HOTAIR, a cell cycle–associated long noncoding RNA and a strong predictor of survival, is preferentially expressed in classical and mesenchymal glioma
NEURO-ONCOLOGY † D E C E M B E R
HOTAIR, a cell cycle - associated long noncoding RNA and a strong predictor of survival, is preferentially expressed in classical and mesenchymal glioma
Jun-Xia Zhang 0 2 3 4 5 6 7
Lei Han 0 2 3 4 5 6 7
Zhao-Shi Bao 0 2 3 4 5 6 7
Ying-Yi Wang 0 2 3 4 5 6 7
Lu-Yue Chen 0 2 3 4 5 6 7
Wei Yan 0 2 3 4 5 6 7
Shi-Zhu Yu 0 2 3 4 5 6 7
Pei-Yu Pu 0 2 3 4 5 6 7
Ning Liu 0 2 3 4 5 6 7
Yong-Ping You 0 2 3 4 5 6 7
Tao Jiang 0 2 3 4 5 6 7
Chun-Sheng Kang 0 2 3 4 5 6 7
for the Chinese Glioma Cooperative Group 0 2 3 4 5 6 7
0 System, Ministry of Education, Tianjin Key Laboratory of Injuries, Variations and Regeneration of Nervous System
1 The Author(s) 2013. Published by Oxford University Press on behalf of the Society for Neuro-Oncology. All rights reserved. For permissions , please
2 Tianjin Neurological Institute, Key Laboratory of Post-trauma Neuro-repair and Regeneration in Central Nervous
3 Department of Neurosurgery, Tianjin Medical University General Hospital, Laboratory of Neuro-Oncology
4 T.J.); Department of Neuropathology, Tianjin Neurological Institute, Tianjin Medical University General Hospital
5 Department of Neurosurgery, Beijing Tiantan Hospital, Capital Medical University , Beijing, China (Z.-S.B., W.Y.
6 Nanjing Medical University , Nanjing, China (J.-X.Z., Y.-Y.W., N.L., Y.-P.Y.); Beijing Neurosurgical Institute
7 Tianjin , China (J.-X.Z., L.H., L.-Y.C., P.-Y.P. , C.-S.K.); Department of Neurosurgery, The First Affiliated Hospital of
Tianjin, China (S.-Z.Y.)
cell cycle; glioma; HOTAIR; molecular subtype; survival
Background. Long noncoding RNA Hox transcript
antisense intergenic RNA (HOTAIR) has been characterized
as a negative prognostic factor in breast and colon
cancer patients. The clinical significance and function of
HOTAIR in glioma remains unclear.
Methods. We analyzed the clinical significance of
HOTAIR in 3 different glioma cohorts with gene
expression data, including correlation with tumor grade,
prognosis, and molecular subtype. The function of HOTAIR
in glioma was explored by performing gene set
enrichment analysis and in vitro and in vivo experiments.
Results. HOTAIR expression was closely associated
with glioma grade and poor prognosis. Multivariate
Cox regression analysis revealed that HOTAIR was an
independent prognostic factor in glioblastoma multiforme
†These authors contributed equally to this paper.
patients. HOTAIR expression correlated with glioma
molecular subtype, including those of The Cancer Genome
Atlas. HOTAIR was preferentially expressed in the
classical and mesenchymal subtypes compared with the neural
and proneural subtypes. A gene set enrichment analysis
designed to show gene set differences between patients
with high and low HOTAIR expression indicated that
HOTAIR expression was associated with gene sets
involved in cell cycle progression. HOTAIR reduction
induced colony formation suppression, cell cycle G0/
G1 arrest, and orthotopic tumor growth inhibition.
Conclusion. Our data establish that HOTAIR is an
important long noncoding RNA that primarily serves as a
prognostic factor for glioma patient survival, as well as
a biomarker for identifying glioma molecular subtypes,
a critical regulator of cell cycle progression.
Oploring novel, noncoding RNAs (ncRNAs) to
ver the last few decades, researchers have been
excharacterize their potential roles in biological
processes and disease development.1 – 4 The human
genome includes a number of ncRNAs, such as
microRNAs (miRNAs), long noncoding RNAs (long
ncRNAs, lncRNAs), Piwi-interacting RNAs, and small
nucleolar RNAs. LncRNAs are nonprotein coding
transcripts longer than 200 nucleotides and are implicated
in a number of important events, such as epigenetic
regulation, transcriptional regulation, and
posttranscriptional regulation.5,6 LncRNAs exhibit unique profiles in
various human cancers, which reflect disease progression
and serve as a predictor of patient outcomes.7 – 9 It was
recently discovered that lncRNAs function in various
aspects of cell biology and can potentially contribute to
Gliomas are the most frequent primary tumors in the
brain. Despite recent advances in cancer treatment, this
statistic has not changed significantly.10,11 Therefore, it
is essential to investigate the mechanism involved in the
development and progression of glioma. In the current
study, we profiled the lncRNA Hox transcript antisense
intergenic RNA (HOTAIR), which was closely correlated
with tumor grade, poor prognosis, and molecular subtype
in glioma. Thus, we clarified the clinical significance and
function of HOTAIR in glioma by analyzing clinical
and molecular pathology features and performing gene
set enrichment analysis (GSEA) and in vitro and in vivo
Materials and Methods
Patients and Samples
In total, 295 glioma samples from the Chinese Glioma
Genome Atlas (CGGA, http://www.cgcg.org.cn/) were
included in this study. The first cohort (CGGA1) comprised
58 astrocytomas, 17 oligodendrogliomas, 22
oligoastrocytomas (OAs), 8 anaplastic astrocytomas (AAs), 11
anaplastic oligodendrogliomas (AOs), 15 anaplastic
oligoastrocytomas (AOAs), 89 tumors of glioblastoma
multiforme (GBM), and 5 normal brain tissue samples. The
second cohort (CGGA2) comprised 8 astrocytomas, 10
oligodendrogliomas, 8 OAs, 7 AAs, 5 AOs, 3 AOAs, and 34
GBM. Anaplastic glioma (AG) includes AA, AO, and
AOA. High-grade glioma (HGG), includes AG and
GBM. Further information is described in the
Supplementary material. Glioma gene expression datasets are
deposited at the Repository of Molecular Brain Neoplasia
Data (REMBRANDT; http://caintegrator.nci.nih.gov/
rembrandt/) and the Gene Expression Omnibus Web
site (http://www.ncbi.nlm.nih.gov/geo/, accession nos.
GSE4290 and GSE7181).
Described previously12 have been whole genome gene
profiling, pyrophosphate sequencing for the isocitrate
dehydrogenase 1(IDH1) mutation (IDH1R132),
pyrophosphate sequencing for O6-DNA
methylguaninemethyltransferase (MGMT) promoter methylation, and
Gene Set Enrichment Analysis With HOTAIR
The gene expression profiles of GBM samples from
CGGA1 were analyzed by GSEA.13 For GSEA,
HOTAIR expression was treated as a binary variable
divided into low or high HOTAIR expression by a
criterion of whether the value was greater than the median.
To determine functional gene sets for GSEA, we used
gene sets from an analysis of global occupancy of
H3K27me3 and enhancer of zeste homolog 2 (EZH2)
induced by HOTAIR overexpression in MDA-MB-231
breast cancer cells.14 As a metric for ranking genes in
the 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.
Cell culture and transfection, colony formation assay, cell
cycle analysis, and western blot analysis are described in
the Supplementary material.
Orthotopic Glioma Model and Treatment
Bagg albino (BALB)/c nude mice at 4 weeks of age were
purchased from the Animal Center of the Cancer
Institute at the Chinese Academy of Medical Science. To
establish intracranial gliomas, 0.5 × 105 U87 cells were
transduced with luciferase lentivirus of small interfering
(si)HOTAIR and then implanted stereotactically. Mice
were imaged for Fluc activity using bioluminescence
imaging on days 1, 10, 20, and 30.
Kaplan – Meier survival analysis was used to estimate the
survival distributions, and the log-rank test was used to
assess the statistical significance between stratified
survival groups using the median value as the cutoff. Cox
proportional hazards regression analyses were performed
using SPSS software for Windows. Pearson correlation
was used to determine significant differences. One-way
ANOVA was used to test for differences among at least
3 groups, and a least significant difference post-hoc test
was used to obtain individual P values followed by
ANOVA. The t test was used to determine differences in
each 2-group comparison. All data are presented as
mean + standard error. A 2-sided P value of ,.05 was
regarded as significant.
HOTAIR Expression Correlates With Glioma Grade
First, we analyzed HOTAIR expression level in whole
genome gene profiling of 220 glioma and 5 normal
tissues. HOTAIR expression was significantly higher in
HGG than in low-grade glioma (LGG; P , .001).
Moreover, as shown in Fig. 1A, GBM demonstrated a
significant increase in HOTAIR transcript levels, compared
with that observed in normal tissues (P ¼ .093), LGGs
(P , .001), or AGs (P ¼ .011). No significant difference
in HOTAIR expression levels was observed between LGG
and AG (P ¼ .326). Next, we employed 2 independent
glioma gene expression datasets (REMBRANDT and
GSE4290) to examine the association between HOTAIR
expression levels and glioma grade (Supplementary Fig.
S1A). One-way ANOVA showed that HOTAIR was
significantly associated with tumor grade (P ¼ .002 and P ¼
.001 for REMBRANDT data and GSE4290 data,
respectively), which was consistent with the CGGA1 data.
These findings suggest that HOTAIR may play an
important role in glioma development.
HOTAIR Overexpression Confers a Poor Prognosis in Glioma Patients
Next, we investigated the correlation between HOTAIR
expression and overall survival using Kaplan – Meier
survival curve analysis with a log-rank comparison.
HGG samples expressing higher than median levels of
HOTAIR were associated with decreased survival relative
to those with HOTAIR levels lower than the median
(P ¼ .0031) in the CGGA1 data (Fig. 1B). Further,
HOTAIR expression was inversely correlated with
overall survival in AG (P ¼ .0284) and GBM (P ¼
.0077) (Fig. 1C and D), and similar results were detected
in the REMBRANDT data (Supplementary Fig. S1B).
Highly statistically significant correlations were observed
between overall survival and the expression levels of
HOTAIR (P , .0001 for HGG; P ¼ .0091 for AG);
however, the P value for GBM (P ¼ .0759) did not
reach statistical significance. To further confirm these
results, we performed microarray analysis to examine
HOTAIR levels in another independent cohort of Chinese
glioma (CGGA2). As shown in Fig. 1E and F, HOTAIR
expression was significantly increased in HGG compared
with LGG (P , .001), and cases of GBM that were highly
positive for HOTAIR had a markedly worse outcome
(P ¼ .0088). Overall, these data indicate that HOTAIR
overexpression correlates with a significantly worse
HOTAIR Is an Independent Prognostic Factor in GBM Patients
High expression of HOTAIR was associated with older
age at diagnosis (P ¼ .012), nonmutated IDH1 (P ,
.001), unmethylated MGMT promoter (P ¼ .027), and
high expression of epidermal growth factor receptor
(EGFR; P ¼ .005) (Table 1). Next, we conducted
univariate Cox regression analysis using clinical and genetic
variables for 89 GBM patients from the CGGA1 cohort and
found that high expression of HOTAIR, high KPS score,
and total resection were statistically associated with
overall survival, while IDH1 mutation and MGMT
promoter methylation were not associated with overall
survival (Table 2). Then we evaluated the factors that
contributed to overall survival using a multivariate Cox
proportional hazards model. The analysis revealed that
HOTAIR expression, KPS score, and total resection
correlated independently with overall survival (hazard ratio
[HR] ¼ 2.933, P ¼ .005; HR ¼ 0.508, P ¼ .048; HR ¼
0.416, P ¼ .034, respectively) when considering gender,
Ki-67, EGFR, proliferating cell nuclear antigen (PCNA),
topoisomerase II, and glutathione S-transferase (GST) –
p expression (P , .3, univariate Cox regression analysis).
HOTAIR Is a Marker for Glioma Molecular Subtype
The Cancer Genome Atlas (TCGA) network described a
robust gene expression – based molecular classification
of GBM into classical, mesenchymal, neural, and
proneural subtypes.15 We applied the TCGA classification
system to the CGGA1, REMBRANDT, and GSE4290
data and annotated the samples according to the 4
TCGA subtypes using the prediction analysis of
microarrays classifier. One-way ANOVA indicated a markedly
significant difference in HOTAIR expression between
the 4 glioma subtypes in the CGGA1 datasets. In
particular, HOTAIR expression in the classical and
mesenchymal subtypes was higher than the expression in the
neural and proneural subtypes (Fig. 2A). Recently, we
proposed a new classification system containing 2 major
subtypes: mesenchymal-like (containing classical and
mesenchymal subtypes) and proneural-like (containing
neural and proneural subtypes), which is primarily
based on the treatment efficacy of temozolomide in the
different subtypes.12 Further, we found that HOTAIR
expression in the mesenchymal-like subtype was
significantly higher than in the proneural-like subtype (P ,
.001; Fig. 2B). Moreover, our recent study has identified
3 major groups of gliomas from 220 CGGA1 samples
(referred to as G1, G2, and G3) that have distinctly
different clinical prognoses and molecular characteristics.16
HOTAIR expression in the G3 subtype was statistically
higher than that in the G1 or G2 subtype (Fig. 2C). A
similar trend was observed in the REMBRANDT and
GSE4290 data (Fig. 2).
HOTAIR regulates Polycomb repressive complex 2
(PRC2) – dependent histone H3 lysine 27 trimethylation
and gene silencing in breast and colorectal cancers.14,17
Because EZH2, suppressor of zeste 12 homolog
(SUZ12), and embryonic ectoderm development (EED)
are the core components of PRC2,18 we first explored
their expression profile in CGGA1. The level of EZH2
Abbreviations: TOPO II, topoisomerase II; PTEN, phosphatase and
Abbreviations: TOPO II, topoisomerase II; PTEN, phosphatase and
expression was the most significantly associated with
glioma grade (Fig. 3A). Then, we tested whether
HOTAIR expression levels in 89 GBM samples from
CGGA1 were highly correlated with the previously
identified gene expression signatures of EZH2. Indeed, gene
signatures with HOTAIR-induced EZH2 occupancy
(P ¼ .134 and false discovery rate [FDR] ¼ 0.141) and
H3K27me3 occupancy (P ¼ .112 and FDR ¼ 0.117)
were not significantly enriched in GBM (Fig. 3B), which
is incompatible with previous data that demonstrated
that HOTAIR expression induced genome-wide
retargeting of PRC2 in breast cancers and colorectal cancers.14,17
Inconsistent results were also obtained from the GBM
data in GSE4290 (Supplementary Fig. S2). These results
indicate that HOTAIR may regulate gene expression
patterns in an EZH2-independent manner in glioma.
HOTAIR Is a Cell Cycle – Associated Long
To identify the mechanism of HOTAIR involvement
in glioma, we first screened differentially expressed
genes in HGG compared with LGG and found 1928
upregulated genes and 815 downregulated genes in HGG.
Further coexpression analysis of these 2743 genes
revealed that 1400 genes and 637 genes were positively
and negatively, respectively, correlated with HOTAIR
expression (P , .05; Fig. 4A). These genes will be called
“HOTAIR-associated genes.” GSEA was used to evaluate
the pathways that were differentially expressed between
patients with high levels of HOTAIR expression and
those with low levels of HOTAIR expression. GSEA
analysis revealed that HOTAIR regulates genes primarily
associated with cell cycle progression (Fig. 4B).
Next, we used a colony formation assay and evaluated
cell cycle distribution to determine the role of HOTAIR in
glioma cell proliferation. As shown in Fig. 5A, LN229 and
U87 cells exhibited a significant reduction in colony
formation after 2 weeks of siHOTAIR treatment compared
with the control group. In addition, a reduction in
HOTAIR expression resulted in a significant increase in
cells in the G0/G1 phase in both LN229 and U87 cell
lines (Fig. 5B). Next, we explored expression changes in
cell cycle proteins. Western blot assays revealed that
HOTAIR knockdown triggered a reduction in cyclin
D1, cyclin E, cyclin-dependent kinase (CDK)4, CDK2,
and E2F1 expression and an induction of p21 and p16
expression (Fig. 5C).
Further, we employed a U87 orthotopic glioma model
to detect the in vivo function of HOTAIR. Statistically
significant difference in tumor volume indicated by
bioluminescence imaging appeared between the control
and siHOTAIR treated group on day 10 (Fig. 5E).
Representative images of mice implanted with
intracranial tumors are shown in Fig. 5D.
HOTAIR has been characterized as a negative prognostic
factor in liver, colon, and laryngeal squamous cancer
patients.17,19,20 In our study, HOTAIR was identified as a
critical marker not only for tumor grade and outcome
but also for molecular subtype in glioma. HOTAIR
expression was low in LGG but high in HGG samples.
Glioma patients with high HOTAIR expression had a
poorer prognosis for overall survival than did those with
low HOTAIR expression. Multivariate Cox regression
analysis revealed that HOTAIR was an independent
prognostic factor in GBM patients. Our further data indicated
that HOTAIR expression in the classical or mesenchymal
subtype was higher than in the neural or proneural
subtype, which was consistent with HOTAIR expression
status in the mesenchymal-like and proneural-like
subtypes. This suggests that HOTAIR plays an important
role in glioma molecular classification and may serve as
a novel therapeutic target for classical and mesenchymal
Gupta and colleagues14 reported that
HOTAIRinduced genome-wide retargeting of PRC2, which is
composed of EZH2, SUZ12, and EED, led to H3K27me3
modification and promoted metastasis of breast cancer
by silencing multiple metastasis-suppressing genes.
Additional data that used GSEA also showed that
HOTAIR expression induced genome-wide retargeting
of PRC2 in colon cancer.17 However, GSEA in our
study suggested that HOTAIR may regulate gene
expression in an EZH2-independent manner in glioma.
Recent data suggest that HOTAIR preferentially occupies
a GA-rich DNA motif to nucleate broad domains of
Polycomb occupancy and H3K27me3 and that HOTAIR
occupancy occurs independently of PRC2.21 These data
suggest that HOTAIR-dependent gene regulation in
glioma cells is complex and differs significantly from the
reports of its activity in other cancer cells.
HOTAIR has been well documented to play a role in
the invasion of breast, colon, and liver cancer cells.
Another recent study showed that HOTAIR knockdown
not only inhibited cell invasion but also decreased
cell proliferation, altered cell cycle progression, and
induced apoptosis in pancreatic cancer cells.22 In our
study, GSEA demonstrated gene set differences between
HOTAIR high- versus low-expressing patients, which
indicated that HOTAIR regulates genes involved in cell
cycle progression. Additionally, a reduction in HOTAIR
expression induced cell cycle G0/G1 arrest and was
accompanied by changes in the expression of cell cycle –
In summary, we show that HOTAIR is a favorable
factor for malignant progression and poor prognosis in
glioma patients and exhibits pro-oncogenic activity by
modulating the cell cycle. To our knowledge, this is the
first study that used clinical glioma samples to
demonstrate the clinical signature and biological function of
HOTAIR in glioma. Therefore, understanding the
regulation of HOTAIR expression in glioma could lead to
the identification of new therapeutic targets for treating
glioma and warrants further investigation.
Supplementary material is available online at
This work was supported financially by grants from the
National High Technology Research and Development
Program 863 (2012AA02A508), the International
Science and Technology Cooperation Program of
China (2012DFA30470), the National Natural Science
Foundation of China (91229121, 81272792, 81172406,
81101901), the Jiangsu Province’s Key Discipline
of Medicine (XK201117), the Key Foundation of
Jiangsu Province Science and Technology
Commission (BL2012028), Provincial Initiative Program for
Excellency Disciplines, Jiangsu Province, and the
Jiangsu Province’s Medical Major Talent program
(RC2011051), Natural Science Foundation of Tianjin
Municipal Science and Technology Commission
Conflict of interest statement. None declared.
1. Sana J , Faltejskova P , Svoboda M , Slaby O . Novel classes of non-coding RNAs and cancer . J Transl Med . 2012 ; 10 : 103 .
2. Cano A , Nieto MA . Non-coding RNAs take centre stage in epithelial-tomesenchymal transition . Trends Cell Biol . 2008 ; 18 : 357 - 359 .
3. Huang K , Zhang JX , Han L , et al. MicroRNA roles in beta-catenin pathway . Mol Cancer . 2010 ; 9 : 252 .
4. Wang WX , Huang Q , Hu Y , Stromberg AJ , Nelson PT . Patterns of microRNA expression in normal and early Alzheimer's disease human temporal cortex: white matter versus gray matter . Acta Neuropathol . 2011 ; 121 : 193 - 205 .
5. Mercer TR , Dinger ME , Mattick JS . Long non-coding RNAs: insights into functions . Nat Rev Genet . 2009 ; 10 : 155 - 159 .
6. Zhang J , Zhang A , Wang Y , et al. New insights into the roles of ncRNA in the STAT3 pathway . Future Oncol . 2012 ; 8 : 723 - 730 .
7. Gibb EA , Brown CJ , Lam WL . The functional role of long non-coding RNA in human carcinomas . Mol Cancer . 2011 ; 10 : 38 .
8. Wapinski O , Chang HY . Long noncoding RNAs and human disease . Trends Cell Biol . 2011 ; 21 : 354 - 361 .
9. Han L , Zhang K , Shi Z , et al. LncRNA profile of glioblastoma reveals the potential role of lncRNAs in contributing to glioblastoma pathogenesis . Int J Oncol . 2012 ; 40 : 2004 - 2012 .
10. Wang Y , Jiang T. Understanding high grade glioma: molecular mechanism, therapy and comprehensive management . Cancer Lett . 2013 ; 331 : 139 - 146 .
11. Hegi ME , Janzer RC , Lambiv WL , et al. Presence of an oligodendrogliomalike component in newly diagnosed glioblastoma identifies a pathogenetically heterogeneous subgroup and lacks prognostic value: central pathology review of the EORTC_26981/NCIC_CE.3 trial . Acta Neuropathol . 2012 ; 123 : 841 - 852 .
12. Zhang JX , Zhang J , Yan W , et al. Unique genome-wide map of TCF4 and STAT3 targets using ChIP-seq reveals their association with new molecular subtypes of glioblastoma . Neuro Oncol . 2013 ; 15 : 279 - 289 .
13. Subramanian A , Tamayo P , Mootha VK , et al. Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles . Proc Natl Acad Sci U S A . 2005 ; 102 : 15545 - 15550 .
14. Gupta RA , Shah N , Wang KC , et al. Long non-coding RNA HOTAIR reprograms chromatin state to promote cancer metastasis . Nature . 2010 ; 464 : 1071 - 1076 .
15. Verhaak RG , Hoadley KA , Purdom E , et al. Integrated genomic analysis identifies clinically relevant subtypes of glioblastoma characterized by abnormalities in PDGFRA, IDH1, EGFR, and NF1. Cancer Cell . 2010 ; 17 : 98 - 110 .
16. Yan W , Zhang W , You G , et al. Molecular classification of gliomas based on whole genome gene expression: a systematic report of 225 samples from the Chinese Glioma Cooperative Group . Neuro Oncol . 2012 ; 14 : 1432 - 1440 .
17. Kogo R , Shimamura T , Mimori K , et al. Long noncoding RNA HOTAIR regulates polycomb-dependent chromatin modification and is associated with poor prognosis in colorectal cancers . Cancer Res . 2011 ; 71 : 6320 - 6326 .
18. Chase A , Cross NC . Aberrations of EZH2 in cancer . Clin Cancer Res . 2011 ; 17 : 2613 - 2618 .
19. Geng YJ , Xie SL , Li Q , Ma J , Wang GY . Large intervening non-coding RNA HOTAIR is associated with hepatocellular carcinoma progression . J Int Med Res . 2011 ; 39 : 2119 - 2128 .
20. Li D , Feng J , Wu T , et al. Long intergenic noncoding RNA HOTAIR is overexpressed and regulates PTEN methylation in laryngeal squamous cell carcinoma . Am J Pathol . 2013 ; 182 : 64 - 70 .
21. Chu C , Qu K , Zhong FL , Artandi SE , Chang HY . Genomic maps of long noncoding RNA occupancy reveal principles of RNA-chromatin interactions . Mol Cell . 2011 ; 44 : 667 - 678 .
22. Kim K , Jutooru I , Chadalapaka G , et al. HOTAIR is a negative prognostic factor and exhibits pro-oncogenic activity in pancreatic cancer . Oncogene . 2013 ; 32 : 1616 - 1625 .