The Long Noncoding RNA NEAT1 Promotes Sarcoma Metastasis by Regulating RNA Splicing Pathways

This article is featured in Highlights of This Issue, p. 1441

Metastasis is the cause of death for 90% of patients with cancer (1, 2). In general, the metastatic process involves multiple inefficient steps: primary tumor cells invade into surrounding tissues; tumor cells intravasate into the circulatory system; tumor cells extravasate into the secondary distant site; and tumor cells finally colonize and form secondary tumors in the distant tissue (3). Despite great progress in cancer biology over the last few decades, the underlying mechanisms by which primary cancer cells metastasize to distant tissues remain poorly understood (2). This knowledge gap limits our ability to develop therapeutic strategies to prevent and effectively treat patients with metastasis (4). Currently, approximately 50% of humans with high-grade soft-tissue sarcoma (STS) develop lung metastasis with a median survival of 15 months (4). Thus, a better understanding of the underlying mechanisms by which sarcoma metastasizes to the lung is critical for developing novel therapeutic approaches.

An understudied class of genomic elements with the potential to regulate metastasis are long noncoding RNAs (lncRNA) named for their sequence length of more than 200 nucleotides. LncRNAs generally do not code for proteins, but instead control key cellular functions by binding to proteins, DNA, and other RNAs, or by producing small functional peptides (5–9). Interestingly, recent studies suggest a critical role of lncRNAs in regulating cancer metastasis (10, 11). For example, the Nuclear Enriched Abundant Transcript 1 (NEAT1) is a conserved lncRNA of two major isoforms (short isoform, NEAT1_1 and long isoform, NEAT1_2), which is positively associated with poor prognosis of several cancer types such as gastric cancer (12) and ovarian cancer (13). However, whether the lncRNA NEAT1 regulates metastasis is unknown.

An emerging mechanism for regulating metastasis is RNA splicing. RNA splicing is the process by which introns are removed from the nascent precursor RNA (pre-mRNA) to form mature RNA (mRNA). RNA splicing is critical for normal gene expression and proteome diversity (14). The recent development of high-throughput RNA-sequencing (RNA-seq) has led to the identification of dysregulated splicing pathways in cancer (14, 15). Moreover, several studies have determined that mis-splicing significantly affects the metastatic process (16). For instance, the alternative splicing regulator serine and arginine-rich splicing factor 1 (SRSF1) induces metastasis by shifting the expression of the kinase ribosomal protein S6 kinase (S6K1) to a short oncogenic splice isoform (17). Understanding how splicing is regulated could lead to the discovery of novel biomarkers and therapeutic targets. In this study, we utilized a genetically engineered mouse model (GEMM) that mimics human undifferentiated pleomorphic sarcoma (UPS) where 40% to 50% of mice develop lung metastasis (3, 18–21). Through RNA-seq of paired primary KP tumors and lung metastases, we found that the lncRNA Neat1 is upregulated in the lung metastases. Using CRISPR/Cas9-mediated knockout (22–26), RNA pulldown assays followed by mass spectrometry, and RNA analysis of mouse and human lung metastases, our results suggest that the lncRNA Neat1 contributes to sarcoma metastasis by regulating RNA splicing.

Kras^LSL-G12D/+ mice (27) were a gift from T. Jacks (Massachusetts Institute of Technology, Cambridge, MA). Trp53^Flox/Flox mice (28) were a gift from A. Berns (The Netherlands Cancer Institute, Amsterdam, the Netherlands). Adenovirus expressing Cre recombinase was injected into the hind limb of Kras^LSL-G12D; Trp53^Flox/Flox (KP) mice (18). Once the primary tumors reached around 250 mm³, the tumor-bearing legs were amputated and the mice were further followed for the development of lung metastases (3). Male (n = 7) and female (n = 7) mice were included for RNA-seq. Paired primary and lung metastases that were used for RNA-seq were developed from 5 male and 2 female mice. Both male (n = 35) and female (n = 17) mice that developed primary sarcomas were chosen for qPCR validation of Neat1 expression in primary tumors and lung metastases. Lung metastasis developed in 18 male and 5 female mice. Seventeen male and 12 female mice did not develop lung metastasis and were euthanized. Athymic nude (nu/nu) mice (5–6 weeks old) for the transplant study were purchased from Taconic Biosciences and maintained in Duke University's accredited animal facility. All animal studies were performed in accordance with protocols approved by the Duke University Institutional Animal Care and Use Committee.

Study of human sarcoma samples was performed under the Duke University Institutional Review Board (IRB, Durham, NC) protocol Pro00087063. Sarcoma samples including tissue microarrays (29) were collected at the following institutions under their respective institutional protocols: the Samuel Lunenfeld-Tanenbaum Research Institute (LTRI, Toronto, Ontario, Canada, Research Ethics Board–approved protocol 01-0138-U), the Memorial Sloan Kettering Cancer Center (New York, NYIRB-approved protocol02-060), the MD Anderson Cancer Center (Houston, TX, IRB-approved protocol Lab-04-0890), and the Cleveland Clinic (Cleveland, OH, IRB-approved protocol 06-977).

Primary and metastatic KP and KI sarcoma cell lines were dissociated from autochthonous KP and KI mouse tumors, respectively. They were cultured in DMEM (Thermo Fisher Scientific) supplemented with 10% FBS and 1% antibiotic–antimycotic (Thermo Fisher Scientific) and incubated at 37°C with 5% CO₂ in a humidified cell culture incubator. These primary mouse sarcoma cell lines were not authenticated or tested for Mycoplasma.

Harvested primary tumors and lung tissues were fixed in 4% formalin, paraffin-embedded, and 5 μm sections were analyzed by hematoxylin and eosin (H & E) staining using standard methods.

Stained tissue microarray sections were captured with a 20× objective on a Brightfield Microscope (Leica DM2000 LED). The same acquisition settings were maintained for each image. Images were named with nonidentifying numbers and forwarded to an investigator blinded to experimental groups who measured the number of positively stained nuclei. Images were batch processed in ImageJ with a script. Briefly, the protocol was as follows (see Supplementary Materials and Methods for details): (i) color Threshold was used to select the positively stained area to measure. (ii) The image was converted to Binary. (iii) Watershed transformation was applied. (iv) The Analyze Particles function was used to count the number of individual stained areas. The same particle size parameters were maintained for each image. An example of measurements on a staining tissue section is shown in Supplementary Fig. S1.

A total of 5 × 10⁵ sarcoma cells were injected into the tail vein of nude mice. Two weeks after injections, the nude mice were euthanized, the lungs were processed, and a second person blinded to the cell lines injected performed the analysis for metastasis. Formalin-inflated, paraffin-embedded lung sections were subjected to H&E staining and images covering overlapping parts of the lungs were captured with using 5 × objective on a Brightfield Microscope (Leica DM2000 LED). The same acquisition settings were maintained for each image. Partial images were stitched into whole lung sections using Leica LAS V4.5 Software and named with nonidentifying numbers. Images were forwarded to an investigator blinded to experimental groups, who measured proportional sarcoma colonization area in each lung. Regions of sarcoma colonization were detected as dense areas of hematoxylin staining. Images were then analyzed using the ImageJ software and any nonlung tissue, if present, was masked out. Images were batch processed with a home-made script (Supplementary Fig. S2). The protocol was as follows: (i) a Gaussian blur was applied to the image, (ii) a color threshold was applied to select the area of interest in the lung with tumor, (iii) the image was then converted to binary, (iv) the positive area was measured, (v) steps (i) to (iv) were repeated with different threshold parameters and an additional “Fill Holes” plugin was applied after step (iii) on the same image to measure the whole lung area, (vi) the proportion of lung colonization area was calculated by dividing area with tumor by the total lung area. An example of measurements on a whole-lung section is shown in Supplementary Fig. S3.

RNA was extracted from tissues and cells using the Direct-zol RNA Miniprep Kit (Zymo Research). cDNA was prepared using iScript Advanced cDNA Synthesis Kit (Bio-Rad). RT-PCR was performed in biological duplicates using QuantStudio 6 System (Applied Biosystems). Details of probes and primers can be found in Supplementary Table S1.

Twenty KP primary tumor samples were snap frozen in liquid nitrogen and then total RNA was isolated from those samples using TRizol Reagent (Invitrogen) per the manufacturer's recommendation. Then RNA was prepared with TruSeq Stranded Total RNA library prep kit with Ribo-Zero Human/Mouse/Rat Set A (Illumina). Libraries were sequenced on Illumina's HiSeq2000. The RNA-seq data are deposited in the Gene Expression Omnibus with accession number of the RNA-seq data assigned as GSE139574.

To analyze the correlation of the altered (copy-number alteration > 1, mutated, and mRNA median z-score > 2 or <−2) KHSRP or SMC2 and the prognosis of patients with STS, we used The Cancer Genome Atlas dataset (30, 31). We chose studies of “sarcomas” and “query by gene,” then input KHSRP or SMC2. After selecting genomic profiles of “mutations,” “putative copy-number alterations from GISTIC,” and “mRNA expression z-scores (RNA Seq V2 RSEM),” we calculated the disease/progression-free Kaplan–Meier estimate based on the cases with or without alteration in query gene. The “sarcoma” dataset we chose in the cBioPortal survival data includes leiomyosarcoma (n = 95), dedifferentiated liposarcoma (n = 54), UPS (n = 48), myxofibrosarcoma (n = 21), synovial sarcoma (n = 10), malignant peripheral nerve sheath tumor (n = 8), desmoid/aggressive fibromatosis (n = 2), pleomorphic liposarcoma (n = 2), and STS (n = 1).

RNA-seq data from Neat1 wild-type and knockout mouse embryonic fibroblasts (MEF) were published previously (32; GSE100098). To perform gene enrichment analysis, the identity of all differentially expressed genes was submitted to EnrichR analysis (33). The enriched spliceossomal assembly signature (P = 0.00546) depicted is from the Biocarta 2016 library.

The quality of the sequencing reads was assessed using FastQC (v0.11.5; ref. 34). The reads were aligned to the UCSC mm10 mouse genome from the iGenome Project using STAR (v2.5.2b; ref. 35) and mapped to the mouse transcriptome annotated by GENCODE (Release M14; ref. 36). Raw gene counts were quantified using the HTSeq (37) tool implemented in the STAR pipeline. Differentially expressed genes were first identified by modeling the raw counts within the framework of a negative binomial model using the R package DESeq2 (v1.24.0; ref. 38), and the genes were considered lncRNA genes when their annotated biotypes fall into the following categories: antisense, bidirectional_promoter_lncRNA, lincRNA. Gene set enrichment analysis was performed on the gene ontology terms (39) and Kyoto Encyclopedia of Genes and Genomes pathways (40) using the R package gage (v2.34.0; ref. 41). Differential alternative splicing events were identified by rMats (v4.0.2; ref. 42) and reported in inclusion level difference values (lung metastases vs. primary sarcoma). All P values were adjusted for multiple testing using the Benjamini–Hochberg method within the 10% false discovery framework. The analyses were scripted in the R statistical environment (https://www.R-project.org/) along with its extension packages from the Comprehensive R Archive Network (https://cran.r-project.org/) and the Bioconductor Project (43).

RNA-seq libraries were prepared using the TruSeq RNA Sample Kits (Illumina). Libraries were sequenced on Illumina's HiSeq2000 and FASTQ files were generated by CASAVA. The reads were aligned to hg19 using STAR (v2.3). Gene expression was quantified using GENCODE version 19 as reference and transcript abundance was estimated using HTSeq-count. Gene expression was then normalized within the sample cohort using DESeq2 (v1.10.1). p53 mutational status in patients with sarcoma is presented in the Supplementary Table S2.

Human NEAT1 (#411531) and NEAT_2 (#411541-C2) probes were purchased from Advanced Cell Diagnostics. NEAT1 RNA ISH in human tissue microarrays was performed using Advanced Cell Diagnostics RNAscope under standard conditions.

T7-Neat1_1 plasmid was linearized by digestion with the BstB1 restriction enzyme and T7-Neat1_1 antisense plasmid was linearized by digestion with the PciI restriction enzyme. Digested plasmids were purified using DNA Clean and Concentrator-5 Kit (Zymo Research). Neat1_1 or Neat1_1 antisense was in vitro transcribed in a reaction containing linearized plasmid DNA, T7 RNA polymerase (New England Bio), murine RNase Inhibitor (New England Bio), and ATP, GTP, CTP, and UTP (New England Bio) including 10% biotinylated UTP (PerkinElmer). Synthesized RNA was purified using RNA Clean and Concentrator Kit (Zymo Research). Purified RNA was incubated in the thermo cycler at 60°C for 10 minutes and then cooled in ice. Then RNA was mixed with the protein lysate from sarcoma cells and the mixed complex was further pulled down, washed, and purified using Streptavidin Beads (Pierce). Next, streptavidin beads were mixed with sample buffer and boiled at 95°C for 5 minutes. Supernatant was transferred into a new tube after samples were spun down. Half of the supernatant was used for Western blotting and the rest was analyzed by mass spectrometry in the Duke Proteomics and Metabolomics Core Facility.

A total of 1.2 × 10⁶ 293T cells (ATCC) were seeded in 6 cm dish (Thermo Fisher Scientific) 24 hours prior to transfection. The lentivector (1 μg) plus psPax2 (0.9 μg; Addgene) and VSVG (0.1 μg; Addgene) were transfected into those cells using the reagent Mirus LT1 (Mirus). After 24 hours, the supernatant media were replaced with fresh media containing 30% FBS and 48 hours later, the supernatant was harvested and filtered with 0.45-μm syringe filter (Thermo Fisher Scientific). Polybrene (Millipore-Sigma) was mixed with 500 μL supernatant containing the lentivirus and 1 mL normal media with a final concentration at 8 μg/mL. The mixture media were then added to sarcoma cells to transduce the lentivirus.

Electrotransfection was performed with 10⁶ mouse primary cancer cells. The cells were resuspended in 100 μL of pulsing buffer (Opti-MEM) with 6 μg of px333-Cas9-Neat1 sgRNA 3+4 on ice. The cell suspension was transferred to electroporation cuvettes with two parallel plate electrodes spaced 4 mm apart. Cells were electrotransfected with the BTX ECM 830 Square Wave Electroporation System (Harvard Apparatus). Cells were treated with eight electric pulses at 240 V/4 mm, 5 millisecond duration, and 1 Hz frequency. The cuvettes were kept at room temperature for 10 minutes following the pulse application to allow the cells to recover before pipetting them to a 6-well plate with full cell culture medium.

Cells were lysed in RIPA Buffer (Thermo Fisher Scientific) and lysate supernatant was separated from debris. Protein concentration was determined by Bradford Assay (Bio-Rad). Lysates were boiled in 4× Sample Buffer (Thermo Fisher Scientific) at 100°C for 5 minutes, then cooled to room temperature. Samples were electrophoresed in 4%–20% precast Protein Gel (Bio-Rad) at 100 V for 60 minutes before transfer to Immobilon-P Polyvinylidene Difluoride Membrane (Millipore-Sigma). Membranes were then blocked using the Odyssey Blocking Buffer (LI-COR) and incubated overnight at 4°C with primary antibodies diluted in TBS buffer supplemented with 0.1% Tween-20 (TBS-T): KHSRP, 1:1,000 dilution (Abcam); SMC2, 1:5,000 dilution (Abcam); and GAPDH, 1:5,000 dilution (Proteintech). Membranes were washed three times in TBS-T for 5 minutes and then incubated with secondary antibodies: goat anti-rabbit IRDye800 (LI-COR Biosciences, P/N 925-32211) and goat anti-mouse IRDye680 (LI-COR Biosciences, P/N 925-68070) both at 1:20,000 dilution in TBS-T for 1 hour at room temperature. The membranes were washed three times in TBS-T for 5 minutes and imaged using an Odyssey CLx (LI-COR Biosciences). Image analysis and quantification were performed using the Image Studio (Version 5.2, LI-COR Biosciences, P/N 9140-500) software. Full blots with protein ladders (LI-COR Biosciences, P/N 928-60000) are shown in Supplementary Fig. S4.

LncRNA Neat1 was assessed using northern blotting as described previously (44). RNA was extracted from cells using the Direct-zol RNA Miniprep Kit (Zymo Research). The probe for Neat1 was labeled with radioactive α-dCTP (PerkinElmer, P/N BLU513H250UC). Full blots with RNA ladders (Thermo Fisher Scientific, P/N SM1823) are shown in Supplementary Fig. S4.

Results are presented as means ± SEM unless otherwise indicated. Before analysis, all data were displayed graphically to determine whether parametric or nonparametric tests should be used. Two-tailed Student t test was performed to compare the means of two groups. To test the difference between groups, one-sided Wilcoxon rank-sum test was used for unpaired samples, and one-sided Wilcoxon signed rank test was used for paired samples. All calculations were performed using Prism 6 (GraphPad).

To identify genes that are differentially expressed between the primary sarcomas and lung metastases, we generated KP mice, which develop autochthonous sarcomas in the muscle following injection of an adenovirus expressing Cre recombinase. Cre activates the expression of a mutant Kras^G12D and deletes both alleles of Trp53 gene (Fig. 1A). As each primary sarcoma (n = 14) reached the size of around 250 mm³, the leg bearing the tumor was amputated and the tumor stored for subsequent RNA-seq. Within 2 months, 6 mice developed lung metastases: 2 mice developed one lung metastases each, 2 mice developed two lung metastases each, and 2 mice developed four lung metastases each. RNA was then extracted from the six primary tumors and the 14 matched lung metastases. Therefore, a total of 20 RNA samples were analyzed by RNA-seq (Fig. 1B).

We identified differentially expressed lncRNAs in the lung metastases compared with paired primary mouse sarcomas using a 5% FDR. (Fig. 2A; Supplementary Table S3). Among the lncRNAs, Neat1 was identified to be significantly upregulated in lung metastases (Supplementary Fig. S5). Several studies have shown that NEAT1 is involved in regulating metastasis (45–49). To validate this result, we generated an independent cohort of primary sarcomas in KP mice and collected 22 primary sarcomas and a paired lung metastasis from each mouse and performed qRT-PCR to examine the expression of total Neat1 by using a probe targeting the common region of the two isoforms Neat1_1 and Neat1_2 (Fig. 2B, red arrow). The expression of Neat1_2 alone was quantified using a separate probe targeting a unique region (Fig. 2B, blue arrow). In this validation cohort, we confirmed that both total Neat1 and Neat1_2 alone were significantly upregulated in lung metastases compared with paired primary KP sarcomas (Fig. 2C). However, we further showed that the expression of Neat1 is unchanged between metastatic primary KP tumors and nonmetastatic primary KP tumors (Supplementary Fig. S6). Next, the expression of NEAT1 was examined in a total of 10 paired, primary STS tumors and lung metastases from patients diagnosed with UPS (n = 1) or myxofibrosarcomas (n = 9). These patient samples were obtained from the LTRI (Toronto, Ontario, Canada) and the Memorial Sloan Kettering Cancer Center (New York, NY). The results demonstrated that total NEAT1 was upregulated in a subset of lung metastases compared with paired primary tumors (Fig. 2D). Among these samples, five pairs of human UPS/myxofibrosarcomas and lung metastases from the LTRI (Toronto, Ontario, Canada) were submitted for RNA-seq and NEAT1 was found to be expressed higher in the lung metastases compared with the primary tumors (Supplementary Fig. S7). We note that some pairs of primary KP sarcomas and lung metastases did not show an increase in Neat1 and that the analysis of the small sample of paired human sarcomas failed to reach a P value of 0.05, which indicates heterogeneity in the mechanisms of sarcoma metastasis.

To further investigate the expression of NEAT1 in human UPS lung metastases, we next used RNA ISH. When the unpaired primary human UPS and lung metastases were analyzed, the expression of NEAT1 appeared to be similar between primary sarcomas and lung metastases from different patients (Supplementary Fig. S8). However, when paired primary UPS and lung metastases from the same patient were compared, the expression of total NEAT1 and NEAT1_2 were upregulated in a subset of the lung metastases (Fig. 3). In particular, the expression of total NEAT1 was significantly upregulated (one-sided unadjusted P = 0.05; median of differences = 1.9) in the paired lung metastases (Fig. 3).

To functionally test the role of Neat1 in sarcoma metastasis, we isolated two highly metastatic KP sarcoma cell lines, which were derived from lung metastases. We designed a pair of single guided RNAs (sgRNA) to target the promoter region of Neat1 based on RNA polymerase II chromatin immunoprecipitation (ChIP) sequencing in the University of California Santa Cruz (UCSC) genome browser (Fig. 4A; ref. 50). We then transduced a lentivirus that expresses Cas9 protein and two sgRNAs targeting the promoter region of Neat1 or control sgRNAs into the two different KP metastatic cell lines. Real time qPCR (Fig. 4B) and northern blot analysis (Fig. 4C) confirmed that total Neat1 expression was significantly reduced by using the CRISPR/Cas9 system to target the Neat1 promoter region in both transduced cell lines. Next, we injected the two KP metastatic cell lines with or without Neat1 knockdown into 5 nude mice per group through a tail vein injection. After 2 weeks, lung tissues were collected and stained with H & E. The tissue sections were then analyzed for the percentage of lung metastases. CRISPR/Cas9-mediated knockdown of Neat1 in both cell lines decreased the area of lung with colonization of sarcoma (Fig. 4D).

The short Neat1_1 isoform has been shown to be a critical modulator of prostate cancer progression and metastasis (51), leading us to evaluate whether depletion of Neat1_1 alone is sufficient to represses sarcoma lung colonization. Accordingly, we designed a new pair of sgRNAs (sgRNA-3 and sgRNA-4) that specifically target the short isoform Neat1_1 (Fig. 5A). The two sgRNAs were then electroporated, alongside another plasmid expressing Cas9, into a metastatic KP cell line. Following single-cell isolation, real time qPCR and northern blots were used to confirm the deletion of Neat1_1 (Fig. 5B). Two individual KP clones were confirmed to have the short isoform deleted. Of note, these clones expressed a truncated version of the long isoform Neat1_2. Having confirmed the deletion of the short isoform, we injected these Neat1_1-knockout clones #2 and #3, into nude mice via the tail vein. Nonedited KP cells that had been electroporated with Cas9 protein and a pair of nontargeted sgRNAs, were also injected into separate nude mice as a control. We confirmed that deletion of Neat1_1 region has no effect on cell proliferation in both cell lines (Supplementary Fig. S9). After 2 weeks, the respective lung tissues were collected, stained with H&E, and analyzed for the percentage of the area containing lung metastases. Remarkably, the deletion of the short isoform is sufficient to significantly suppress lung colonization (Fig. 5C).

Because Neat1 can be regulated by Trp53 (44, 52, 53), we next evaluated the effect of Neat1_1 depletion on lung metastasis from sarcomas in a Trp53 wild-type context. Sarcoma cell lines were derived from Kras^LSL-G12D; Ink4a/Arf^Flox/Flox (KI) mice, following intramuscular injection of adenovirus expressing Cre. Cre-mediated recombination activates mutant Kras^G12D and deletes both alleles of Ink4a/Arf (18), effectively inactivating oncogene-induced tumor suppression, yet maintaining wild-type Trp53. As described for the KP cell line, in the KI cell line we deleted the Neat1 short isoform using the CRISPR/Cas9 system. Deletion was confirmed using real time qPCR and northern blot assays (Fig. 5D). KI clones #6 and #10 showed Neat1_1 deletion and truncation of the long isoform Neat1_2. The edited KI cells were then assayed for lung colonization capacity by tail vein injection into nude mice and compared with nonedited KI control cells. After 2 weeks, lung tissues were collected and analyzed for area of lung colonization. Again, Neat1_1 depletion suppressed lung colonization (Fig. 5E). Together, these results suggest that regardless of Trp53 status, depletion of Neat1_1 region is sufficient to repress lung colonization.

To begin to investigate the underlying mechanisms by which Neat1_1 promotes lung metastasis, we performed an RNA pulldown assay combined with mass spectrometry to identify Neat1_1 interacting proteins. Neat1_1 sense and antisense RNA was in vitro transcribed and labeled with biotin, and then mixed with KP cell lysate (Fig. 6A). The Neat1_1–protein complexes were then pulled down and purified using streptavidin beads. Silver staining indicated that several unique proteins were selectively pulled down by Neat1_1-biotinylated sense RNA (Fig. 6B). The isolated Neat1_1--protein complexes were then analyzed by mass spectrometry. Following analysis of the Neat1_1 sense RNA interacting proteins (Supplementary Table S4), we discovered that numerous proteins that interact with Neat1_1 were associated with pathways regulating RNA splicing (Fig. 6C).

Among the proteins identified to interact with Neat1_1, alterations of far upstream element-binding protein 2 (KHSRP; Fig. 7A) and structural maintenance of chromosomes protein 2 (SMC-2; Supplementary Fig. S10A) were both previously associated with poor prognosis in STS (30, 31). We first used Neat1_1 RNA pulldown combined with Western blotting to confirm that KHSRP (Fig. 7B) and SMC2 (Supplementary Fig. S10B) specifically interact with the Neat1_1 sense RNA. Next, we tested the effect of KHSRP and SMC2 depletion on sarcoma lung colonization. We designed two individual sgRNAs targeting either Khsrp or Smc2, respectively, and delivered lentivirus to stably express each of the sgRNAs in two different KP cell lines. The protein expression of KHSRP (Fig. 7C) or SMC2 (Supplementary Fig. S10C) in their respective cell lines was significantly decreased following CRISPR/Cas9 editing. While decreased expression of SMC2 had little effect on lung colonization after tail vein injection (Supplementary Fig. S10D), decreased expression of KHSRP reduced lung colonization in both cell lines (Fig. 7D).

A previous study reported that KHSRP is a critical regulator of RNA splicing (54) and we showed that its depletion led to reduced lung colonization after tail vein injection (Fig. 7D). Therefore, we next examined whether RNA splicing pathways are dysregulated in mouse lung metastasis. By running pathway analysis on our initial RNA-seq data of paired primary sarcomas and lung metastases, we found that expression of genes involved in RNA splicing pathways is downregulated in lung metastases compared with primary sarcomas (Supplementary Tables S5 and S6). Interestingly, when the pathway analysis was performed in wild-type and Neat1-knockout MEFs (44), we observed that RNA splicing pathways are significantly upregulated in Neat1-knockout MEFs (Supplementary Fig. S11). Collectively, these results suggest that metastatic lesions may rely on a broader downregulation of the expression of genes involved in RNA splicing, and moreover, that one mechanism by which this occurs is via upregulation of Neat1.

The rarity and heterogeneity of STSs makes it challenging to identify genes that are critical for the metastatic process. In this study, we utilized GEMMs to investigate mis-regulation of gene expression programs in sarcoma metastases. Through RNA-seq of several paired primary KP sarcomas and lung metastases, we found that the lncRNA Neat1 is significantly upregulated in lung metastases compared with primary KP sarcomas (Supplementary Table S3). A relatively small sample size was used for RNA-seq studies (primary tumor, n = 6 and lung metastases, n = 14). Therefore, to validate our findings, we used an independent murine cohort of 22 paired primary KP tumors and lung metastases. However, in a small subset of lung metastases in mice, we detected downregulation of Neat1, which may reflect heterogeneity in the mechanisms of sarcoma metastasis (20) and alternative pathways to promote metastasis. Furthermore, we found that the expression of Neat1 is unchanged in metastatic primary KP tumors (n = 22) compared with nonmetastatic primary KP tumors (n = 17; Supplementary Fig. S6). These results suggest that NEAT1 may not be required for sarcoma cell invasion or intravasation, but rather critical for later steps in sarcoma metastasis, such as colonization (2).

RNA-seq and real time qPCR studies of five paired human primary UPS/myxofibrosarcomas and lung metastases from the LTRI (Toronto, Ontario, Canada) showed upregulation of NEAT1 in a subset of human lung metastases, while another five paired human primary UPS/myxofibrosarcomas and lung metastases from Memorial Sloan Kettering Cancer Center (New York, NY) demonstrated that NEAT1 is only upregulated in a subset of lung metastases from myxofibrosarcomas (n = 5). Furthermore, NEAT1 RNA ISH experiments in human UPS tissue microarrays also suggests a heterogeneous role of NEAT1 in sarcoma metastasis because only when compared with paired primary UPS, but not unpaired, samples was NEAT1 upregulated in lung metastases. In the future, additional analysis of the RNA-seq data of paired primary and lung metastases from KP mouse sarcomas as well as human sarcomas has the potential to identify additional genes that regulate metastasis.

To investigate whether the lncRNA Neat1 regulates sarcoma metastasis or is just a passenger in the metastatic process, we used CRISPR/Cas9 technology to decrease expression of Neat1 in highly metastatic mouse sarcoma cell lines. Deletion of the promoter of a noncoding gene without disrupting a large area in the genome has been reported to successfully knockout lncRNAs (23). We successfully knocked down the expression of total Neat1 in two KP sarcoma cell lines using a lentivirus to stably express Cas9 and a pair of sgRNAs that target a 1 kb region of the Neat1 promoter. Knockdown of Neat1 repressed sarcoma lung colonization after tail vein injection in the different KP cell lines tested. Furthermore, we designed another pair of sgRNAs that specifically targeted the short isoform of Neat1 in the mouse genome. We determined that knockout of the Neat1_1 locus significantly repressed sarcoma lung colonization regardless of Trp53 status in both KP and KI cell lines. Our findings suggest that Neat1, or at least the short isoform Neat1_1, may promote sarcoma lung metastasis via dysregulation of mRNA splicing. This could be further validated by rescuing Neat1_1 expression in Neat1_1-knockout cell lines in the future (Supplementary Fig. S12). However, our results cannot rule out the possibility that truncation of Neat1_2 also plays a role in suppressing sarcoma metastasis. It should be noted that in a small subset of lung metastases in mice, we detected downregulation of Neat1, which likely reflects that there are multiple molecular pathways that lead to sarcoma metastasis (20).

Understanding the mechanisms by which cancer cells adapt to a new distant microenvironment may lead to novel targeted therapies to prevent or treat metastasis. Accumulating evidence suggests that dysregulated RNA splicing impacts metastasis (16). Through alternative splicing of multiple genes, a cancer cell clone could acquire more aggressive phenotypes to survive and colonize a distant organ (16, 20). For instance, CD44 antigen (CD44) contains variable exons (exons v2–v10) that are alternatively spliced giving rise to two isoforms: CD44s (lacking exon v2–v10) and CD44v (including exon v2–v10; ref. 42). An aberrant switch from the expression of CD44s variant to the expression of the CD44v isoform contributes to breast cancer metastasis as well as to the plasticity of breast cancer stem cell phenotypes (55–58). Through a combination of Neat1 RNA pulldown and mass spectrometry, we identified that the majority of proteins that interact with Neat1 are involved in regulating RNA splicing pathways. Remarkably, RNA splicing pathways are significantly downregulated in lung metastases compared with paired KP primary tumors, but significantly upregulated in Neat1-knockout MEFs.

We further found that one protein interacting with Neat1 is the splicing regulator, KHSRP, whose alteration correlates with poor prognosis in patients with sarcoma (30, 31). KHSRP modulates RNA stability and gene expression by interacting with single-stranded RNA (57, 58) and mediates exon inclusion by forming a multiprotein complex that binds to G-U–rich introns (54). While other studies have previously suggested that KHSRP may be important for regulating metastasis (26, 59), here we have shown that KHSRP interacts with the lncRNA Neat1 leading us to propose that the Neat1–KHSRP complex may promote sarcoma metastasis by regulating RNA splicing.

In conclusion, we identified that the lncRNA Neat1 is upregulated in a subset of human and mouse sarcoma metastases. Through modulation of Neat1 expression, we determined that Neat1 can promote lung colonization of sarcoma cells. Furthermore, we found that the Neat1 interacting protein, Khsrp, which regulates splicing, also promotes KP sarcoma lung colonization. These results suggest that genes regulating RNA splicing, including Khsrp, are critical for a subset of KP sarcoma metastasis and should be further investigated as potential candidates for the prevention and treatment of sarcoma metastasis.

T.J. Robinson reports grants from Radiologic Society of North America (resident research grant award) during the conduct of the study. N. Gokgoz reports grants from FDC Foundation and McLaughlin Centre during the conduct of the study. I.L. Andrulis reports grants from FDC Foundation and McLaughlin Foundation during the conduct of the study. D.G. Kirsch reports grants from NCI (R35, during the conduct of the study), Merck (support for clinical trial), Eli Lilly (research support), and Bristol-Myers Squibb (research support, outside the submitted work) and other from Lumicell (member of scientific advisory board; stock, stock options; consultant; patents; royalties; and research support) and XRad Therapeutics (co-founder; stock; and research support). No potential conflicts of interest were disclosed by the other authors.

The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH.

J. Huang: Conceptualization, resources, validation, investigation, methodology, writing-original draft. M. Sachdeva: Conceptualization, methodology, writing-review and editing. E. Xu: Data curation, software, methodology, writing-review and editing. T.J. Robinson: Data curation, methodology, writing-review and editing. L. Luo: Methodology, writing-review and editing. Y. Ma: Methodology, writing-review and editing. N.T. Williams: Methodology, writing-review and editing. O. Lopez: Methodology, writing-review and editing. L.D. Cervia: Resources, data curation, methodology, writing-review and editing. F. Yuan: Resources, writing-review and editing. X. Qin: Formal analysis, writing-review and editing. D. Zhang: Formal analysis, writing-review and editing. K. Owzar: Formal analysis, writing-review and editing. N. Gokgoz: Resources, investigation, writing-review and editing. A. Seto: Data curation, writing-review and editing. T. Okada: Resources, investigation, writing-original draft. S. Singer: Resources, writing-review and editing. I.L. Andrulis: Resources, writing-review and editing. J.S. Wunder: Resources, writing-review and editing. A.J. Lazar: Resources, writing-review and editing. B.P. Rubin: Resources, writing-review and editing. K. Pipho: Resources, investigation, writing-review and editing. S.S. Mello: Resources, writing-review and editing. J. Giudice: Resources, writing-review and editing. D.G. Kirsch: Conceptualization, resources, supervision, funding acquisition, writing-original draft, project administration, writing-review and editing.

We thank the Duke University School of Medicine Proteomics and Metabolomics Shared Resource, which performed mass spectrometry and was supported by the Duke Cancer Center Support Grant (P30CA14236). This work was supported in part by the NCI of the U.S. NIH under awards R35CA197616 (to D.G. Kirsch) and P01CA142538 (to K. Owzar), the National Institute of General Medical Sciences under award R01GM130866 (to J. Giudice), the American Heart Association under award 19CDA34660248 (to J. Giudice), and by the Canada Foundation for Innovation/Ontario Research Fund, FDC Foundation, and McLaughlin Centre (to I.L. Andrulis and J.S. Wunder).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.