Pathogenesis and Consequences of Uniparental Disomy in Cancer Free

Chromosomal aberrations constitute a hallmark of a cancer genome. Recurrent balanced chromosomal abnormalities, such as those found in distinct types of leukemia and lymphoma, can be diagnostic and often explain the pathogenesis of such conditions. Chromosomal defects also serve as excellent clonal markers and are essential for the diagnosis of a malignant clone or the detection of minimal residual disease or relapse, especially for cancers that arise in the hematopoietic system (1–3). However, the frequent complexity of chromosomal defects and inability to obtain viable cells make it difficult to apply diagnostic cytogenetic testing in solid tumors. Consequently, relatively few chromosomal defects that can function as diagnostic or prognostic markers have been identified to date, although this has begun to change in recent years. Loss of heterozygosity (LOH) owing to segmental or numerical chromosomal deletion is particularly important and is one of several paradigms of malignant transformation, including the concept of tumor suppressor gene inactivation and Knudson's 2-hit hypothesis (4). After the loss of chromosomal materials containing one allele, the remaining allele can be affected by somatic mutation or harbor a disease-prone polymorphic variant. Similarly, loss of chromosomal material can lead to LOH and the conversion of a heterozygous inherited (potentially functionally silent) mutation to a hemizygous mutation. However, the discovery of uniparental disomy (UPD), also referred to as copy-neutral LOH, suggests that LOH may not necessarily be due to the loss of chromosomal material. Under normal circumstances, each autosome has 2 copies (paternal and maternal) that carry discrete differences encoded by single nucleotide polymorphisms (SNPs), and these differences can be used to distinguish between them. In regions of UPD, portions of one of the chromosomes are lost and replaced by the exact copy of the remaining chromosome (either paternal or maternal), resulting in the retention of 2 copies of genetic information but the loss of polymorphic differences that existed due to the presence of maternal and paternal genes in this region of a diploid chromosome set. Because of the lack of change in the copy number, UPD remains undetected by metaphase cytogenetics. However, studies using microsatellite analysis or genotyping of sequential SNPs combined with copy number determination have shown that various types of UPD frequently occur in the cancer genome (5, 6). In this review, we discuss the genesis and types of UPD seen in malignant and normal cells. We focus on mechanisms of selective growth advantage that can result from this lesion and discuss their mechanistic role in cancer pathogenesis. We also review the recurrent regions of UPD identified in various forms of cancer and the clinical implications associated with these defects.

Until recently, LOH has been most consistently linked to deletions of chromosomal material in somatic cytogenetic defects encountered in cancer. In contrast, UPD has been identified through the study of inherited diseases, because UPD can occur as a germline lesion leading to isodisomy. Heterodisomy, another possible outcome of germline UPD, does not result in LOH (7). Inherited UPD was first described by Engel (8) in 1980. It can affect whole chromosomes or fragments of chromosomes, and can be interstitial or telomeric (Fig. 1). Principally, UPD corresponds to a duplication of either paternal (unipaternal disomy) or maternal (unimaternal disomy) alleles, and thus to homozygosity for germline allelic variants. LOH due to deletion results in hemizygosity, whereas UPD results in homozygosity (Fig. 1). Theoretically, it is also possible that trisomy is associated with LOH in the form of uniparental trisomy, which is invariably related to numerical aberrations (Fig. 1). Conceptually, any trisomy might represent a form of UPD without LOH, as both parental alleles are retained while one is duplicated.

In contrast to constitutional UPD, the genesis of somatic UPD is not well understood. However, it may be a result of mitotic recombination or a successful attempt to correct for the loss of chromosomal material by duplicating the remaining allele (Fig. 2).

The chromosomal mechanics behind UPD have been intensely investigated in embryology and inherited conditions. In consanguineous populations, homozygosity is frequent and can accumulate in the form of stretches of autozygosity. Although such changes do not always result in pathology, autozygosity represents a risk for inherited disease, including genetic predisposition to various cancers (9). In contrast to autozygosity, germline UPD can arise as a result of mistakes in meiosis, in which all cells in the organism contain the change, or in the initial mitoses after fertilization, resulting in tissue mosaicism. Autozygosity and meiotic UPD are not distinguishable without a pedigree analysis (Fig. 3). Germline UPD can be due to meiotic chromosomal missegregation and subsequent mitotic reassortment leading to a balanced genome. Various scenarios can lead to germline meiotic UPD, and heterodisomy and isodisomy (resulting in LOH) have to be distinguished from numerical chromosomal defects. Trisomic rescue following errors in meiosis I or II can result in UPD as one of the possible outcomes (Fig. 3) (10).

Unlike germline UPD and autozygosity, somatic UPD results from mitotic homologous recombination events, or it may represent an attempt to correct for the unbalanced loss of chromosomal material by using the remaining alleles as a template (Fig. 2). Numerical somatic UPD can occur as a result of mitotic errors, including nondisjunction or loss of a chromosome due to anaphase lag followed by reduplication of the remaining chromosome (11). Segmental telomeric UPD may be due to mitotic homologous recombination between highly homologous, low-copy-number repeats (12, 13). However, such a mechanism would be more difficult to invoke for segmental interstitial UPD because it would require 2 consecutive or simultaneous homologous recombination steps. It is possible that segmental UPD results from initial deletion followed by a compensatory reduplication of the remaining chromosomal fragment (Figs. 2 and 3). For diagnostic and investigational purposes, the ability to distinguish between a somatic, clonal UPD and germline UPD or autozygosity is of the utmost importance.

The pivotal description of UPD in a hematological malignancy came from a study of polycythemia vera (PV) (14), in which UPD9p was found in 33% of patients, constituting the most common chromosomal lesion in this disease (15). Later this chromosomal defect was linked to a JAK2V617F mutation (16). More-comprehensive studies demonstrated that JAK2V617F mutations with UPD9p can also be found in other myeloproliferative neoplasms (MPNs) (17, 18). For example, primary myelofibrosis (PMF) reveals a high frequency of UPD9p with JAK2V617F mutations (44%) (19). However, the homozygous mutational burden varies because of differences in the population size of mutant cells. Even in purified myeloid cell populations, heterozygous and homozygous cells can be found. Moreover, patients with essential thrombocytosis (ET) exhibit a lower frequency of UPD9p, and the resultant JAK2V617F mutational burden in ET is low compared with PMF and PV (16, 20).

Systemic application of SNP arrays as a karyotyping tool (Fig. 4) led to further discoveries of recurrent regions of UPD in various myeloid and lymphoid malignancies, with secondary acute myeloid leukemia (sAML), myelodysplastic syndrome (MDS) or MPN, and chronic myelomonocytic leukemia (CMML) showing particularly high frequencies of this type of chromosomal lesion. Investigations of solid tumors produced comparable results with the identification of recurrent areas of acquired UPD in a wide spectrum of cancers, some of which show a remarkable predilection for this type of chromosomal defect (Table 1).

Tumor-specific recurrent regions of UPD can be mapped for AML and MDS (Fig. 5), as well as for a representative collection of cell lines derived from various malignancies (Supplemental Fig. 1). Many of the commonly affected areas contain important genes that have been implicated in malignant pathogenesis. It should be noted that many studies probably overestimated the frequency of somatic UPD owing to the lack of distinction from germline-encoded UPD. Nevertheless, the somatic nature of these alterations can be clearly deduced from the size and location of the reported regions of UPD and their recurrence. The impact that specific regions of UPD have in terms of prognosis or diagnosis is currently being evaluated. For example, UPD7q, UPD11q, and UPD17p have been linked to poor outcomes in myeloid malignancies (21–23).

The fact that certain cancers, such as MUTYH-associated polyposis colon carcinomas (24), display a higher frequency of somatic UPD and also can accumulate multiple areas of UPD (complex UPD) implies that there is an inherited or acquired predisposition to this type of defect owing to the presence of fragile sites prone to recombination or a specific type of chromosomal instability. Particularly high frequencies of somatic UPD have been described for some malignancies, suggesting that this type of defect may be related to pathological pathways that are common in some cancers but absent in others. For instance, in sporadic colon cancer, physical loss of chromosomal material is characteristic and UPD is less common (25).

Preferred sites of mitotic recombination leading to UPD, with a clustering of the centromeric and telomeric breakpoints, have been identified (12). In a study of mantle cell lymphoma (MCL), the breakpoints flanking all of the genomic UPDs were significantly associated with genomic regions enriched in copy number variants and segmental duplications, suggesting that recombination in these regions may play a role in the genomic instability of MCL (26). Similarly, a careful analysis of the sites of acquired UPD origin in low-risk MDS showed that 43% of UPD regions were localized to or formed part of a previously identified fragile site (27). Fragile sites correspond to known locations of frequent genomic instability and are associated with breakpoints occurring in the generation of chromosomal aberrations in hematological malignancies (28). Fragile sites have also been associated with regulatory micro RNA amplifications and deletions (29).

Risk factors for the acquisition of UPD also include the presence of BRCA mutations, which have been associated with an increased frequency of UPD that is not observed in cases of spontaneous breast cancer (30). In ovarian cancer, UPD is frequently observed in tumors with an inherited BRCA mutation (31). Microsatellite instability (MSI) has been shown to be associated with an increased frequency of UPD (32). In one study (33), MSI was present in 60% of patients who had AML and regions of UPD (Fig. 5), whereas single-locus MSI was absent in patients with AML in whom UPD was not detected.

Although it is likely that chromosomal deletions occur randomly, those that result in a proliferative advantage or resistance to, e.g., physiological apoptosis could initiate clonal outgrowth. Selection for clones with a specific region of LOH could be related to a somatic or germline loss of a wild-type allele, resulting in hemizygosity for an SNP-encoded, disease-prone allele or a somatic or germline mutated allele (Fig. 1). If the affected area includes promoters of alleles that are differentially silenced (imprinted), deletion can lead to either a gain of imprinting (GOI) or loss of imprinting (LOI). This can result in changes in gene expression. UPD can also lead to the duplication of an imprinted expressed allele or a silenced (methylated), imprinted allele. When the transcription of both alleles is required for normal cellular physiology, deletions can result in pathological haploinsufficiency, and thus LOH is less likely to play a pathogenic role (Fig. 1).

There are similarities and important differences between the consequences of LOH due to deletion or UPD. UPD may convey a permissive growth advantage when, in accordance with the 2-hit hypothesis, an initial heterozygous mutation is duplicated through UPD. This may result in homozygosity of a somatic mutation that inactivates a tumor suppressor gene, such as occurs with TP53 in UPD17p, RUNX1 in UPD21q, and many others (Fig. 6). Activating oncogenic mutations can be duplicated through UPD, leading to increased proliferative drive though a double dose of the mutated gene product. Such a scenario has been encountered with JAK2 (UPD9p) (14, 16, 34), FLT3 internal tandem duplication (UPD13q) (35, 36), WT1 (UPD11p) (37, 38), and MPL (UPD1p) (19, 39). Recently, we and other groups found loss-of-function mutations of EZH2, encoding trimethyltransferase of H3K27, in patients with UPD7q (Fig. 6) (40–42). Because methylation of H3K27 is a histone repressive mark associated with condensation of chromatin, its loss-of-function mutation results in chromatin relaxation and accelerated gene transcriptions, as in WNT1 and HOXA family genes. EZH2 mutations are more commonly homozygous (UPD7q) than heterozygous (40, 42). In myeloid malignancies, UPD is frequently observed on chromosome 11q specifically in the MDS/MPN phenotype. CBL mutations were observed in 76% of patients with UPD11q, but were relatively rare (<5%) in patients with deletion or without LOH 11q. In mutations associated with recurrent UPD, homozygosity may provide a malignant clone with a further growth advantage. Theoretically, a similar effect could be produced by LOH as a result of deletion, but for some genes, such as CBL, most mutations are homozygous, and corresponding deletions have not been found to harbor hemizygous mutations (22, 43, 44). Consequently, the CBL knockout is less leukemogenic than ring finger domain mutant knockin CBL null mice (43, 45). By contrast, TP53 or TET2 mutations are associated with both deletions and UPD. UPD can also affect germline heterozygous mutations. Examples of such mutations include UPD11q and UPD17q leading to duplication of CBL and NF1 mutations in juvenile myelomonocytic leukemia (JMML) (Fig. 6) (46–48).

Somatic UPD can also lead to duplication and hence homozygosity of disease-prone SNPs that are silent in a heterozygous configuration. Typically, such SNPs show an exceeding low frequency of homozygosity for the minor allele in the general population. Of note, however, although duplication of initially heterozygous mutations is a driving force for clonal dominance, areas of UPD contain a large number of genes that can include germline polymorphisms and imprinted sites. For example, CNTLN, KANK1, DMRT1, TOPORS, and MLANA genes have been reported to be either imprinted or differentially methylated on chromosome 9p, which is frequently affected by UPD. Therefore, UPD at this site can lead to GOI and LOI for specific genes (49–52). In addition, UPD9p contains a large number of nonsynonymous polymorphisms for which either minor or major alleles can be duplicated and result in discrete changes of the phenotype. These findings indicate that association of the high allelic burden of JAK2V617F mutation due to UPD9p may be influenced by homozygous SNPs and/or loss or gain of expression of monoallelically expressed genes in the region affected by UPD (53). More recently, multiple groups described the relationship between JAK2's genetic predisposition and JAK2V617F, and reported that the 46/1 JAK2 haplotype predisposes to the development of JAK2V617F-associated MPN (54–56). Subsequently, Tefferi and colleagues (57) observed that a JAK2 germline genetic variation (rs12343867 genotype CC) was less frequent in PMF with a high JAK2V617F burden. This suggests that the allelic distortion from acquired UPD contributes to the appearance of a more pronounced effect on disease susceptibility in JAK2V617F-positive patients.

Examples of changes in LOI and GOI can also be found in cancer-prone inherited disorders associated with UPD, such as GOI of H19 and LOI of IGF2 (58, 59). A similar alteration of imprinting patterns has been found in hepatoblastoma, a tumor characterized by frequent UPD11p affecting H19 and IGF2 (60). LOI for IGF2 and H19 due to UPD is evident in colon carcinoma (61) and Wilms' tumor (WT) (62), ARH1 LOI is evident in ovarian and breast cancer (63), and H19 LOI is evident in AML (64). Some of these events may be due to a shared mechanism of UPD. Thus, it is possible that although deletion or duplication can randomly affect each parental chromosome, clonal selection may favor the expressed or silenced imprinted allele and thus may not be random. In MDS, for example, the FZD9 promoter has been found to be consistently hypermethylated in patients with LOH7q as a result of UPD or chromosomal deletion (65). Theoretically, several of these mechanisms could be operative in UPD, affecting a large number of genes and contributing to the heterogeneity of the resulting tumor phenotype and clinical behavior (Fig. 6).

UPD is a previously cryptic type of chromosomal aberration that occurs ubiquitously in cancer and often maps to regions that are affected by loss of chromosomal material. During malignant evolution, the clonal selection process favors duplication of heterozygous somatic or germline mutations, disease-prone SNP alleles, or imprinting patterns that can produce a selective advantage. The development of whole-genome scanning technologies with SNP arrays has greatly facilitated the detection of UPD. In addition to the somatic form of UPD, SNP arrays can detect stretches of homozygosity due to inherited UPD or autozygosity, which requires truly clonal events to be distinguished from nonclonal homozygosity. Inherited UPD or autozygosity may constitute an independent predisposition factor for the development of malignancy. Although this theory is supported by the increased prevalence of cancer found in inbred populations, it needs to be further explored. Similarly, the mechanisms that lead to the acquisition of somatic UPD remain to be clarified. However, distinct pathways are probably involved in the development of numerical and segmental (interstitial or telomeric) UPD. In similarity to other chromosomal lesions, including gains, losses, and balanced translocations, various acquired or inherited defects of the mitotic machinery or DNA repair pathways may be involved in UPD. By elucidating the mechanisms behind this process, we may be able to identify people at risk for developing cancer as a result of UPD, and to develop new drug targets for the treatment of tumors associated with UPD. In the diagnostic setting, identification of UPD allows one to distinguish between truly homozygous genetic changes resulting from contamination with wild-type cells and those caused by cells with heterozygous mutations. This distinction can be used to more precisely assess the extent of a mutation and to better understand the effect of specific mutations in a particular genetic context. For example, mutations in TP53 most often occur in homo- or hemizygous form, and the presence of a wild-type allele is protective. The prognostic significance of some of the recurrent regions of UPD has been evaluated, and such defects could be included in a cytogenetically based prognostic scoring system. Should UPD prove to be of diagnostic value, cytogenetic methods could be introduced to allow for routine detection of this type of defect.

No potential conflicts of interest were disclosed.