The Quest for the 1p36 Tumor Suppressor

Genomic analyses of late-stage human cancers have uncovered deletions encompassing 1p36, thereby providing an extensive body of literature supporting the idea that a potent tumor suppressor resides in this interval. Although several genes have been proposed as 1p36 candidate tumor suppressors, convincing evidence that their encoded products protect from cancer has been scanty. A recent functional study identified chromodomain helicase DNA-binding protein 5 (CHD5) as a novel tumor suppressor mapping to 1p36. Here, we discuss evidence supporting the tumor-suppressive role of CHD5. Together, these findings suggest that strategies designed to enhance CHD5 activity could provide novel approaches for treating a broad range of human malignancies. [Cancer Res 2008;68(8):2551–6]

Deletions of the short arm of human chromosome 1 were first reported in neuroblastoma in 1977 (1), presaging a flurry of reports indicating that the region referred to as 1p36 is frequently deleted in a broad range of human cancers, including those of neural, epithelial, and hematopoietic origin (Fig. 1; refs. 2–22). These findings suggest that loss of a tumor suppressor mapping to this region is a critical event in tumorigenesis. Because such a diverse array of human cancers has been associated with deletion of 1p36, loss or inactivation of this region might be a common initiating event for several malignancies. The search for the 1p36 tumor suppressor has led to some exciting candidates (Fig. 1). Although some of these proposed tumor suppressors have tumor-protective capabilities in specific cellular contexts, none can account for the wide range of tumor types that have been associated with the 3 decades of literature documenting 1p36 deletions. Hence, the search for the 1p36 tumor suppressor has continued.

Why has it been so difficult to identify the tumor suppressor in this region of the genome? One reason could be that more than one 1p36 tumor suppressor exists. Support for this idea is that although the prevalence of tumor-derived deletions attests to 1p36 loss as being an important event in cancer, these deletions do not always overlap and therefore fail to identify a single genomic interval (Fig. 1). Perhaps distinct tumor suppressors function tissue specifically; this might be the reason that the various malignancies in which 1p36 deletions have been documented cannot be explained by any of the individual candidate tumor suppressors described to date. A feature of most late-stage tumors with 1p36 deletions is that the interval involved is typically extremely large, often encompassing the entire short arm; therefore, informative tumors that might reveal the location of the tumor suppressor and lead to its identity are rare. One idea is that 1p36 is inherently unstable. Another possibility is that several 1p36 tumor suppressors work together and, therefore, that loss of a combination of these genes is a prerequisite for cancer. Because hereditary cancer predisposition syndromes do not exist for this genomic region, linkage-based methodologies that were so crucial for discovering tumor suppressors such as retinoblastoma, are not helpful for identifying tumor suppressors in this interval. In short, 1p36 has presented a major challenge for the cancer community.

The underlying assumption for the approach that we took was that there was a region of 1p36 that when deleted would predispose to cancer. Therefore, increased tumorigenicity would provide the functional readout needed to reveal the location of the tumor suppressor. Chromosome engineering, a Cre/loxP-based methodology (23) reviewed by Mills and Bradley (24), was used to systematically delete regions of the mouse genome corresponding to human 1p36 (25). This seemed a logical approach because genes are conserved between mouse and man in this region of the genome, with genes present on human 1p36 mapping to distal mouse chromosome 4 (Fig. 2). This strategy identified a tumor-suppressive interval that when deleted caused enhanced proliferation, loss of contact inhibition, and spontaneous immortalization in cultured cells, hallmarks of increased tumorigenic potential. These cellular phenotypes culminated as cancer in vivo; mice heterozygous for this deletion developed spontaneous tumors, including many types associated with deletion of the corresponding region in humans. Further functional evidence that this region harbored a potent tumor suppressor was garnered when it was discovered that mice engineered to have an extra copy of this region had excessive tumor suppression. Three copies of this interval caused an increase in the tumor-suppressive mechanisms of cellular senescence in cultured cells and markedly enhanced apoptosis in vivo. This combined loss-of-function and gain-of-function approach identified a potent tumor-suppressive region corresponding to human 1p36.

Although a tumor-suppressive region of mouse chromosome 4 had been identified, this genomic interval was 4.3 Mb and encompassed 52 genes (25). Therefore, the next phase of this work was to identify the tumor suppressor in the interval. It had been shown that cells with an extra copy of the region proliferated very poorly in culture and that this proliferative defect was rescued by “correcting” dosage of the interval (i.e., cells with an extra copy of the region on one chromosome and a deletion of the same interval on the other chromosome were effectively diploid for that region). Mice carrying both a deletion and a duplication of the 4.3-Mb interval were viable and fertile, and cells from these animals proliferated at rates comparable with those of wild-type cells. Therefore, Occam's razor—the principle that the simplest solution is usually the best one—posed the possibility that cells with an extra copy of the region proliferated poorly because they had increased expression of a single gene within the interval. This meant that the tumor suppressor could be identified by systematically reducing expression of candidate genes in cells with an extra copy of the interval. The assay was simply looking for the gene that restored proliferation when its dosage was reduced. Genes were prioritized based on gene ontology terms, and RNA interference (RNAi) was then used to reduce expression of the top candidates in the region (25). Knockdown of all but one gene tested failed to rescue the proliferative defect. For example, knockdown of p73, Dnajc11, and Camta1, genes within the interval that had previously been proposed as 1p36 tumor suppressor candidates, failed to enhance proliferation, thereby excluding these genes as being responsible for the tumor-suppressive phenotypes observed. In contrast, two different hairpins against chromodomain helicase DNA-binding protein 5 (Chd5) effectively restored proliferation to these poorly proliferating cells and even enhanced proliferation of wild-type cells. This identified Chd5 as the tumor suppressor candidate in the region.

Human CHD5 had previously been proposed to function as a chromatin remodeling protein based on its homology with members of the chromodomain superfamily of proteins (26). CHD5 is one of nine members of the CHD family, named because of their unique combination of chromatin organizing modulator, helicase, and DNA-binding domains. The CHD family can be further subdivided; CHD5 is similar to CHD3 and CHD4 in that it is endowed with plant homeodomain motifs (27). Interestingly, CHD3 and CHD4 are components of the nucleosome remodeling complex, an assembly of proteins that remodels chromatin by sliding nucleosomes out of the way, thereby providing polymerase with the access needed to activate gene expression. Other than this homology between CHD5 and previously described chromatin remodeling proteins, very little functional data existed for this protein. How might this hypothesized chromatin remodeling protein function to prevent cancer?

The tumor-suppressive region modulates p16^Ink4a/Rb- and p19^Arf/p53-mediated pathways. To explore the mechanism whereby Chd5 protects from tumorigenesis, models with loss and gain of the region encompassing Chd5 were first examined (25). The first clue was that the phenotypes of poor proliferation, enhanced cellular senescence, and augmented apoptosis that were caused by just one extra copy of the 4.3-Mb interval were dependent on p53. When p53 was reduced, either by RNAi-mediated knockdown in cultured cells or by germ-line deficiency in vivo, the phenotypes of exacerbated tumor suppression disappeared completely; cells proliferated normally or mice were viable and fertile. This showed that the tumor suppressor in the region positively regulated p53.

An important advantage of the chromosome engineering approach is that the consequence of gain and loss of the same genomic interval can be directly compared and contrasted (24). Therefore, the phenotypes resulting from incremental changes in dosage (zero, one, two, three, and four copies) can be analyzed, something that cannot be easily done using conventional mouse models. Because an extra copy of the region encompassing Chd5 intensified p53 function, it was predicted that cells with the corresponding deletion would have compromised p53 (25). Cells with the deficiency expressed reduced p53 at the protein (but not the transcript) level, showing that the tumor suppressor was regulating p53 posttranscriptionally. Despite this reduction in p53 expression, cells heterozygous for the deficiency were as efficient as wild-type cells in their ability to induce p53 in response to DNA damage, indicating that DNA damage-induced pathways upstream of p53 were intact. This suggested that oncogene-induced pathways that mediate p53 function were compromised by deletion. Indeed, cells heterozygous for the 4.3-Mb region were susceptible to transformation by oncogenic Ras, further substantiating their p53-compromised status.

Having implicated p53 in the tumor-suppressive effect, the mechanism whereby p53 was regulated was next explored (25). Because deficient cells were sensitized to oncogenic transformation and oncogenic stimuli are known to function through p19^Arf (28), we asked whether p19^Arf was compromised when the tumor-suppressive region was deleted. Indeed, compared with wild-type cells, p19^Arf expression was reduced at both the transcript and the protein level in cells with the deficiency but was expressed more highly in cells with an extra copy of this interval. Furthermore, p19^Arf depletion significantly enhanced proliferation of cells with an extra copy of the region, and both the extent and time course of rescue by p19^Arf reduction were essentially identical to that caused by depletion of Chd5 itself. This clearly implicated a second tumor suppressor, p19^Arf, in the tumor-suppressive phenotype.

p19^Arf is encoded by the Ink4/Arf locus, which also encodes a distinct tumor suppressor: p16^Ink4a. As was the case for p19^Arf, p16^Ink4a was compromised at both the transcript and the protein level in cells with the deletion but was expressed at higher levels in cells with an extra copy of this region (25). But in contrast with the ability of p19^Arf depletion to restore proliferation in cells with three copies of the interval, p16^Ink4a depletion caused only a modest increase in proliferation that was not statistically significant. This indicated that, in the context of these cultured cells, the phenotype caused by an extra copy of the region was dependent on p19^Arf, whereas p16^Ink4a was dispensable. Because p19^Arf inhibits Mdm-2–mediated degradation of p53, the finding that p19^Arf expression correlated with dosage of the tumor-suppressive interval provided a mechanistic explanation for how p53 was being modulated.

Chd5 depletion phenocopies deletion of the tumor-suppressive interval. The chromosome engineering strategy provided a powerful approach for functionally mapping the tumor-suppressive region. But because the rearrangement encompassed a large interval (hence allowing the interrogation of multiple loci simultaneously), it was important to determine which of the genes were responsible for the observed phenotype. In this study, it was essential to show that Chd5 depletion alone accounted for the phenotype characteristic of the 52-gene deletion (25). Importantly, for every assay performed, wild-type primary cells in which Chd5 was knocked down closely phenocopied cells with an engineered heterozygous deficiency of the 4.3-Mb region encompassing Chd5. For example, Chd5-compromised cells had reduced expression of p16^Ink4a, p19^Arf, and p53, and expression levels correlated closely with those of cells heterozygous for the deficiency. The finding that both p16^Ink4a and p19^Arf were cosuppressed by heterozygous deficiency of the interval encompassing Chd5 was consistent with the view that Chd5 is a chromatin remodeling protein that mediates expression of the Ink4/Arf locus.

The consequence of crippling the tumor-suppressive network in cultured cells was enhanced proliferation and susceptibility to oncogenic transformation, with the phenotype of wild-type cells in which Chd5 was specifically knocked down closely paralleling that of cells heterozygous for the engineered deficiency. Heterozygosity of the 4.3-Mb region predisposed to spontaneous tumors. To conclusively show that Chd5 deficiency was responsible for this tumor-prone phenotype, wild-type cells in which Chd5 had been depleted were compared with cells with the deletion for their ability to form tumors in the context of activated Ras in vivo. This analysis revealed that cells in which Chd5 was specifically knocked down were as tumorigenic as cells with the heterozygous deficiency: tumors grew with the same kinetics and had indistinguishable histopathology. These findings provided genetic and molecular evidence that Chd5 was the gene within the region that was responsible for tumor suppression.

Because Chd5 could account for each of the tumor-suppressive properties analyzed, it is likely that Chd5 is the sole tumor suppressor within the 4.3-Mb interval. Although it cannot be excluded that other genes in the region affect aspects of tumorigenesis not assessed in this study (e.g., angiogenesis, metastasis, or as yet undiscovered mechanisms of tumor suppression), it is clear that Chd5 deficiency is an initiating event in tumorigenesis because (a) heterozygosity of the region encompassing Chd5 predisposes to spontaneous tumors and (b) specific depletion of Chd5 cooperates with activated oncogenes to enhance tumor growth in nude mice. Despite these lines of evidence, it will be imperative to assess spontaneous tumorigenesis in Chd5 heterozygous null mice as well as to determine whether gain of Chd5 function invokes the tumor-suppressive mechanisms of senescence and apoptosis. These approaches will conclusively determine whether Chd5 is solely responsible for the phenotypes caused by loss and gain, respectively, of the 52-gene interval.

Identification of Chd5 as a tumor suppressor in the mouse agrees with the rich body of literature describing 1p36 deletions in human cancer; many of these encompass CHD5 (see Fig. 1; refs. 2–22). Our analyses revealed that CHD5 was frequently deleted in human glioma (25), and more recent reports found that CHD5 was deleted in leukemia/lymphoma (29) and neuroblastoma (see Fig. 2; ref. 30). The Brodeur group used expression analyses to prioritize between candidate genes mapping to a 2.2-Mb minimally common region of deletion (mcr) that they identified in neuroblastoma and concluded that CHD5 was the strongest candidate in the interval (30). Maser et al. (29) identified a mcr encompassing Chd5 in a mouse model of chromosomally unstable lymphoma and also showed that CHD5 is one of only three genes mapping to a mcr that they discovered in human T-cell acute lymphoblastic leukemia/lymphoma (T-ALL). Furthermore, CHD5 mutations were recently reported in human breast cancer (31) and neuroblastoma (30). Combined, these studies define a mcr that not only includes CHD5 but also excludes nearby tumor suppressor candidates such as p73, Icat, and Riz1 (see Figs. 1 and 2). Similarly, these analyses exclude miR34, a microRNA that was recently reported as being regulated by p53 (reviewed in ref. 32). It should be emphasized that although this shows that these loci are not responsible for the tumor-suppressive phenotypes addressed in the study by Bagchi and colleagues (25), it does not eliminate the possibility that these genes contribute to carcinogenesis in other cellular contexts or cooperate with CHD5 in a manner that has not been revealed to date. Together, these findings support the tumor-suppressive role of CHD5, leading to several important questions.

The observation that 1p36 deletions are typically very large has led to the speculation that more than one tumor suppressor resides in this region. Although there is substantial support for the tumor-suppressive role of CHD5, these studies do not exclude the possibility that additional 1p36 tumor suppressors exist. It should be noted that the 5.7-Mb interval that functionally identified CHD5 as a tumor suppressor comprises less than one quarter of the 24.9-Mb 1p36 region (see Fig. 2; ref. 25). Therefore, additional tumor suppressors may map to the portions of 1p36 that were not studied. Interestingly, two independent groups recently identified KIF1Bβ as a 1p36 tumor suppressor candidate (33).¹

This gene clearly maps outside of the tumor suppressive region described by Bagchi et al. (see Figs. 1 and 2). In addition, mcrs in several hematopoietic malignancies do not encompass CHD5 (see Fig. 1). Although this does not support CHD5 deficiency as being a causal event in these cancers, the finding that mice heterozygous for the 4.3-Mb interval encompassing Chd5 are prone to spontaneous lymphoma (25) and an independent report that Chd5 and CHD5 are deleted in lymphoid cancers of mouse and man, respectively (29), provide evidence that CHD5 deficiency plays a key role in these hematopoietic malignancies. Generation of a series of chromosome-engineered deletion/duplication mouse models that in sum span the entire region corresponding to human 1p36 could be used to functionally address the issue of whether additional 1p36 tumor suppressors exist. Another approach would be to more fully evaluate spontaneous tumorigenesis in Chd5-compromised mice and to compare these tumors with those associated with 1p36 deletions in humans. Can loss of the region encompassing Chd5 account for each of the tumor types that have been associated with 1p36 deletion? Mice heterozygous for the engineered deletion develop lymphoma and squamous cell carcinoma, two of the three major tumor types in which 1p36 deletions have been reported. Whereas CHD5 loss has been documented in human glioma (25) and neuroblastoma (30), neural-associated malignancies have not yet been reported in mice heterozygous for the deficiency. Whether neural tumors develop in Chd5-compromised mouse models warrants further investigation. Approaches that directly assay for cooperation between CHD5 and previously reported 1p36 candidate tumor suppressors may also be used to test the idea that multiple tumor suppressors reside in this region of the genome.

Although it is clear that heterozygous deletions encompassing Chd5 predispose to cancer, there is currently no evidence that the remaining wild-type Chd5 allele is lost or mutated in the tumors that develop. In the small number of tumors analyzed to date in mice with a deficiency encompassing Chd5, as well as in cells from these mice that spontaneously immortalize, the Chd5 allele is retained, suggesting that complete loss of Chd5 is not a prerequisite for tumorigenesis (25). This idea is consistent with results of an independent study in human neuroblastoma (30). In addition, CHD5 mutations in human breast cancer (31) and in neuroblastoma (30) are heterozygous, with the wild-type CHD5 allele being retained. CHD5 status needs to be examined in additional types of human cancers to more fully appreciate the scope of CHD5 loss in various malignancies and to assess whether retention of the wild-type CHD5 locus is cancer specific. This endeavor could be guided by the spontaneous tumor spectrum of mice with compromised Chd5, especially given the recent report that modeling cancer in the mouse can accurately reflect genomic alterations in human cancer (29). CHD5 expression should be examined at the protein level to determine whether the CpG-rich CHD5 promoter is silenced by methylation. If it holds that CHD5 function is partially retained, this would represent a paradigm shift in the current view of carcinogenesis. Perhaps, this would provide an explanation for why studies relying on complete loss as a criterion for tumor suppressors did not identify CHD5.

Although mutation of one copy of CHD5 accompanies tumorigenesis (30, 31), it is not currently known whether these mutations are driver mutations or whether they are passenger mutations that simply reflect the genomic instability of end-stage cancers. It is not immediately apparent how these tumor-derived mutations thwart the tumor-suppressive function of CHD5, as the affected amino acids are not within any of the known functional domains. Do these mutations affect the ability of Chd5 to epigenetically regulate expression of the Ink4/Arf locus? Although Chd5 is proposed to function as a chromatin remodeling protein, there is currently no functional evidence that it directly binds to or modulates Ink4/Arf. Assessing the motifs of CHD5 required for tumor suppression and elucidating the biochemical properties of CHD5 may provide a foundation for designing strategies for effectively modulating its tumor-suppressive function.

Although heterozygosity of CHD5 predisposes to cancer, additional cooperating events undoubtedly contribute to this process. Chd5 facilitates both p16^Ink4a/Rb- and p19^Arf/p53-mediated pathways (25). Although p16^Ink4a is dispensable for regulating proliferation in the context of cultured cells, it is intriguing that the very first tumor observed in mice heterozygous for the deficiency was a squamous cell carcinoma of the skin, a tumor type that does not develop in p53-compromised mouse models. Therefore, it is likely that p16^Ink4a/Rb-mediated pathways affect the tumor-suppressive effects of Chd5 in vivo. The ability of Chd5 to modulate the Ink4/Arf locus was evident because Chd5-compromised cells were sensitized to oncogenic transformation. However, it is likely that Chd5 regulates chromatin structure globally, as was elegantly shown for Bmi-1 (33). Therefore, additional genes that are modulated by Chd5 could be discovered using a combination of expression and chromatin immunoprecipitation microarrays using cells/tissues with altered dosage of Chd5. We propose that a partial crippling of this tumor-suppressive network culminates in cancer, and predict that combined heterozygosity of Chd5 and other members of the tumor-suppressive network that is modulated by Chd5 leads to carcinogenesis. This hypothesis can be directly tested by assessing whether loss or inactivation of Chd5 cooperates with a compromise in p16^Ink4a/Rb- and p19^Arf/p53-mediated pathways or in combination with compromised function of additional members of the Chd5-modulated network that are yet to be discovered. Additionally, in light of the observation that Chd5 deficiency renders cells sensitive to oncogenic transformation and that 1p36 deletions in neuroblastoma frequently have amplification of N-MYC, it will be important to assess whether CHD5 deficiency cooperates with activated oncogenes in human cancer.

The above findings represent the initial stages of investigating the role of CHD5 in cancer. It will be important to assess the role of CHD5 in diverse types of human cancer and to more fully elucidate the mechanism by which the tumor-suppressive effects are thwarted. In addition, biochemical studies designed to interrogate the mechanistic basis of the tumor-suppressive role of CHD5 are needed. Finally, it will be important to determine whether this chromatin remodeling protein affects gene expression globally. These studies are likely to have a significant impact on our understanding of tumorigenesis that could ultimately pave the way for better cancer diagnostics and pharmaceutics.

We thank Hannes Vogel, Markus Bredel, and Cristian Papazoglu for contributing to the work discussed in this review.