The role of the CCDC 26 long noncoding RNA as a tumor suppressor

CCDC26 on chromosome 8q24 is considered to encode a long intergenic noncoding RNA because the short open reading frame within the mRNA transcribed from this gene is not conserved in any other species. Genome-wide analysis has revealed association of CCDC26 with certain tumors, for instance low-level glioma. Moreover, moderate amplifications of the whole or part of the CCDC26 genetic locus have been observed in pediatric acute myeloid leukemia patients. The CCDC26 gene is amplified in the HL-60 acute myeloid leukemia cell line, in which double minute chromosomes—abnormal tiny chromosomes—harbor the CCDC26 gene. We examined the function of CCDC26 by gene knock-down (KD) using short hairpin RNAs in K562 human myeloid leukemia cells. In four stable KD clones, CCDC26 expression was suppressed to 1% of its normal level by transcriptional gene suppression, not post-transcriptional suppression. The growth rates of these KD clones were reduced compared with those of control cells in media containing high serum concentrations. In contrast, in media containing much lower serum concentrations, the KD clones exhibited significantly higher growth rates than controls, and increased survival after serum withdrawal. Enhanced expression of a receptor tyrosine kinase, KIT, was detected in the KD clones, and treatment with ISCK03, a KIT inhibitor, eliminated their increased survival in the absence of serum. Therefore, CCDC26 seems to control myeloid leukemia cell growth through regulation of KIT expression. These observations suggest that CCDC26 is a tumor-suppressive long noncoding RNA because it suppresses the KIT oncogene that supports survival of cancer cells in the stem cell state.


Introduction
Noncoding RNAs (ncRNAs) have intrinsic functions without being translated into polypeptides.Classically, ncRNAs include 18S and 28S rRNAs that are transcribed by RNA polymerase I; and tRNA, 5S and other rRNAs that are transcribed by RNA polymerase III.ncRNAs also include two subsets: microRNAs (miRNAs) and long noncoding RNAs (lncRNAs) that are usually transcribed by RNA polymerase II.miRNAs are approximately 20 bp long and silence specific target genes, while lncRNAs extend from 200 bp to several kb in length and have multiple roles in cellular functions and gene regulation.While the regulatory mechanisms of miRNAs acting on their targets through Argonaute protein are well documented, that of most lncRNAs remains to be fully elucidated [1] .
In the last few years, there has been an exponential increase in the number of reports of lncRNA involvement in diseases such as cancer [2] .The role of lncRNAs, especially their regulatory role in gene expression, has received much attention because they could provide new diagnostic and  [73] .therapeutic targets.Many lncRNAs have been shown to function in an oncogenic or anti-oncogenic manner.For example, ANRIL (breast cancer, and others) [3][4][5] , CCAT1 (colorectal cancer, gastric carcinoma, hepatocellular carcinoma) [6] , H19 (bladder cancer) [7] , HOTAIR (breast cancer, and others) [8,9] , MALAT-1 (lung cancer, and others) [10, 11]   , PCA3 (prostate cancer) [12,13] , and XIST (bladder cancer and others) [14,15] have been shown to be correlated with tumor progression, while GAS5 (gastric carcinoma and others) [16,17] , LOC285194 (osteosarcoma) [18] , MEG3 (pituitary tumor and others) [19][20][21] , and NDM29 (neuroblastoma) [22] are considered putative tumor suppressors.Another several tens of lncRNAs suspected to be associated with tumors can be found in public databases [23]   .The expression of several lncRNAs correlates well with cancer cell malignancy and could be used as a biomarker useful for diagnosis and prognosis.In particular, PCA3 has been established as one of the most reliable biomarkers of prostate cancer [24] .Meanwhile, there is also an increasing understanding of the molecular mechanisms of lncRNA regulation of cancer cell growth and metastasis.ANRIL, H19, HOTAIR, NDM29, and XIST interact with polycomb repressive complex 2 (PRC2) or its components to repress their target genes transcriptionally and epigenetically to progress malignancy [3,7,9,15,22] .CCAT-1 antagonizes its target miRNAs as a "sponge" to progress hepatocellular carcinoma [6] .MALAT-1 adjusts splicing mechanisms to regulate gene expression post-transcriptionally [25] .GAS5 lncRNA functions as a tumor-suppressor through the mechanisms including suppression of certain steroid hormone receptors and sequestration of the miRNA, miR-21, in multiple cancer cells [17] .LOC285194 inhibits tumor cell growth through regulation of its target miRNA [18] .MEG3 affects protein levels of the tumor suppressor p53 in lung cancer cells [21] .PCA3 is encoded on the antisense strand within the intron of the tumor suppressor gene PRUNE2.A unique lncRNA transcribed from PCA3 was shown to form double-stranded RNA with the premature RNA of PRUNE2 to suppress it through ADAR (adenosine deaminase acting on RNA)-dependent adenosine-inosine RNA editing [13] .Although the mechanisms vary widely, most lncRNAs affect cancer cells through regulation of protein-encoding genes.

RESEARCH HIGHLIGHT
Recently, clinical research has revealed that CCDC26, which encodes a protein with unknown function, is related to tumors including acute myeloid leukemia (AML) and glioma.In this review, we discuss the role of CCDC26 in myeloid leukemia with respect to its contribution to cell growth.

CCDC26 is a putative intergenic long noncoding RNA
Human transcriptome analysis revealed the existence of CCDC26, an mRNA-type lncRNA expressed in hematopoietic cells [26] .Figure 1 depicts an overview of the CCDC26 genetic locus.CCDC26 is located in a 350-kb region between LOC728724, another lncRNA, and GSDM, a hypothetical protein-coding proto-oncogene (Figure 1a).Therefore, CCDC26 is classified a long intergenic noncoding RNA (lincRNA), the function of which is unknown.There are two major splicing variants of the CCDC26 mRNA with common terminal exons (exons 3 and 4) (Figure 1b).
A 109-amino-acid open reading frame (ORF) could be encoded within exon 4. The actual protein (Coiled Coil Domain Containing), however, has not been observed.Furthermore, homologous proteins are not found in any other species [26] .Therefore, this ORF could have arisen simply by accidental absence of the terminal codon.Regardless of whether the protein is functional, this ORF can work to prevent rapid degradation of CCDC26 RNA by nonsense-mediated RNA decay [27] , which is a mechanism associated with quality control of mRNA-degrading mRNAs without a certain extent of ORF that are not useful for protein synthesis.The ORF on CCDC26 mRNA will prolong its half-life and may maintain the function (if any) of the CCDC26 RNA.The absence of a functional protein encoded by CCDC26 implies that it functions as an lncRNA not a protein-encoding mRNA.
Several lines of evidence support a relationship between CCDC26 and tumors, including AML and glioma.Within this locus, particular regions are frequently rearranged in AML and AML-derived cells [28][29][30] .Single nucleotide polymorphisms (SNPs) associated with glioma [31,32] and a retrovirus insertion site, where a viral insertion makes AML cells resistant to retinoic acid [33] indicated in Figure 1b will be discussed in detail later.
High levels of histone methylation and acetylation were observed in several regions within the CCDC26 intron (Figure 1b), meaning that these locations may be actively transcribed.Furthermore most of these regions are highly conserved among mammals, again implying functionality.Moreover, transcripts other than the mature CCDC26 mRNA have been reported in the intron.Among them, miR-3686, an miRNA encoded in the intron of CCDC26 and registered in miRBase (http://www.mirbase.org/) [34]could play a role still unknown.
Highly conserved regions in the introns of CCDC26 suggest the existence of some further biological functions.By comparison of the syntenic region in mouse and other mammals, we found a homologous region in the CCDC26 locus including CCDC26 exon 4 itself.A mouse lncRNA, AK015428, is considered to be related to mouse spermatogenesis.The human homolog of AK015428, hAK015428, is another homologous region in the CCDC26 locus: located upstream of the CCDC26 promoter and actively transcribed in human myeloid cells.hAK015428 might be a promoter-related noncoding RNA of CCDC26 [35]   .Of note, the locations of CCDC26 exon 4 and hAK015428 are coincident with cis-elements, more than 1 Mb from the MYC oncogene locus, NDME (Notch dependent Myc enhancer element), and BDME (Brd4-dependent Myc enhancer) (Figure 1b), which regulate MYC expression in human and mouse T cell acute lymphoblastic leukemia cell lines [36] .However the significance of this coincidence is unclear.It is possible that CCDC26 exon4 and hAK015428 function as enhancer-associated RNAs [37] .In summary, there might be biologically functional regions in the introns and upstream of CCDC26 in addition to in its exons.

CCDC26 in glioma
Associations between particular SNPs with primary brain tumor (PBT) were discovered by genome-wide association studies (GWAS) [31] .Glioma is a tumor of glial cells accounting for a major part of PBT, and contains cases with different stages of malignancy, namely benign glioma (grade I), diffuse astrocytoma (grade II), anaplastic astrocytoma (grade III), and glioblastoma (grade IV) [38] .Genetic mutations characteristic of glioma have been reported to occur in genes for six DNA repair enzymes, (PRKDC, XRCC, PARP1, MGMT, ERCC1 and ERCC2), epidermal growth factor (EGF) and interleukin 13 (IL13) [38] .Furthermore, a GWAS revealed association of another five genes with glioma, including TERT (telomere reverse transcriptase), RTEL1 (regulator telomeric elongation helicase 1), CDKN2A/2B (a tumor suppressor), PHLDB1 (pleckstrin homology-like domain family B member 1), and CCDC26 [31] .All these genes except CCDC26 are protein-coding genes that are interconnected in biological networks.On the other hand, CCDC26 is a non-protein-coding gene, the function of which remains utterly unknown [38] .The CCDC26 gene locus was strongly associated with glioma through five SNPs including rs4295627, rs16904140, rs6470745, rs891835, and rs10464870 [39][40][41][42] .Another, higher-resolution GWAS revealed the strongest association in the CCDC26 locus: rs55705857 with PBT [32,43] .A specific nucleotide in a CCDC26 SNP is a risk factor in low-grade (I-II) glioma, but not necessarily in high grade (III-IV) glioma cases [44] .A different SNP in the intergenic region, 400 kb downstream of CCDC26, rs987525, was linked to cleft palate that is also a risk factor for PBT [45] .Therefore, cleft palate might have relationship with PBT through CCDC26.
There might be a synergistic effect with other cancer genetic risk factors.CCDC26 genotype is associated with mutation of the metabolic enzymes isocitrate dehydrogenase 1 and 2 (IDH1/2) in low-grade glioma [46] .These IDH1/2 mutations affect enzymatic activity to reduce α-ketoglutarate levels and elevate 2-hydroxyglutarate levels.This results in suppression of certain gene-regulatory proteins, including HIF1 through altered histone modification [47] .Considering the synergy of CCDC26 with IDH1/2, CCDC26 may be linked to a certain subpopulation of gliomas [46,48] .A recent GWAS on pediatric PBT demonstrated associations with genes including RTEL1, TERT, and CCDC26; these are in common with associations identified in adult PBT.Additional to these, another noncoding RNA was identified-ANRIL (CDKN2B-AS), which down-regulates CDKN2B via an antisense transcript-and the tumor suppressor gene CDKN2B is related to adult PBT [49] .Interestingly, although ANRIL and RTEL1 are significantly related to astrocytoma, TERT and CCDC26 are stronger risk factors in the non-astrocytoma subtype of pediatric PBT.These observations suggest that pediatric and adult PBT share common genetic risk factors involving CCDC26, and similar etiological pathways that are related to the specific subcategory of PBT.

CCDC26 in acute myeloid leukemia
Acute myeloid leukemia (AML) is characterized by mutations in a set of genes that can be divided into two classes [50][51][52] .One class includes genes related to the proliferation or survival of cells and includes FLT3, ABL, RAS, and KIT, and the other consists of genes related to cell differentiation: HOXA9, AML1, MLL, NPM1, and RARα.Mutations in genes of both classes are required to give rise to AML.
CCDC26 was reported to be associated with differentiation and apoptosis of PLB985 cells (a myelocytic leukemia cell line) induced by treatment with retinoic acid (CCDC26 is also known as RAM, for 'retinoic acid modifying').In isolated cells that had acquired resistance to retinoic acid after transduction with a retrovirus vector, insertion of the viral DNA in the intron of CCDC26 was observed Although knock-down (KD) of CCDC26 with RNA interference had little influence on the cells in ordinary conditions, retinoid acid-induced differentiation and apoptosis was significantly promoted [33] .Retinoic acid induces differentiation and apoptosis of not only some types of leukemia cells but also of neuroblastoma and glioblastoma cells, through modification of transcriptional regulation.Therefore CCDC26 might affect the differentiation state of these cells, too.
A genome-wide study revealed that CCDC26 is most frequently changed in childhood AML patients.Radtke and colleagues investigated copy number alterations (CNAs) in pediatric AML using comprehensive SNP array analysis, and found the most common CNA, occurring in 14% (15 of 111) of pediatric AML patients, to be on chromosomal loci 8q24 with a low-burden copy number increase (2.83-3.77copies) [53]   .Four out of 15 cases indicated focal CNA and the other 11 indicated CNA in a broader region including trisomy 8, which occurs frequently in AML [54] .They had a common altered region (MAR)-a 20-Mb region of 8q24-which interestingly contains part of the CCDC26 intron and exon (Figure 1b).
Notably, in accord with the situation for low-grade glioma, mutations of IDH1/2 are important for diagnosis and prognostic prediction in AML patients [56] .While cytogenetically normal AML patients with an NPM1 (nucleophosmin1) mutation and a normal FL3 gene is associated with favorable outcomes, AML patients with IDH1/2 mutation in addition to the same genetic background show adverse prognosis with poorer remission.Combined with the observation that CCDC26 is up-regulated in the NPM1-mutated subset of AML [57] , CCDC26 and IDH1/2 are also implied to share synergistic responsibility for AML.

The CCDC26-containing 8q24 locus is a hot spot for chromothripsis in AML
Our own studies and those of others have shown that all or part of the CCDC26 gene is often amplified in AML cells harboring aberrant tiny extrachromosomal elements or double minute chromosomes (referred to as dmins).Dmins are cytogenetic abnormalities sometimes observed in AML cells although less commonly [28][29][30] .In HL-60 cells, two or four repeats of a 1.8-Mb amplification unit (amplicon) constitute their dmins.The amplicon is derived from several truncated regions of a 4.6-Mb region of chromosomal loci 8q24, contains an intact MYC oncogene.In addition to MYC, several other genes, including CCDC26 and tribbles homolog 1 (TRIB1), are also encoded on the amplicon (Figure 1a) and actively transcribed.Transcription of all these genes was significantly suppressed accompanied with differentiation of HL-60 induced by treatment with anticancer drug, indicating that they might be related to the cancerous nature of cells.The anticancer drug hydroxyurea blocks proliferation of HL-60 cells by promoting exclusion of dmins.Even after exclusion of dmins, MYC of the original chromosomal locus remained intact, but was no longer transcribed [58] .These observations suggest aberrant gene(s) on dmins other than MYC might be responsible for proliferation of HL-60 cells.
As a result of chromosomal rearrangement, the CCDC26 gene on dmin in HL-60 cells is changed to an incomplete form without exon 4 (which encodes the ORF) to produce abnormal transcripts [30] .Common breaking points in the CCDC26 locus have been found in dmin of AML patients and AML cell lines (Figure 1b), suggesting the common mechanism behind the generation of dmins in AML cells.Dmin in HL-60 consists of several segments spanning a 200 to 700-kb genomic region, which have been joined by non-homologous recombination (our published [30] and unpublished observations).Such defective chromosomal rearrangements occurring in an extensive but limited chromosomal region are termed "chromothripsis", a mechanism that explains the cytogenetic abnormalities often observed in cancerous cells [59] .

CCDC26 is a tumor suppressor gene, but might not be an oncogene
Fusion of BCR and ABL (BCR/ABL) is well known to be sufficient to cause CML [60] .However, an acute blastic crisis of CML is often accompanied by mutation of genes that are also mutated in AML, such as HOXA9 or AML [61,62] .Thus, hematopoietic tyrosine kinases, including FLT3, ABL, RAS, and KIT, seem to play a similar role in AML and CML [63] .Considering the importance of imatinib, an ABL tyrosine kinase inhibitor used in the treatment of BCR/ABL-positive CML, inhibitors of other tyrosine kinases are likely to become increasingly important for the treatment of AML [60,64] .
KIT is a receptor tyrosine kinase that is considered to play a role in AML because of its frequent up-regulation in patients.Expression of KIT in leukemia stem cells (LSCs) from pediatric AML patients who relapsed after chemotherapy was increased compared with that in patients who did not relapse [65] .In work we have recently published [66] , we obtained K562 cells in which CCDC26 expression was constitutively knocked down with a short hairpin RNA to 1% of the level in the original K562 cells (called KD cells).Although KD cells grew at similar rate to the original cells in normal conditions, their growth slowed down in high-serum (15%) conditions.On the other hand, KD cells proliferated at a higher rate than the original cells in low-serum (0.1%) conditions.Moreover, in no-serum conditions, significantly fewer CCDC26-KD cells died than did original K562 cells.In these conditions, the original K562 cells stopped proliferating and died slowly, and cell death occurred in the G2/M cell cycle phase rather than the G1 phase.Therefore, under serum starvation, suppression of CCDC26 resulted in a prevention of cell death at a specific point in the cell cycle, and more surviving cells.
DNA microarray analysis demonstrated that the gene expression profile in CCDC26-KD cells was shifted from that of the original cells, indicating that CCDC26 regulates the expression of a set of genes, although the mechanism remains unknown.Moreover, CCDC26 and some neighboring genes including GSDM and LOC728724 undergo transcriptional down-regulation in KD cells.This regulation seems to involve epigenetic transcriptional repression because treatment with 2′-deoxy-5-azacytidine and trichostatin A to release epigenetic silencing partially reversed this repression.However an influence on MYC has yet to be observed.Among the genes that show altered expression in CCDC26-KD cells, the activation of endogenous KIT is highly interesting.Point mutations in KIT that enhance tyrosine kinase activity are frequently accompanied by the chromosomal translocation t(8;21) in adult AML, which results in poor prognosis [67] .We confirmed the presence of endogenous KIT protein, with a normal molecular weight and canonical localization, in KD cells.ISCK03, a KIT inhibitor, abolished this prolonged survival.This provides evidence for a new role for CCDC26 in myeloid leukemia through the regulation of a set of genes including KIT.The concept that CCDC26-KIT regulation is key to survival of leukemia cells is important from the therapeutic viewpoint.KIT is an important gene in many different biological aspects [68] .KIT protein is a receptor for stem cell factor, which is expressed in the stromal cells of bone marrow and is necessary for self-renewal and maintaining hematopoietic stem cells in an undifferentiated state.In leukemia, malignant potential is considered to be tightly related to LSCs continuing self-renewal in the undifferentiated state.LSCs with CD34+ and CD38− surface markers are reportedly a cause of AML recurrence because they have the potential to survive in a special microenvironment, namely niche sites where they can escape the influence of drugs [69] .CCDC26-KD cells might share some properties with LSCs, including relatively slow growth and the ability to survive under certain conditions, such as a shortage of growth factors.Constitutive activation of KIT might contribute to the survival of these cells by an autocrine mechanism involving stem cell factor [70] .Although LSCs are usually negative for surface KIT protein in de novo AML [71] , the existence of KIT-positive LSCs is related to an increased tendency of pediatric AML recurrence after chemotherapy.This can arise from constitutive activation of KIT protein [72] .Survival of KD cells was sensitive to the KIT inhibitor ISCK03 in a dose-dependent manner.After ISCK03 treatment, the survival of KD cells was suppressed to the same level as that of non-KD cells.Therefore, KIT inhibitors are a potential strategy for the treatment of myeloid leukemia harboring an altered CCDC26 genetic locus.The down-regulation of KIT, a well-known oncogene, by CCDC26 suggests that it functions as a tumor suppressor gene.On the other hand, the amplified copy number of CCDC26 in some AML cases is conflicting.This paradox is explicable as follows.The low-burden increased copy number of CCDC26 is only partial and does not extend across the whole gene.Therefore this CNA might reflect the situation that CCDC26 is broken and corrupted into partially multiplicate forms at the internal chromothripsis hotspot.Moreover, the partial increased copy number and high expression of part of CCDC26 observed in HL-60 cells might be able to interfere with the function of the intact original gene.This explanation is consistent with the hypothesis that CCDC26 functions in a tumor-suppressive manner, and is not tumor promoting.

Conclusion
The current evidence suggests that the lncRNA CCDC26 is not oncogenic, but an anti-oncogene that down-regulates the oncogene KIT.KIT inhibitors might be effective in the treatment of cancers with CCDC26 mutation, because the KIT receptor tyrosine kinase plays a role in the survival of cancer stem cells in their niche.

Figure1
Figure1.(A) Summary of the genetic locus 8q24.21.The 8q24.21 region spans 126.1 Mb to 130.8 Mb (right to left; numbering is based on human genome assembly 37.1).Three of the regions amplified in HL-60 are indicated by filled rectangles.Distal and peripheral breakpoint clusters (DBC and PBC, respectively) in dmin-positive AML patients are shown.(B) An enlarged part of 8q24.21 including the CCDC26 genetic locus, from 130.35 Mb to 130.80 Mb (right to left).Marks beneath the scale indicate the locations of six SNPs that are linked to glioma (rs4295627, rs16904140, rs55705857, rs6470745, rs891835, and rs10464870; open circles, left to right), and a retrovirus insertion site (filled circle).The locations of the amplified region in HL-60 cells and the commonly amplified region (MAR) in childhood AML are shown by colored rectangles.The locations of NDME (Notch dependent MYC enhancer) and BDME (BRD dependent MYC enhancer) are indicated.Below them, which covers a 350-kb region, are shown the major variants of CCDC26 mRNA (long and short).The long transcript consists of four (1-2-3-4) exons, and the short transcript comprises three (1a-3-4) or four (1a-2a-3-4) exons.All variants share exons 3 and 4, in which the hypothetical ORF is encoded.Actively transcribed regions hAK015428, THS1, and THS2, which are not major exons of CCDC26 mRNA, and the location of mir-3686, are indicated.The transcription scores are shown from RNA-seq experiments with whole cell and nuclear RNA from K562 myeloid leukemia cells, as are the patterns of histone H3K4 methylation and histone H3K27 acetylation.The bars at the bottom indicate regions with conserved synteny among mammals, as obtained from the UCSC Genome Browser[73] .