Long Noncoding RNA HOTAIR Regulates Polycomb-Dependent Chromatin Modification and Is Associated with Poor Prognosis in Colorectal Cancers

Noncoding RNAs (ncRNA) are found throughout the genome. However, the functions of ncRNAs are only partially understood. The function and clinical significance of short ncRNAs, such as miRNA and siRNA were elucidated first and then, long ncRNAs were reported more recently. Most long ncRNAs work with DNA-binding proteins, such as chromatin-modifying complexes, and epigenetically regulate the expression of multiple genes (1–3). Hox transcript antisense intergenic RNA (HOTAIR) is a long ncRNA that was identified from a custom tilling array of the HOXC locus (12q13.13; ref. 2). HOTAIR trimethylates histone H3 lysine-27 (H3K27me3) of the HOXD locus with the polycomb-repressive complex 2 (PRC2), which is composed of EZH2, SUZ12, and EED, and inhibits HOXD gene expression (2). Thus, HOTAIR epigenetically regulates HOXD expression, located on a different chromosome.

Gupta and colleagues reported that HOTAIR induced genome-wide retargeting of PRC2, leading to H3K27me3, and promoted metastasis of breast cancer by silencing multiple metastasis suppressor genes (4). In particular, they concluded that HOTAIR suppressed tumor suppressor genes such as HOXD10, PGR, and the protocadherin gene family in breast cancer cells. HOTAIR expression was low in normal breast epithelia but high in primary breast cancer as well as metastatic lesions. Moreover, breast cancer patients with high HOTAIR expression had a poorer prognosis for overall survival and for metastasis-free survival than did those with low HOTAIR expression.

Colorectal cancer (CRC) is the one of the most common cancers in the world. However, the existence of multiple known carcinogens and varying genetic backgrounds makes it difficult to determine which factors are most important in the development of CRC. Therefore, the identification of a bona fide molecule involved in progression of CRC has been greatly sought after. In the current study, we clarified the clinical significance of HOTAIR expression in CRC. Moreover, to determine the function of HOTAIR in CRCs, we used cDNA microarray data from another subset of 32 CRC samples obtained by laser microdissection (LMD). We conducted gene set enrichment analysis (GSEA) and investigated whether HOTAIR expression was highly correlated with previously curated gene expression signatures of PRC2 (4).

One hundred CRC samples (bulk samples) were obtained from patients who underwent surgery at the Medical Institute of Bioregulation Hospital, Kyushu University, between 1993 and 2000. Another 32 CRC samples (LMD samples) were obtained from Medical Institute of Bioregulation Hospital, Kyushu University, Tokyo Medical and Dental University Hospital, Kitazato University Hospital, National Cancer center, and National Defense Medical College Hospital. All specimens were immediately frozen in liquid nitrogen and stored at −80°C until RNA extraction. Written informed consent was obtained from all patients. No patient received chemotherapy or radiotherapy before surgery. The follow-up periods ranged from 2 months to 11 years, with a mean of 3 years. HEK293T, HCT116, and SW480 cell lines were provided by the American Type Culture Collection and were maintained in Dulbecco's Modified Eagle's Media, McCoy 5A, or RPMI-1640, respectively, containing 10% FBS with 100 units/mL penicillin and 100 μg/mL streptomycin and cultured in a humidified 5% CO₂ incubator at 37°C.

Total RNAs from frozen CRC samples were extracted using ISOGEN (Nippon Gene) following the manufacturer's protocol.

Total RNAs were extracted using QIAamp DNA Micro Kit (Qiagen) following the manufacturer's protocol.

As previously reported, cDNAs from all samples were synthesized from 8.0 μg of total RNA (5). HOTAIR levels were quantified using LightCycler 480 Probes Master kit (Roche Applied Science) following the manufacturer's protocol with the following specific HOTAIR primers (forward, 5′-CAGTGGGGAACTCTGACTCG-3′ and reverse, 5′-GTGCCTGGTGCTCTCTTACC-3′). HOTAIR levels were normalized to GAPDH (forward, 5′-GTCAACGGATTTGGTCTGTATT-3′ and reverse, 5′-AGTCTTCTGGGTGGCAGTGAT-3′).

RNAs from another 32 CRC tissues were collected for LMD. CRC tissues were microdissected using the LMD System (Leica Microsystems) as previously described (6).

HOTAIR/GAPDH levels in 32 CRC tissues (LMD samples) were measured by quantitative real-time PCR. Gene expression profiles of 32 CRC samples were measured by Agilent Whole Human Genome Microarray 4 × 44K G4112F and analyzed by GSEA (7). The expression profiles were quintile normalized. The batch effect in microarray experiments was also adjusted by an empirical Bayesian approach (8). To collapse each probe set on the array to a single gene, the probe with the highest variance among multiple probes that corresponded to the same gene was selected, which produced a 19,749 (genes) × 32 (CRCs) expression matrix. For GSEA, HOTAIR expression was treated as a binary variable divided into low or high HOTAIR expression by a criterion of whether or not its value was greater than 0.273. As a result, 9 CRC samples were categorized as high and 23 were labeled as low, of 32 tumors. For functional gene sets for GSEA, we used gene sets from global occupancy of H3K27me3, PRC2 subunits, SUZ12 or EZH2, and PRC2 (all of H3K27me3, SUZ12, and EZH2) induced by HOTAIR overexpression in MDA-MB-231 breast cancer cells (4). As a metric for ranking genes in GSEA, the difference between the means of samples with low and high HOTAIR expression was used, and the other parameters were set by their default values. Gene expression arrays have been deposited in the National Center for Biotechnology Information (NCBI) Gene Expression Omnibus (GEO) database with accession code GSE21815.

To generate HOTAIR expression lentiviral vector, we amplified insert (full-length human HOTAIR) by PCR from MCF7 cDNA. Lentiviruses were produced by transient transfection of HEK293T cells with pCMV-VSV-G-RSV-Rev, pCAG-HIVgp, and either CSII-CMV-HOTAIR or CSII-CMV-MCS (empty) plasmid DNAs (5′XhoI and 3′NotI site) plus Lipofectamine 2000 (Invitrogen) following the manufacturer's protocol. Forty-eight hours after cotransfection, the lentivirus-containing supernatant was collected and passed through a 0.45-μm filter. The titer of the lentivirus vector in filtered supernatants was estimated by measuring the concentration of HIV p24 gag antigen with an ELISA kit (Perkin-Elmer Life Science).

Two individual siRNAs (siRNA HOTAIR1 and siRNA HOTAIR2) and negative control siRNA (silencer negative control siRNA) are purchased from Ambion. siRNA oligonucleotides (10 nmol/L) in Opti-MEM (Invitrogen) were transfected into SW480 cells using Lipofectamine RNAiMAX (Invitrogen) following the manufacturer's protocol. Forty-eight hours posttransfection, HOTAIR expression levels were measured, and Matrigel invasion assays were conducted. Target sequences for HOTAIR siRNAs were as follows: siRNA1, 5′-UUUUCUACCAGGUCGGUAC-3′ and siRNA2, 5′-AAUUCUUAAAUUGGGCUGG-3′.

The Matrigel invasion assay was done using the BD Biocoat Matrigel Invasion Chamber (pore size: 8 μm, 24-well; BD Biosciences) following the manufacturer's protocol. Cells (5 × 10⁴) were plated in the upper chamber in serum-free medium. The bottom chamber contained medium with 10% FBS. After 48 hours, the bottom of the chamber insert was stained with Calcein AM (Invitrogen). The cells that had invaded through the membrane to the lower surface were evaluated in a fluorescence plate reader at excitation/emission wavelengths of 485/530 nm. Each Matrigel invasion assay was conducted in at least 3 replicates.

The significance of differences between 2 groups was estimated with the Student t test and χ² test. Overall survival curves were plotted according to the Kaplan–Meier method, with the log-rank test applied for comparison. Variables with a value of P < 0.05 by univariate analysis were used in subsequent multivariate analysis on the basis of Cox proportional hazards model. All differences were considered statistically significant at the level of P < 0.05. Statistical analyses were done using JMP 5 (SAS Institute).

HOTAIR expression levels in 100 cancerous and noncancerous tissues were examined by quantitative real-time PCR. HOTAIR levels in cancerous tissues were significantly lower than those in the noncancerous tissues (P = 0.002; Fig. 1A). We divided the 100 patients with CRC into a high HOTAIR expression group (n = 20) and a low expression group (n = 80), according to a HOTAIR/GAPDH ratio of 0.273 in cancerous tissue (Fig. 1B). Clinicopathologic factors were analyzed in the high and low HOTAIR expression groups (Table 1). The high HOTAIR expression group (n = 20) showed a less differentiated histology, greater tumor depth, and liver metastasis than the low HOTAIR expression group (n = 80; P < 0.05). In particular, high HOTAIR expression was strongly associated with liver metastasis (P = 0.006). With regard to overall survival, patients with high HOTAIR expression had a significantly poorer prognosis than those with low HOTAIR expression (P = 0.0046; Fig. 1C). Univariate analysis of overall survival revealed that the relative level of HOTAIR expression, histologic grade, depth of tumor, lymph node metastasis, lymphatic invasion, and venous invasion were prognostic indicators (Table 2). Variables with a value of P < 0.05 were selected for multivariate analysis. Multivariate analysis showed that HOTAIR expression was an independent prognostic indicator for overall survival in patients with CRC (relative risk: 5.62, P = 0.008; Table 2). Moreover, we asked whether high HOTAIR expression contributed to poor prognosis in another independent subgroup. Using exon array data from 320 patients with CRC, we divided the cases according to the median, yielding a high HOTAIR expression group (n = 160) and a low expression group (n = 160). The patients in the high HOTAIR expression group had a significantly poorer prognosis than those in the low expression group (P < 0.0001; Supplementary Fig. S1).

Next, we used a cDNA array on the basis of cancer tissue samples from 32 patients with CRC obtained by LMD and asked whether HOTAIR expression levels in the 32 CRC samples were highly correlated with previously curated gene expression signatures of PRC2 (4). First, we divided 32 patients with CRC into a high HOTAIR expression group (n = 9) and a low expression group (n = 23), according to the HOTAIR/GAPDH ratio of 0.273 as in Fig. 1B. We expected that CRC with high and low HOTAIR expression levels would be significantly enriched for these gene sets, as HOTAIR overexpression induced localization of H3K27me3 and PRC2 subunits, SUZ12 and EZH2, in MDA-MB-231 breast cancer cells (4). Indeed, gene signatures with HOTAIR-induced SUZ12 occupancy [P = 0.001 and false discovery rate (FDR) = 0.004], EZH2 occupancy (P = 0.029 and FDR = 0.044), H3K27me3 occupancy (P = 0.028 and FDR = 0.032), and PRC2 occupancy (occupancy of SUZ12, EZH2, and H3K27me3; P = 0.032 and FDR = 0.024) were confirmed to be significantly enriched in CRC (Fig. 2A and B). Moreover, these enriched genes were expressed at lower levels in tumors with high HOTAIR expression than in those with low HOTAIR expression (Supplementary Fig. S2). These results suggest that HOTAIR expression induced genome-wide retargeting of PRC2 not only in breast cancer but also in CRC.

Finally, we examined the effect of HOTAIR in CRC cells. We determined HOTAIR levels in CRC cell lines by quantitative real-time PCR (Supplementary Fig. S3A). We generated lentiviral HOTAIR expression vectors which were transduced to HCT116, a CRC cell line (Supplementary Fig. S3B). HOTAIR overexpression in HCT116 significantly promoted invasion in Matrigel (P < 0.05; Fig. 3A). Conversely, suppression of HOTAIR in SW480 CRC cells (that expressed endogenous HOTAIR) with specific siRNAs decreased invasion in Matrigel (P < 0.05; Fig. 3B and Supplementary Fig. S3C).

In our current study, we found that HOTAIR expression levels in CRC tissues were higher than those in corresponding noncancerous tissues. Moreover, high HOTAIR expression in CRC tissues was associated with a poorer prognosis. As for clinicopathologic variables, HOTAIR expression levels were intimately linked to liver metastasis. Recently, Gupta and colleagues showed that HOTAIR expression was associated with metastasis of breast cancer (4). Therefore, we speculated that HOTAIR expression was also associated with metastasis in CRC. GSEA based on cDNA microarray data showed that HOTAIR expression was significantly associated with genome-wide retargeting of PRC2 genes as shown in breast cancer by Gupta and colleagues (4). Using in vitro data, we showed that HOTAIR overexpression increased the invasiveness of CRC cells. These results indicate that HOTAIR might also play a role in promoting metastasis of CRC.

In recent years, many long ncRNAs have been identified and their involvement in human disease has been reported. Long ncRNAs, such as lincRNA-p21 (lung cancer; ref. 9), uc.73 (CRC; ref. 10), and uc.338 (hepatocellular carcinoma; ref. 11), have been associated with human malignancies. However, the detailed function and the clinical significance of the long ncRNAs have not yet been elucidated. This is the first report using clinical CRC samples to show that HOTAIR works in cooperation with the PRC2. It was previously shown that the PRC2 bound to the 5′ terminus of HOTAIR and trimethylated H3K27 (12). Also, the LSD1/CoREST/REST bound to the 3′ terminus of HOTAIR, which demethylated H3K4 (12). Thus, the modifications of those DNA-binding proteins by HOTAIR regulate global gene expression.

EZH2 and SUZ12, the components of PRC2, are overexpressed in several cancers (13, 14). In particular, SUZ12 is reportedly overexpressed in CRC (15–17). Thus, it is interesting that the current study identified an association between HOTAIR and SUZ12 in cancer-specific GSEA. Moreover, PRC2-targeted genes were identical to gene sets that were silenced by pluripotent stem cell–related transcription factors such as Oct4, Sox2, and Nanog (18, 19). Therefore, HOTAIR overexpression in CRC might be associated with multipotent differentiation of CRC cells. Gene pathway analysis also indicated that HOTAIR-regulated gene sets included CDH1 (E-cadherin) target genes (data not shown), whose expression is lost in metastatic cancer cells of the mesenchymal phenotype. Therefore, we suggest that HOTAIR might maintain mesenchymal and undifferentiated cancer cells in cooperation with the PRC2. The strong correlation between HOTAIR expression and liver metastasis might indicate an important role of HOTAIR in the proliferation of mesenchymal and undifferentiated cancer cells, which enhances the metastatic ability of CRC.

In conclusion, HOTAIR regulates expression of multiple genes in cooperation with PRC2 and is a novel molecule involved in the progression of CRC. HOTAIR might increase the number of undifferentiated cancer cells and contribute globally to cancer metastasis.

No potential conflicts of interest were disclosed.

The authors thank T. Shimooka, K. Ogata, and M. Kasagi for their technical assistance and H. Miyoshi (RIKEN BioResource Center) for providing lentivirus vector plasmid DNA.

This work was supported in part by the following grants and foundations: CREST, Japan Science and Technology Agency (JST); Japan Society for the Promotion of Science (JSPS) Grant-in-Aid for Scientific Research, grant numbers 20390360, 20591547, 20790960, 21591644, 21791295, 21791297, 215921014, and 21679006; the Funding Program for Next Generation World-Leading Researchers (LS094); NEDO (New Energy and Industrial Technology Development Organization) Technological Development for Chromosome Analysis; and Grant-in-Aid from the Tokyo Biochemical Research Foundation.

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.