-Seq analyses, we aimed to evaluate the maximal effect of HOTAIR expression. We generated hTERT-hMSC cell lines with stable HOTAIR and EWS-FLI1 expression consistent with that seen in ESFT cell lines. However, to maximally reduce HOTAIR expression in ESFT cell lines, we used GapmeRs for transient knockdown. This approach may limit the ability to measure HOTAIR's activity in modulating epigenetic regulation, which has some rapidly altered features but also additional characteristics only seen over time. These different methods used to modulate HOTAIR expression likely contribute to the difference in the number of genes identified in the different models. In the stable hTERT-hMSC system, thousands of genes were found to be differentially expressed, while comparatively fewer were identified in the transient ESFT models. Nonetheless, key effects were still identified. In particular, there is a set of genes whose expression is regulated by HOTAIR independently of EWS-FLI1 and conversely another set regulated specifically by EWS-FLI1, suggesting complementary functions of the two genes in ESFT tumorigenesis.
In both ESFT and hTERT-hMSC models of HOTAIR and EWS-FLI1 expression, differential expression of genes integral to tumor formation were identified, including cytoskeletal and adhesion proteins (collagens, keratins, cadherins and protocadherins) and matrix metalloproteases. The specific genes affected in each individual cell line are variable, consistent with data showing significant diversity in the gene expression profiles of ESFT in general (26, 28). Regardless, we identified key pathways affecting tumorigenesis, including cell motility and migration, across the different models. It is particularly noteworthy that these pathways were previously identified to be biomarkers of survival in patients(28).
We also identified other critical pathways that are differentially affected by HOTAIR, including DNA repair pathways and normal differentiation and development. ESFT are characteristically undifferentiated, which contributes to the inability to as yet define the cell of origin for these tumors. Additionally, errors in DNA repair pathways have been hypothesized to contribute to oncogenesis(29, 30). HOTAIR may function in the cell of origin to maintain the cells in an undifferentiated and EWS-FLI1 receptive state, tolerant of DNA damage induced by pathways activated by the fusion protein.
HOTAIR was originally demonstrated in fibroblasts to function as a scaffold for the LSD1/REST complex at its 3’ end and PRC2 at its 5’ end(9, 31). We confirmed the interaction of HOTAIR and LSD1, the correlated effect on histone methylation and gene expression in ESFT cell lines, and the necessity of this interaction for the anchorage-independent growth phenotype seen in the hTERT-hMSC model. It is important to acknowledge that the effects of HOTAIR varied among the disease models, with far greater effects seen on gene expression in the hMSC models than in ESFT cell lines. This may have been due to greater effects of stable HOTAIR expression in contrast to the transient effects of HOTAIR repression by the GapmeRs, as noted above. Alternatively, HOTAIR may induce gene expression changes during ESFT transformation but may not be required to maintain gene expression at all sites after transformation occurs. A comprehensive evaluation of HOTAIR binding across the genome, and a comparison of that binding to histone methylation, will elucidate the direct effects of HOTAIR in ESFT.
Additional work is warranted in investigating other functions of HOTAIR on gene expression, including its interaction with PRC2 and other regulatory mechanisms, such as effects on DNA methylation independent of LSD1 or PRC2, as recently described(32). We have begun additional work on the interaction of HOTAIR and PRC2, but these studies are much more complicated. A prior study showed a lack of effects on H3K27 methylation despite the presence of HOTAIR(9), which may be due to additional cofactors that allow PRC2 binding but prevent methyltransferase activity(33). We felt it important to make our present findings available while we perform more expansive studies on HOTAIR in ESFT, particularly because of the demonstration of LSD1's importance in ESFT(34) and the development of novel LSD1 inhibitors(19) for use in this disease. These prior studies showed functions of LSD1 through the Nucleosome Remodeling and Demethylation (NuRD) Complex, but our studies would support evaluation of the effects of the LSD1 inhibitors on HOTAIR-directed H3K4 modifications.
A recent publication by Amandio et al., has called the biological relevance of HOTAIR into question(35). They examined the effect of the mouse Hotair ortholog on mouse developmental biology, in follow-up to prior studies showing discordant effects in different mouse strains. They concluded that Hotair had little effect on mouse morphology or on H3K27 methylation patterns, questioning the necessity of the lncRNA and its role in disease. While intriguing, this work by Amandio et al. has its limitations. They did not examine the effects of Hotair on LSD1 function or targeting, a key pathway identified in our studies in ESFT. Additionally, they examined gene expression effects of Hotair deletion at a single embryonic timepoint in whole tissues consisting of dozens of cell types. HOTAIR expression and its effects have been shown to be widely variable across cell types and degree of differentiation(36). Our studies also examined HOTAIR in ESFT tumors and models of disease, which while heterogeneous in their composition, are largely comprised of a single cancer cell type. Loss of Hotair may not be associated with premature differentiation, but HOTAIR expression is associated with inhibition of differentiation, a feature observed by Amandio et al. Finally, prior study by that group demonstrated poor sequence homology between Hotair and HOTAIR, with the mouse lncRNA showing poor conservation of the domains responsible for EZH2 and LSD1 interaction(37). As they state, “it nevertheless suggests that the function of this RNA in mice is not identical to that described for its human cognate(37).” This fact is further supported by the scores of reports on the expression of HOTAIR in multiple cancers, its correlation with disease outcome, and its biological functions in these cancers. While studies are needed to discriminate the direct function of HOTAIR from its indirect effects on downstream targets, our work supports these effects and merits such additional study.
The identification of HOTAIR in ESFT has implications on the understanding of the disease and on its treatment. The driver of HOTAIR expression has yet to be confirmed, but its expression was previously described in normal embryonic stem cells(11) and cancer stem cells(38, 39). The ESFT cell of origin may inherently have high HOTAIR expression as a normal part of development or an error in that process. HOTAIR may prime these cells for EWS-FLI1-mediated malignant transformation. We are currently evaluating what additional genetic, transcriptional,or epigenetic events can collaborate with HOTAIR and EWS-FLI1 to induce tumor formation in vivo in our 1o and hTERT-hMSC models. Further study of the regulation and function of HOTAIR in these models and in primary ESFTs may elucidate the oncogenesis of these cancers.
HOTAIR offers novel therapeutic options for ESFT. Indirectly, its specific activity through LSD1 suggests that these pathways may be specifically targetable in ESFT in a fashion that would be synergistic to attacks on EWS-FLI1. LSD1 inhibitors are being developed against ESFT(19), though that work was based on LSD1 interaction with the NuRD complex. Our findings suggest LSD1 may have diverse functions in ESFT, and disruption of HOTAIR-dependent LSD1 function may augment antitumor effects. New therapeutics may also be developed against HOTAIR itself, such as antisense oligonucleotides that have been previously generated against lncRNAs(40). As HOTAIR is not normally expressed in most tissues, this strategy may have fewer adverse effects systemically. Additional drugs could be identified or designed to interrupt HOTAIR's interactions with the epigenetic complexes, avoiding the normal HOTAIR-independent functions that LSD1 has in nonmalignant tissues. These avenues have promise in the generation of more specific and effective therapies for ESFT and other HOTAIR-expressing cancers.
MATERIALS AND METHODS
Cell Culture and Plasmids
ESFT cell lines ES1, ES2, ES3, ES4, ES6, ES7, ES8, A673 and TC-71 were obtained from Peter Houghton (UT Health Science Center, San Antonio, TX), and 5838, RD-ES, SK-ES and TC-32 were obtained from Timothy Cripe (Nationwide Children’s Hospital, Columbus, OH). IMR90 cells were obtained from Ryan Roberts (NCH). Cell line identities were confirmed upon receipt, start of studies, and annually by short tandem repeat (STR) profiling, using the Promega Powerplex 16 system, last performed on October 18, 2016 (completion of studies). Cells were tested for Mycoplasma with the SouthernBiotech Mycoplasma detection kit (Birmingham, AL), tested every 3 months, last in October 2016. Mycoplasma-contaminated cells were either replaced with uninfected cells from prior stores or treated with Plasmocin (Invivogen, San Diego, CA). The immortalized human marrow-derived mesenchymal stem cell line (hTERT-hMSC) was purchased from Applied Biological Materials, Inc. (British Columbia, Canada). Primary hMSCs were given as a generous gift from Edwin Horwitz (NCH). Cell lines were maintained in DMEM media supplemented with 10% FBS and Penicillin-Streptomycin at 370C with 5% CO2.
pLenti-HOTAIR-GFP was constructed by inserting the HOTAIR sequence at the PmeI restriction site in the pLenti-CMV-GFP-2A-Puro vector (Applied Biological Materials, Inc.) using the Infusion HD kit (TakaraBioUSA, Mountain View, CA). The HOTAIR deletion variants were then cloned from this plasmid by inverse PCR and Gibson cloning. For pLenti-EWS-FLI1-GFP, EWS-FLI1 coding sequence containing a self-cleaving peptide sequence E2A at the 3’ end was inserted 5’ to the GFP-2A-PURO sequence using inverse PCR and Gibson cloning. All cloning primers are listed in the supplementary methods.
HOTAIR shmiRNA construct shHOTAIR2 was obtained from Dharmacon (Lafayette, CO, USA). The shmiRNA sequences from the inducible TRIPZ vector were cloned into the constitutively active pGIPZ vector by Gibson cloning. Antisense LNA GapmeR for HOTAIR (300631–724) and nonsilencing control (300615–08) was purchased from Exiqon (Vedbaek, Denmark).
siRNA-mediated EWS-FLI1 knockdown
The Luciferase-RNAi (Luc-RNAi) and EWS/FLI-RNAi (EF-2-RNAi) constructs are previously described(4, 41). ESFT cells were infected with Luc-RNAi or EF-2-RNAi. Cells were grown in appropriate selection media for 72 hours and total RNA was extracted using an RNAeasy Kit (Qiagen). RT-qPCR was performed using SYBR green fluorescence for quantitation. Fold-change of genes was determined relative to the control gene, RPL19.
Transfection, Virus Production and Transduction
The NEON Transfection System (ThermoFisher Scientific, Waltham, MA) was used for plasmid transfection. GapmeRs were transfected using Lipofectamine 3000 (ThermoFisher Scientific) per manufacturer protocol.
Lentivirus was generated using HEK-293T cells. Briefly, cells were seeded onto a 10 cm tissue culture plate at 70% confluency, then transfected with shmiRNA plasmid using the calcium phosphate precipitation method, pCMV-dR8.2, and pMD2.G at a ratio of 4:3:1. Lentivirus-containing media were collected at 48 and 72 hours, pooled and concentrated using Lenti-X Concentrator (Takara Bio USA). Target cells were then transduced with lentiviral particles using polybrene (8 μg/ml), with exposure for 24 hours then selection with puromycin for 3–7 days.
RNA extraction and RT-qPCR
Total RNA from cells was extracted using NucleoSpin RNA purification kit (Takara Bio USA) per manufacturer instructions, then used for cDNA synthesis using Maxima RT cDNA First Strand Synthesis kit (ThermoFisher Scientific). qPCR was performed using KiCqStart SYBR Green qPCR ReadyMix (Sigma-Aldrich, St. Louis, MO) using the ABI PRISM 7900HT thermal cycler (ThermoFisher Scientific). Primers used are listed in the Supplementary Methods.
Soft Agar Assay
Colony formation assay was performed on soft agar as previously described(42), with 5×103 cells seeded in soft agar/media. After 2 or 3 weeks, fluorescent or brightfield photomicrographs were taken of the wells, and colonies were counted with the ImageQuant TL software (GE), with a colony defined as a cohesive collection of viable cells > 50 mcm in diameter. Experiments were completed in independent triplicates.
Cell Proliferation Assay
Cell proliferation was measured using the alamarBlue cell viability assay (ThermoFisher Scientific). 2.5 × 103 cells of each type were seeded in triplicate wells of a 96-well plate and grown in complete media with a separate plate for each time point. At each time point, alamarBlue reagent was added to each well and incubated for 2 hours. Fluorescence intensity was then measured, normalized for background fluorescence, then compared to day 0 readings. Experiments were completed in independent triplicates.
Flow cytometry analysis
hTERT-hMSCs overexpressing wild-type HOTAIR or deletion mutants were analyzed on a BD LSRII flow cytometer (BD Biosciences, San Jose, CA) using allophycocyanin-conjugated anti-human CD105 and phycoerythrin-conjugated anti-human CD73 (BD Biosciences). Data were analyzed using FlowJo software v7.6 (Tree Star, Inc., Ashland, OR).
Osteogenic and Adipogenic Differentiation
5 × 10^4 cells/well were plated in 24-well plates and were cultured in DMEM supplemented with 10% FBS. At about 90% confluency, the medium was switched to either NH OsteoDiff Medium or AdipoDiff Medium (Miltenyi Biotec Inc, Auburn, CA) and the cultures were maintained for 21 days. Osteoblastic differentiation was detected by calcium deposition visualized by staining with 1% alizarin red S solution (Sigma-Aldrich). Adipogenic differentiation was detected by Oil Red O staining (Sigma-Aldrich).
Tumor xenograft growth assay
1×106 ESFT cells (ES2, SK-ES, TC32) or 1×107 1o hMSC, hTERT-hMSC, or IMR90 HOTAIR-EWS-FLI1 expressing cells were resuspended in 50 mcL PBS and mixed with 50 mcL of Matrigel (Corning) and injected subcutaneously into the flanks of SCID mice (Envigo), with 5 mice per injection group. Mice were monitored twice weekly for weight, body condition, and tumor size. Tumor volumes were estimated using the formula V=(length × (width)2)/2, per institutional protocol. Mice with SK-ES tumor xenografts were monitored for 17 days; mice with ES2 and TC32 tumor xenografts were monitored for 21 days; mice with hTERT-hMSC, 1o hMSC, or IMR90 cell tumor xenografts were monitored for 12 weeks. Mice were then euthanized at the time endpoint and tumors harvested and fixed in formalin. All studies were designed in accordance with Nationwide Children’s Hospital IACUC guidelines and performed under IACUC-approved protocols.
RNA immunoprecipitation (RIP)
RIP was performed as described(43, 44). 2×107 cells were grown to 80% confluency, cross-linked with 1% formaldehyde, quenched with 2.66 M glycine, and then washed and lysed in IP lysis buffer. Samples were sonicated by probe sonication. 10 mg of lysate was incubated rotating overnight with 5 mcg of anti-LSD1 antibody (C69G12), or IgG control, Cell Signaling) at 4oC then with Protein A/G+ Agarose Beads (TFS). Samples were then centrifuged, washed, and heated to 70oC to reverse crosslink. RNA was isolated with the Nucleospin XS kit (Clontech) including DNAse treatment. RNA was then used for reverse transcription with Superscript IV, then used for qPCR as above. RNA expression was compared to expression in input. RIP experiments were performed in independent triplicate.
RNA-Seq and data analysis
Total RNA was isolated from cells using the Nucleospin kit as above, with RNA quality assessed using Agilent 2100 bioanalyzer and RNA NanoChip kit (Agilent Technologies, CA). RNA was DNase-treated then subjected to rRNA removal with the Ribo-Zero rRNA removal kit (Illumina). Direction-oriented libraries were constructed from first strand cDNA using ScriptSeq v2 RNA-Seq library preparation kit (Epicentre Biotechnologies, WI). Briefly, 50 ng of rRNA-depleted RNA was fragmented and reverse transcribed using random primers containing a 5’ tagging sequence, followed by 3’end tagging with a terminal-tagging oligo to yield di-tagged, single-stranded cDNA. Di-tagged cDNA was purified via magnetic beads then amplified by limit-cycle PCR using primer pairs that anneal to tagging sequences and add adaptor sequences required for sequencing cluster generation. Amplified RNA-seq libraries were purified using AMPure XP System (Beckman Coulter). Library quality was determined via Agilent 2200 Tapestation using High Sensitivity D1000 screen tape, and quantified by Qubit flourometer with dsDNA BR assay (Thermo Fisher Scientific). Paired-end 150 bp sequence reads were generated using the Illumina HiSeq 4000 platform.
A minimum of 69 million paired-end 150 bp RNA-Seq reads were generated for each sample (range 69–111 million). Each sample was aligned to the GRCh38.p5 assembly of the H.Sapiens reference from NCBI (http://www.ncbi.nlm.nih.gov/assembly/GCF_000001405.31/) using version 2.5.0c of the RNA-Seq aligner STAR (http://bioinformatics.oxfordjournals.org/content/early/2012/10/25/bioinformatics.bts635). Transcript features were identified from the GFF file provided with with the GRCh38.p5 assembly and raw coverage counts were calculated using HTSeq (http://www-huber.embl.de/users/anders/HTSeq/doc/count.html). The raw RNA-Seq gene expression data was normalized and post-alignment statistical analyses performed using DESeq2 (http://genomebiology.com/2014/15/12/550) and custom analysis scripts written in R. Comparisons of gene expression and associated statistical analysis were made between different conditions of interest using the normalized read counts. All fold change values are expressed as test condition/control condition, where values less than one are denoted as the negative of its inverse. Transcripts were considered significantly differentially expressed using a 10% false discovery rate (DESeq2 adjusted p value <= 0.1) and a fold-change cut-off of 2 between the control and test samples. Complete dataset will be accessioned to GEO. Annotated gene lists of differentially expressed features are available in the Supplementary Data and were used for analysis of statistical overrepresentation of biological pathways using PANTHER (45), overrepresentation test release 20160715, GO database release 2016–11–30.
Chromatin Immunoprecipitation-Sequencing (ChIP-Seq)
ChIP-Seq was performed as described(46). Chromatin was crosslinked with fresh 1% formaldehyde for 4 minutes and sheared by the Covaris ME220 for 12 minutes. Anti-H3K4Me1 (ab8895) and anti-H3K4Me2 (ab32356, AbCam) antibodies were used with M-280 Sheep Anti-Rabbit IgG Dynabeads™ (TFS) for ChIP. Isolated chromatin was amplified with the Kapa Library Amplification Kit (Kapa Biosystems) and samples analyzed by HiSeq Illumina Genome Analyzer. Peak calling was performed using the USeq software. Complete data analysis of ChIP-Seq data is included in the supplementary data.
Statistical Analysis
All statistical analyses excluding the RNA-Seq data were completed using the GraphPad Prism 7 software (GraphPad Software, Inc.). Where appropriate, the two-tailed Student’s t-test was used to calculate significant differences between comparison groups in the experiments above. For multiple comparisons, one-way ANOVA was used with the Bonferroni correction for multiple comparisons against a single control.
COMPETING INTERESTS
None
ACKNOWLEDGEMENTS
This work was supported by internal funding from the Research Institute at Nationwide Children’s Hospital.