A broad range of human malignancies is associated with nonrandom 1p36 deletions, suggesting the existence of tumor suppressors encoded in this region. Evidence for tumor-specific inactivation of 1p36 genes in the classic “two-hit” manner is scarce; however, many tumor suppressors do not require complete inactivation but contribute to tumorigenesis by partial impairment. We discuss recent data derived from both human tumors and functional cancer models indicating that the 1p36 genes CHD5, CAMTA1, KIF1B, CASZ1, and miR-34a contribute to cancer development when reduced in dosage by genomic copy number loss or other mechanisms. We explore potential interactions among these candidates and propose a model where heterozygous 1p36 deletion impairs oncosuppressive pathways via simultaneous downregulation of several dosage-dependent tumor suppressor genes. Cancer Res; 72(23); 6079–88. ©2012 AACR.
Deletions of the distal short arm of chromosome 1 (1p) are frequently observed in a broad range of human cancers, including breast cancer, cervical cancer, pancreatic cancer, pheochromocytoma, thyroid cancer, hepatocellular cancer, colorectal cancer, lung cancer, glioma, meningioma, neuroblastoma, melanoma, Merkel cell carcinoma, rhabdomyosarcoma, acute myeloid leukemia, chronic myeloid leukemia, and non-Hodgkin lymphoma (1, 2). These nonrandom aberrations suggest that loss of genetic information mapping to this region contributes to cancer development. This is supported by constitutional 1p aberrations in neuroblastoma patients (3, 4) and the association of 1p deletion with poor survival of neuroblastoma (5), breast cancer (6, 7), and colon cancer (8, 9) patients. Deletion of 1p in premalignant lesions and/or early tumor stages of colorectal, breast, and hepatocellular cancer (10–12) points to a role for 1p genes during the early steps of carcinogenesis in these entities. This is supported by loss of 1p material during in vitro progression in a cell culture model of colon carcinogenesis (13). Furthermore, transfer of 1p chromosomal material suppresses tumorigenicity of both neuroblastoma and colon carcinoma cells (14, 15).
Since the first report of 1p deletions in neuroblastomas in 1977 (16), smallest regions of overlapping heterozygous deletions (SRO) have been defined in various tumor entities in the pursuit of cancer-related genes. That 1p36 is a hot spot of chromosomal aberrations became clear early on (1), with the most detailed mapping picture appearing for neuroblastoma (Fig. 1; refs. 1, 17–29). Despite extensive 1p36 candidate gene sequence analyses, success was limited for identifying tumor-specific mutations in neuroblastomas or other malignancies, which led some to conclude that a deletion mapping approach was unlikely to deliver tumor suppressor genes. Many tumor suppressor genes, however, do not require inactivation in a classic “two-hit” manner but contribute to tumor development when their dosage is reduced, sometimes only subtly, by mechanisms such as copy number change, transcriptional repression, epigenetic downregulation, or aberrant miRNA regulation (30). Unlike in a classic “two-hit” mutational inactivation scenario, definite proof for dosage-sensitive tumor suppressor gene involvement is not offered by a single straight forward assay. Instead, evidence must be accumulated from genetic, epigenetic, and transcriptional analyses of human tumors and functional in vitro and in vivo assays. This review discusses five 1p36 genes, CHD5, CAMTA1, KIF1B, CASZ1, and miR-34a, recently suggested as tumor suppressor candidates and likely to be impaired by partial reduction as suggested by both their status in human cancers and their activity in functional cancer models.
Chromodomain helicase DNA binding (CHD) genes encode a class of ATPase-dependent DNA-binding proteins interacting with histones to modulate chromatin structure and transcription. CHD5 resides in 1p36.31, is preferentially expressed in neuronal tissues, and its product regulates genes involved in neuronal function, cell-cycle control, and chromatin remodeling (31). Functional evidence for a tumor suppressive role of mouse Chd5 derives from an elegant approach using chromosome engineering to generate mouse models with loss or gain of genomic regions corresponding to human 1p36 (22). Deletion of a Chd5-containing 4.3 Mb genomic subinterval, corresponding to 5.7 Mb of human 1p36, enhanced proliferation, loss of contact inhibition, spontaneous immortalization, and sensitivity to oncogenic transformation of cultured mouse embryonic fibroblasts. Mice with heterozygous deletion of this subinterval were prone to hyperplasia in a variety of tissues (22). Duplication of this subinterval in mouse embryonic fibroblasts inhibited proliferation and increased the senescent cell fraction. Mice with subinterval duplication had developmental abnormalities characterized by an increased apoptotic cell fraction in various tissues, including the neural tube (22). The identified subinterval includes 52 genes. Among 11 tested candidate genes, knockdown of only Chd5 functionally rescued the proliferative defect of mouse embryonic fibroblasts with duplication of the subinterval. Chd5 knockdown in wild-type cells induced phenotypic changes closely resembling the effects of the engineered deletion, including enhanced proliferation, sensitivity to oncogenic transformation, and inhibition of p19Arf/p53 (22). This suggests that Chd5 is a dose-dependent gene within the identified 4.3 Mb genomic subinterval that mediates the tumor suppressive mechanisms seen in mouse models. The functional role of human CHD5 in a cancer background was analyzed in neuroblastoma cells, where its overexpression had no impact on proliferation, morphology, differentiation, or apoptosis but significantly inhibited clonogenic growth in soft agar and xenograft tumor growth in mice (32). The absence of an impact on proliferation may indicate an already impaired p14Arf (human p19Arf homolog)/p53 pathway in these cells, a defect frequently seen in neuroblastomas (33).
CHD5 is one of 23 genes mapping to a 2 Mb SRO in neuroblastoma (34) and a 5.4 Mb SRO spanning 1p36.32 to 1p36.22 in glioma (Fig. 1; refs. 22, 34). An SRO containing Chd5 was identified in a lymphoma mouse model with chromosomal instability, and syntenic CHD5-containing deletions were discovered in human T-cell acute lymphoblastic leukemia/lymphomas (T-ALL; ref. 35). CHD5 mutations are rare in the entities analyzed so far. Heterozygous missense mutations were found in one of 30 neuroblastoma cell lines (34), 2 of 14 metastatic prostate tumors (36) and 3 of 123 primary ovarian cancers (37). Reports of CHD5 mutation frequency in breast cancer are controversial, ranging from 0% (0/60; ref. 37) to 8.5% (3/35; ref. 38). None of the studies have identified nonsense or frameshift mutations, and whether the missense mutations impair CHD5 function, remains to be investigated. Aberrant CHD5 promoter methylation in primary tumors was identified in 73% of gastric cancers, 17% of colon cancers, 10% of breast cancers, 10% of ovarian cancers, and 4% of gliomas (37, 39, 40), indicating that CHD5 downregulation via promoter methylation mediates a selective advantage in the development of a subset of human tumors. In neuroblastomas, low CHD5 expression is associated with high-risk features such as 1p deletion, amplified MYCN oncogene, and advanced stage (41). Furthermore, low CHD5 mRNA and protein expression in neuroblastomas are significantly associated with poor patient outcome, even when adjusted for established prognostic variables (32, 42). Together, this suggests that CHD5 is a neuronal gene whose dose reduction contributes to tumor development by inhibiting the p14Arf/p53 pathway. This function is likely to be mediated by CHD5 acting as a transcriptional regulator via chromatin remodeling, an idea supported by the presence of CHD5 in a multiprotein complex highly similar to NuRD chromatin remodeling complexes (31).
CAMTA1 encodes a member of the calmodulin-binding transcription activator (CAMTA) protein family (43, 44), is localized in 1p36.31-p36.23 and predominantly expressed in neural tissues, including brain and spinal chord (45). CAMTA1 maps to virtually all recently described 1p36 neuroblastoma SROs (summarized in ref. 46 and Fig. 1), is the only gene mapping to a 150 Kb SRO in glioma (Fig. 1; ref. 23) and is homozygously deleted in a subgroup of glioblastomas (47). In colorectal cancers, deletion of a small region at 1p36.31-p36.23, including only the CAMTA1 gene, had the strongest impact on survival among all identified genomic alterations (48). A missense mutation was seen in one of 26 colorectal cancers (48), but somatic CAMTA1 mutations were not observed in neuroblastomas or gliomas (23, 49); however, CAMTA1 expression is significantly lower in high-risk tumors of both entities (46, 50). In neuroblastoma, low CAMTA1 mRNA expression is significantly associated with prognostic markers of poor outcome, including amplified MYCN and advanced tumor stage (46). Low CAMTA1 expression was also identified as a new independent marker of poor outcome adding prognostic information to existing risk stratification (46). Consequently, CAMTA1 is included in most recent prognostic neuroblastoma expression classifiers (51–54). Low CAMTA1 expression is also significantly associated with shorter survival in glioblastoma patients (50), and, intriguingly, low CAMTA1 expression emerged as a new independent predictor of poor outcome in patients with a tumor that is not of neural origin, colorectal cancer (48).
Functional evidence for a tumor suppressive role of CAMTA1 comes from analyses in neuroblastoma and glioblastoma cells. In neuroblastoma cells with low endogenous CAMTA1 levels, ectopic CAMTA1 expression inhibits proliferation, induces accumulation of cells in the G1–G0 phase of the cell cycle and inhibits anchorage-independent colony formation and xenograft tumor growth (55). CAMTA1 induction shifts neuroblastoma cell morphology toward a more differentiated type, including induction of neuron-specific markers. The transcriptome of CAMTA1-induced cells reflects their phenotype and is significantly enriched for genes that mediate cell-cycle inhibition and neuronal function (55). CAMTA1 is upregulated in neuroblastoma cells prompted to differentiate by retinoic acid or other stimuli (55). In glioblastoma cell models, CAMTA1 overexpression reduces both neurosphere formation and xenograft tumor growth, probably mediated by activation of natriuretic peptide A (NPPA), a secreted peptide with a strong antiproliferative effect on glioblastoma cells (50). The mechanisms downregulating CAMTA1 in high-risk tumors are largely unknown. CAMTA1 expression is significantly lower in neuroblastomas, gliomas, and colorectal cancers with 1p deletion compared with tumors retaining 1p (23, 46, 48). This conforms to a haploinsufficiency model where a single CAMTA1 copy would result in insufficient transcript levels. However, even in neuroblastomas without 1p deletion, low CAMTA1 expression predicts poor outcome (46), indicating additional CAMTA1-repressive mechanisms. No evidence for aberrant methylation of CAMTA1-associated CpG islands was found in neuroblastomas or colorectal cancers (48, 55), but other epigenetic mechanisms might be relevant, as indicated by upregulation of CAMTA1 in neuroblastoma cells treated with histone deacetylase inhibitors (55). Another mechanism of CAMTA1 downregulation was identified in glioblastoma cells, where it is targeted by miR-9/9*, a miRNA pair that is highly abundant in glioblastoma stem cell-enriched CD133+ cell populations (50). Summing up, nonrandom CAMTA1 deletions in tumors, the role of CAMTA1 in differentiation and growth suppression, and the strong association between CAMTA1 downregulation and poor survival in neuroblastoma, glioma, and colorectal cancer patients support its assignment as a dosage-dependent tumor suppressor gene.
Further evidence for an involvement of CAMTA1 in cancer comes from epithelioid hemangioendothelioma (EHE), a rare vascular sarcoma difficult to diagnose because of considerable morphological overlap with other epithelioid vascular tumors. Two independent studies identified an EHE-characteristic translocation, t(1;3)(p36.3;q25), involving WWTR1 (WW domain-containing transcription regulator 1) and CAMTA1 (45, 56). The WWTR1/CAMTA1 translocation was found in 100% (56) and 87% to 89% (45) of 2 EHE cohorts but in no other vascular neoplasm analyzed. The translocation results in a fusion gene encoding the N-terminus of WWTR1 fused in frame to the C-terminus of CAMTA1. This EHE-specific fusion has the potential to (i) serve as a new marker for tumor detection, diagnosis, and monitoring; (ii) provide a highly specific therapeutic target; and (iii) act as a study model to gather insights into the physiological functions of WWTR1 and CAMTA1. The identification of a highly specific CAMTA1-involving genomic rearrangement seen in virtually all tumors of a single cancer entity further implicates CAMTA1 in cancer development.
The KIF1B kinesin motor protein is involved in axon myelination and outgrowth as well as axonal transport of mitochondria and synaptic vesicles (57–59). KIF1B maps to 1p36.22 and encodes two alternatively spliced isoforms, KIF1Bα and KIF1Bβ, conferring different axonal cargo specificity. KIF1B is one of 6 genes within a 500 Kb homozygous deletion found in a neuroblastoma cell line (Fig. 1; refs. 20, 60). Sequence analysis did not reveal mutations in oligodendrogliomas or a panel of pediatric solid tumor cell lines, including rhabdomyosarcoma and Ewing sarcoma cells (61, 62). A missense variant of unknown functional significance was detected in 6 of 100 neuroblastomas (63). Another mutation screen identified missense variants in three of 111 neuroblastomas, 2 of 52 pheochromocytomas, and 1 of 14 medulloblastomas (64). Intriguingly, all variants identified in the latter study were shown to impair KIF1Bβ function in vitro (64), and one of these loss-of-function variants was present in the germline of a three-generation cancer-prone family, segregating with predisposition to pheochromocytoma, neuroblastoma, ganglioneuroma, and lung adenocarcinoma (65). This indicates that KIF1B sequence variants/mutations are infrequent but may be pathogenic in a subset of tumors, which is further supported by a KIF1B single-nucleotide polymorphism (SNP) highly associated with hepatitis virus B (HBV)-related hepatocellular carcinoma (66). KIF1B expression is significantly lower in advanced neuroblastoma stages (63, 67, 68), as is KIF1B expression in hepatocellular carcinomas from chronic HBV carriers compared with tumor-adjacent tissue (66).
Consistent with a tumor suppressive function, KIF1Bβ induction triggers apoptosis in neuroblastoma cells, pheochromocytoma cells, and rat sympathetic neurons (63, 64). KIF1Bβ knockdown in rat sympathetic neurons prevents apoptosis following nerve growth factor (NGF) withdrawal, indicating that KIF1Bβ plays a crucial role in neuronal apoptosis upon NGF limitation (64). KIF1Bβ knockdown also enhances anchorage-independent colony formation and xenograft tumor growth (63). Downregulation of KIF1B in advanced tumors can be mediated by heterozygous 1p loss, as indicated by significantly lower KIF1B levels in both gastrointestinal stromal tumors and neuroblastomas with heterozygous KIF1B deletion (63, 69). No evidence for methylation of KIF1B-associated CpG islands was found in neuroblastomas (63, 67), but chromatin remodeling mechanisms have been suggested to be relevant as the BMI1 Polycomb group protein strongly represses KIF1Bβ by direct binding to the KIF1B promoter (70). Intriguingly, MYCN/MYC oncoproteins directly bind to the BMI1 promoter and induce its transcription (70, 71), suggesting a model where MYCN/MYC represses KIF1B via BMI1-mediated epigenetic chromosome modification. Most MYCN-amplified neuroblastomas harbor 1p deletions indicating that KIF1B expression may be inhibited by both reduced copy number and amplified MYCN in this subgroup. A positive regulator of KIF1Bβ is the proapoptotic hydroxylase EGLN1, placing KIF1B in a pathway to eliminate excess neuroblasts during embryonal development (64) that is likely to be implicated in the pathogenesis of neural crest-derived tumors such as neuroblastomas and pheochromocytomas (72). The identification of functionally impairing KIF1B mutations in neuroblastoma, pheochromocytoma, and medulloblastoma, together with KIF1Bβ downregulation in advanced tumors and its inhibitory effect on cancer cells in vivo and in vitro, support a tumor suppressive function for KIF1B that is linked to its proapoptotic role.
Castor zinc finger 1 (CASZ1), localized in 1p36.22, is the human homolog of the Drosophila zinc finger transcription factor Castor, which is expressed in a subset of central nervous system neuroblasts and is involved in late stage neurogenesis (73). CASZ1 maps near the border of a 3 Mb SRO defined by integrating 1p deletions of neuroblastomas and germ cell tumors (Fig. 1; ref. 21). Sequence analysis did not reveal evidence for tumor-specific CASZ1 mutations (74), but low CASZ1 expression is significantly correlated with unfavorable clinical and biologic features and poor overall survival in neuroblastoma (75). Ectopic restoration of CASZ1 enhanced cell adhesion, induced morphological differentiation, accompanied by expression of neuron-specific markers, and inhibited migration, proliferation, and tumorigenicity (75). Furthermore, CASZ1 was increased in differentiating neuroblastoma cells treated with retinoic acid or cAMP-inducing agents (75, 76), which is in line with murine studies suggesting a developmental role for Casz1 in controlling neuronal subtype specification and differentiation (77). Transcriptome analysis of CASZ1-overexpressing neuroblastoma cells revealed signatures consistent with CASZ1-induced phenotypes and a significant enrichment of genes involved in cell growth regulation and developmental processes (75). One means of CASZ1 downregulation in cancer cells is genomic copy number loss, as indicated by lower CASZ1 expression in 1p-deleted tumors (75, 78). Evidence for tumor-specific promoter CpG methylation has not been reported (74, 75), but CASZ1 induction by histone deacetylase inhibitors in neuroblastoma cells (74, 75) suggests that suppressive histone modifications inhibit CASZ1 expression. This idea is corroborated by the finding that the CASZ1 locus is repressed by the enhancer of zeste homolog 2 (EZH2) Polycomb complex histone methyltransferase, an oncoprotein overexpressed in advanced tumors of various entities, including neuroblastoma (79). Together, these data suggest that CASZ1 is a cell growth and differentiation regulator that, when impaired by dosage reduction, contributes to the malignant phenotype of cancer cells.
MiRNAs are small noncoding RNAs involved in posttranscriptional control of gene expression, and their deregulation has been linked to a variety of diseases, including cancer. MiR-34a, localized in 1p36.22 (Fig. 1), is ubiquitously expressed, with highest levels in the brain (80). Aberrant miR-34a downregulation has been detected in many cancer types, including breast cancer (81), epithelial ovarian cancer (82), prostate carcinoma (80), pancreatic ductal adenocarcinoma (83), hepatocellular carcinoma (84), colon cancer (85), non–small cell lung cancer (86), neuroblastoma (87), glioblastoma (88), malignant peripheral nerve sheath tumors (89), melanoma (80), chronic lymphocytic leukemia (90–92), and acute myeloid leukemia (93). In many cancer entities, low miR-34a expression is associated with advanced disease and/or poor patient survival, as seen in neuroblastoma (87), epithelial ovarian cancer (82), peripheral nerve sheath tumors (89), breast cancer (81), and pancreatic ductal adenocarcinoma (83). Ectopic miR-34a expression induces cell cycle arrest, apoptosis, and senescence, and inhibits migration and invasion of cancer cells (94). MiR-34a is directly induced by p53 (95–99) and plays a pivotal role within a p53-activating positive feedback loop where mirR-34a downregulates the SIRT1 class III histone deacetylase, leading to accumulation of active acetylated p53 (100). Additional negative regulators of p53 (MTA2, HDAC1, and YY1) were identified as mir-34a targets (101, 102), indicating the existence of several mir-34a-dependent mechanisms functioning within p53-activating feed back loops. Bioinformatic analyses and global proteomic approaches indicate that regulation of hundreds of additional mir-34a targets contributes to mir-34a–associated cellular functions (101, 102). Validated direct targets include factors involved in G1/S transition (E2F3, cyclin E2, cyclin D1, CDK4, CDK6, MYC, MYCN), apoptosis (BCL2, Survivin), metastatic potential (MET, AXL), Wnt signaling (WNT1, LEF1), and glycolysis (LDHA; discussed in refs. 94 and 102). Other direct miR-34a targets play important roles in the Notch pathway (NOTCH1, NOTCH2, JAG1, DLL1; refs. 103–105), cancer stem cell functions (CD44; ref. 106), or growth factor signaling (ARAF, PIK3R2; ref. 107). Considering this broad spectrum of potentially oncogenic miR-34a targets, repression of miR-34a expression is likely to create a selective advantage for cancer cells, also supported by miR-34a downregulation during progressive carcinogenesis in a rat liver cancer model (108). Functional impairment of p53 downregulates miR-34a, as seen upon p53 knockdown in vivo (99). In line with this, low miR-34a expression is significantly associated with either deleted or mutated p53 in various malignancies (88, 90–92, 109–111). In lymphocytic leukemia, a SNP within the promoter of the p53 inhibitor MDM2 is associated with higher MDM2 levels and consequently lower miR-34a expression (111). In HPV-induced cervical cancer, the E6 oncoprotein destabilizes p53, resulting in miR-34a downregulation (112, 113). A p53-independent mechanism of miR-34a downregulation is seen in acute myeloid leukemia, where the C/EBPα gene, encoding a transcriptional activator of miR-34a, is mutated in 10% of the cases (93). Aberrant promoter CpG methylation is another frequent mechanism of miR-34a inhibition, reported with concomitant inhibition of expression in 79% of primary prostate carcinomas (80) and 74% of non–small cell lung cancer samples (86). Moreover, miR-34a promoter methylation was detected in melanoma (63%), colorectal cancer (74%), pancreatic cancer (64%), mammary cancer (60%), ovarian cancer (62%), urothelial cancer (71%), renal cell cancer (58%), soft tissue sarcoma (68%), chronic lymphocytic leukemia (4%), multiple myeloma (6%), and non-Hodgkin lymphoma (19%; refs. 80, 114, 115). An additional parameter affecting miR-34a expression is its genomic status, as indicated by significant association between lower miR-34a levels and 1p36 deletion in neuroblastomas (116, 117). In conclusion, miR-34a acts as a pivotal element in the p53 tumor suppressive pathway, and its recurrent downregulation by a broad range of mechanisms in various malignancies together with its inhibitory effect in cancer cell models suggests that aberrant reduction of miR-34a levels contributes to cancer development.
It has been recognized early on that 1p36 harbors genetic information mediating tumor suppression. Here, we summarize recent efforts substantiating the candidacy of CHD5, CAMTA1, KIF1B, CASZ1, and miR-34a to be 1p36 genes contributing to tumor development when reduced in dosage. Cancer-specific mutations of these candidates are infrequent but might play a role in a subset of tumors, as exemplified by rare KIF1B mutations impairing its in vitro function and segregating with cancer predisposition (65). However, the candidate genes seem to be impaired mainly on the transcriptional rather than the genetic level. Each of them was reported to be aberrantly downregulated in one or more cancer entity, and lower levels were associated with advanced disease and/or poor patient survival. Compensation of their dosage in cancer models via ectopic expression inhibited features of malignancy, including tumorigenicity in xenografts, and low expression of all candidates was significantly associated with 1p deletion in at least one tumor type. This may indicate that their dose reduction via single-copy loss compromises tumor suppressive functions and promotes cancer, as also seen for a growing number of haploinsufficient tumor suppressor genes (30).
Multiple tumor suppressive genes simultaneously downregulated via genomic deletion might cooperate in an additive or synergistic way, and targeting more than one component of a pathway or regulatory loop could be a mechanism of circumventing redundant backup mechanisms that compensate for the loss of single genes. Even within the limited set of 1p36 genes discussed here, potential interactions are found. It is tempting to speculate that a repressive effect of MYCN/MYC on KIF1B via BMI1 (70) may be inhibited by mir-34a impairing its direct target MYCN/MYC, thereby leading to an indirect activation of KIF1B by miR-34a. In this scenario, 1p36 deletion would affect KIF1B both by direct copy number loss and downregulation of its putative activator miR-34a. This cascade can be extended by adding p73, another 1p36 tumor suppressor candidate that has been extensively reviewed elsewhere (118, 119). p73 drives expression of miR-34a (120), so that 1p36 deletion would target an p73/miR-34a/MYC(N)/BMI1/KIF1B axis at 3 levels (Fig. 1, gray box). Another interaction level may be convergence of signals downstream of 1p36 genes on identical pathways. In an adequate cellular context, CHD5, CAMTA1, CASZ1, and miR-34a all activate genetic programs implicated in neuronal function and differentiation (31, 55, 75, 120). Expression of neuron-specific gene sets is associated with a markedly better prognosis in both neuroblastomas and gliomas (68, 121, 122). A simultaneous dosage-dependent impairment of the proneural regulators CHD5, CAMTA1, CASZ1, and miR-34a via 1p36 deletion could shift the transcriptome toward dedifferentiation, thus, contributing to neural tumor development. Taken together, dosage-dependent 1p36 genes are likely to interact on more than one level to suppress malignancy.
Considering that most 1p deletions in human tumors extend beyond 1p36 and that a certain fraction of genes reflects the copy number loss on the expression level, genes from other 1p regions may contribute to tumor suppression. Array-based expression analyses of 1p genes in oligodendrogliomas and neuroblastomas detected considerable copy number-dependent expression (123, 124). In neuroblastomas, 15% (124), 31% (125), and 61% (126) of all 1p genes were expressed significantly lower in 1p-deleted tumors compared with tumors retaining 1p, being in favor of a strong impact of heterozygous loss on the 1p transcriptome. Thus, repression of CHD5, CAMTA1, KIF1B, CASZ1, and miR-34a via genomic loss in an individual tumor is likely to be accompanied by downregulation of a set of other, potentially cancer-relevant genes, depending on extension and nature of the 1p deletion. Other 1p genes could contribute to tumor suppressive functions either by interaction with 1p36 genes or via independent routes. An example for an interregional interaction of proximal 1p genes with 1p36 genes can be proposed for CASZ1 regulation. EZH2, a potent suppressor of the CASZ1 locus (79) is a direct target of the 1p31.3-encoded miR-101 (127), and a large terminal deletion including 1p31 would affect both CASZ1 and its indirect activator, miR-101 (Fig. 1, gray box). A model where genes mapping to proximal 1p add to the tumor suppressive effect of 1p36 genes is further supported by the observation that neuroblastomas with large terminal deletions are more aggressive than neuroblastomas with small deletions confined to 1p36 (128). Interaction of cancer-relevant genes whose expression follows copy number aberrations is certainly not limited to 1p, considering that 1p deletion is associated with other genomic aberrations in most tumors. Expression of a substantial fraction of genes is altered consistently with the underlying genomic changes in a variety of malignancies, including colon cancer (129), prostate cancer (130), glioblastoma (131), neuroblastoma (125, 126), and multiple myeloma (132). An interchromosomal interaction of dosage-dependent genes can also be illustrated by the p73/miR-34a/MYC(N)/BMI1/KIF1B axis. Besides KIF1B, BMI1 suppresses CADM1 (70), a candidate for haploinsufficient tumor suppression mapping to 11q23. This region is recurrently lost in a broad range of solid tumors and hematological malignancies (133). Thus, the putative interaction cascade can be extended to p73/miR-34a/MYC(N)/BMI1/KIF1B-CADM1 and, accordingly, might be targeted by copy number–dependent gene deregulation via 1p36 deletion (p73, miR-34a, KIF1B), 11q23 deletion (CADM1), and/or MYC(N) amplification (Fig. 1, gray box). Collectively, in contrast to tumor suppressor inactivation by copy number-neutral gene-specific mutations, copy number–dependent downregulation of 1p36 candidates is likely to be accompanied by deregulation of a considerable set of other cancer-related genes. These may be either targeted by the same event (1p deletion) or genomic aberrations associated with 1p deletion. However, a large fraction of genes may passively reflect copy number change on the expression level without contributing to cancer development. Survival analyses adjusting for the respective copy number alteration may clarify whether the prognostic value of a gene's expression profile is independent of the underlying genomic alteration, as seen for CHD5 and CAMTA1 (32, 46). Such prognostic independency suggests that some tumors evolve additional, copy number–independent mechanisms to regulate these genes, and identifying such mechanisms may strengthen the position of the respective candidates.
A range of inhibitory mechanisms other than copy number loss were identified for the 1p36 candidates discussed here. Mir-34a is frequently downregulated by inactivation of upstream transcription factors, including p53 (88, 90–93, 109–111). Aberrant promoter CpG methylation for miR-34a and CHD5 was observed in a broad range of tumors (37, 39, 40, 80, 86, 114, 115). Epigenetic mechanisms acting via histone modifications have also been linked to impairment of tumor suppressive activities (134) and are likely to play a role in regulating CAMTA1 (55), CASZ1 via EZH2 (79), KIF1B via BMI1 (70), and CHD5, acting as a histone-interacting chromatin remodeler itself.
In conclusion, the 1p36 genes CHD5, CAMTA1, KIF1B, CASZ1, and miR-34a may not necessarily require biallelic inactivation in a classic “two-hit” manner but contribute to cancer development by partial dosage reduction via copy number loss or other mechanisms, including epigenetic inhibition. Functional studies indicate that these candidates cooperate to suppress tumorigenesis. Their codeletion may be one way for a developing cancer cell to acquire selective advantage by inhibiting an antioncogenic network at different positions in a single event. Aberrant expression of other dosage-dependent cancer-relevant genes on 1p or from associated copy number alterations in other chromosomes is likely to further contribute to this selective advantage. In such a setting, deletion mapping and SRO identification will not lead to identification of a single tumor suppressor gene that is completely inactivated by a second hit, but may guide the identification of a minimum gene set, whose reduction is required for tumorigenesis in a certain cellular context. Human genetics alone can deliver definitive proof of biallelic inactivation of a classic tumor suppressor gene, but unequivocal support for the dosage dependency of a tumor suppressor requires other lines of evidence, including mouse models. An elegant example of a tight correlation between tumor suppressor expression level and its functionality was shown by the generation of an allelic series of genetically engineered mice expressing varying levels of the Pten tumor suppressor (135). Surprisingly, even in animals with a subtle 20% reduction of the normal Pten level, tumor incidence was increased and survival was decreased, with the wild-type allele remaining fully functional in all induced tumors (135). A similar mouse-based assay may clarify whether a comparable dosage dependency can be modeled for the candidate genes discussed here.
In contrast to genetic inactivation of tumor suppressor genes by mutation, inhibitory mechanisms acting on the expression level are in principle reversible as long as an intact allele is present. This may pave the way for intervention strategies that restore expression of dosage-sensitive tumor suppressor genes. The genes discussed in this review are targeted by aberrant epigenetic mechanisms, at least in a subset of tumors, and recent advancements in depicting the enzymatic processes controlling the cancer epigenome should open doors for developing new therapeutic approaches (134).
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Conception and design: K.-O. Henrich, M. Schwab, F. Westermann
Development of methodology: K.-O. Henrich, M. Schwab
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): K.-O. Henrich, M. Schwab, F. Westermann
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): K.-O. Henrich, M. Schwab, F. Westermann
Writing, review, and/or revision of the manuscript: K.-O. Henrich, M. Schwab, F. Westermann
Study supervision: K.-O. Henrich, M. Schwab, F. Westermann
BMBF: NGFNPlus #01GS0896 (K.-O. Henrich, M. Schwab, and F. Westermann), MYC-NET, CancerSys #0316076A (F. Westermann), EU (FP7): ASSET #259348 (F. Westermann).