The consequences of uniparental disomy and copy number neutral loss-of-heterozygosity during human development and cancer

Authors

  • Pablo Lapunzina,

    1. Instituto de Genética Médica y Molecular (INGEMM), IDIPAZ-Hospital Universitario La Paz, Universidad Autónoma de Madrid, Spain
    Search for more papers by this author
  • David Monk

    Corresponding author
    1. Cancer Epigenetic and Biology Program (PEBC), Institut d'Investigació Biomedica de Bellvitge (IDIBELL), Hospital Duran i Reynals, Barcelona, Spain
      (email dmonk@idibell.cat).
    Search for more papers by this author

(email dmonk@idibell.cat).

Abstract

UPD (uniparental disomy) describes the inheritance of a pair of chromosomes from only one parent. Mechanisms that lead to UPD include trisomy rescue, gamete complementation, monosomy rescue and somatic recombination. Most of these mechanisms can involve aberrant chromosomes, particularly isochromosomes and Robertsonian translocations. In the last decade, the number of UPD cases reported in the literature has increased exponentially. This is partly due to the advances in genomic technologies that have allowed for high-resolution SNP (single nucleotide polymorphism) studies, which have complemented traditional methods relying on polymorphic microsatellite markers. In this review, we discuss aberrant cellular mechanisms leading to UPD and their impact on gene expression. Special emphasis is placed on the unmasking of mutant recessive alleles and the disruption of imprinted gene dosage, which give rise to specific and recurrent imprinting phenotypes. Finally, we discuss how copy number maps determined from SNP array datasets have helped identify not only deletions and duplications but also recurrent copy number neutral regions of loss-of-heterozygosity, which have been reported in many cancer types and that may constitute an important driving force in cancer. These tiny regions of UPD also alter imprinted gene dosage, which may have cumulative tumourgenic effects in addition to that of unmasking homozygous cancer-associated mutations.

Abbreviations used:
AS

Angelman syndrome

BWS

Beckwith Wiedemann syndrome

CPM

confined placental mosaicism

IUGR

intrauterine growth restriction

LOH

loss-of-heterozygosity

LOI

loss-of-imprinting

PHP-1A

pseudo-hypoparathyroidism 1A

PWS

Prader—Willi syndrome

SRS

Silver—Russell syndrome

UPD

uniparental disomy

pUPD

paternal UPD

mUPD

maternal UPD

SNP

single nucleotide polymorphism

TDNM

transient neonatal diabetes mellitus

Introduction

UPD (uniparental disomy) describes the inheritance of a pair of chromosomes/segments from only one parent. Lack of alleles from the mother or father is termed pUPD (paternal UPD) or mUPD (maternal UPD), respectively. UPD was first described in a seminal publication by Engel in 1980, who reported that constitutional UPD was likely to occur due to the high rates of meiotic error (Engel, 1980). If fertilization took place between a disomic egg and a sperm with nullisomy in the same chromosome, then UPD would result. In these cases, all cells of the body would carry the same UPD due to the germline establishment and transmission. It is now known that the extent of UPD can range from a small chromosomal segment to the entire chromosome, as proposed by Engel. Recent advances in molecular genetics have allowed the precise mapping and frequency of UPD to be assessed. Highly polymorphic markers have revealed that UPD can occur as either heterodisomy, the inheritance of both parental homologues or isodisomy, in which both identical chromosomes/segments are inherited from only one parental homologue, resulting in a reduction in homozygosity (Figure 1).

Figure 1.

Common mechanisms leading to UPD

(A) Maternal UPD revealed by means of tetranucleotide repeat analysis. The left panel shows heterodisomy, with the affected proband inheriting both maternal chromosomes; the middle panel shows isodisomy, with the proband inheriting two copies of one maternal allele with no contribution from the father; and the right panel showing normal biparental inheritance. (B) UPD revealed by means of SNP-array analysis. The log R ratio shows the normal DNA copy number, with a lack of heterozygosity in the allele frequency panel.

Techniques for identifying UPD

Routine cytogenetic analysis cannot identify UPD. Molecular diagnosis currently utilizes DNA polymorphisms, either highly informative microsatellite repeats or SNPs (single nucleotide polymorphisms). Microsatellite marker analysis is generally more informative than the use of SNPs, since each repeat may have several possible genotypes (depending on the expansion of the number of repeat units), thus increasing informativity. However, these markers are relatively rare, and only a few highly polymorphic repeats are routinely analysed per chromosome. Recently, new technologies, such as SNP arrays, have facilitated the detection of UPD. As shown in Figure 1, regions of homozygosity and heterozygosity can be precisely mapped throughout the genome. However, despite SNP arrays having the advantage of precise mapping of segmental UPD regions, this technology can only identify isodisomic cases. Heterodisomy is only identifiable by SNP arrays if parental DNA genotypes are available for comparison. Recently, the use of allelic DNA methylation profiling of imprinted DMRs (differentially methlyated regions) has been used to diagnose UPD. Techniques such as Southern blot, MS-PCR (methylation-specific-PCR) and MS-MLPA (methylation-specific multiplex ligation-dependent probe amplification) have all been successfully employed to identify and confirm UPD. The analysis of allelic methylation allows for the parental origin of the UPD to be determined without the need for parental DNA samples, but it cannot distinguish isodisomy from heterodisomy (Kubota et al., 1996; Kosaki et al., 2000).

Developmental mechanisms of UPD

Mechanisms leading to UPD include (A) gamete complementation, (B) trisomic rescue, (C) monosomy rescue, (D) compensatory UPD (also known as heterochromosomal substitution) and (E) post-fertilization errors including mitotic recombination (Figure 2). The first three result from the fertilization between one or more aneuploid gametes with the aneuploidy, originating from segregation errors in either meiosis I or II, and its subsequent return to diploidy. Generally, non-disjunction at meiosis I will lead to heterodisomy, whereas non-disjunction at meiosis II gives rise to isodisomy (Spence et al., 1988; Engel, 1993). In gamete complementation, fertilization occurs between abnormal aneuploid gametes, one being nullisomic and the other disomic. The disomic gamete contains a chromosome pair resulting from non-disjunction in meiosis I or II (Spence et al., 1988). Contrary to Engel's original predictions, only a small proportion of UPD cases seems to arise from gamete complementation. Instead, the majority arises from trisomy rescue, where a normal gamete and a diploid gamete combine. This trisomy is then reduced to diploidy by post-zygotic extrusion of one chromosome, which in one-third of the cases will result in UPD. Most cases of UPD via this mechanism are maternal in origin, since the incidence of aneuploidy is significantly higher in oocytes than in spermatozoa, due to the meiotic errors that can occur as a result of the prolonged period of risk between meiosis I (during embryonic development) and II (which is completed upon fertilization). Gamete complementation and trisomy rescue may lead to isodisomy or heterodisomy through the entire length of the chromosome, but combined iso-heterodisomy may also occur (Spence et al., 1998; Preece et al., 1999). The intervening iso-heterodisomy arises from meiotic recombination and the number of segments correlated with the number of recombinations.

Figure 2.

Mechanisms leading to UPD

(A) Gamete complementation, a non-disjunction event in maternal meiosis, leads to an oocyte with two copies of a chromosome that is fertilized by a nullosomic sperm lacking the same chromosome. (B) Trisomic rescue, which is the fertilization of a disomic oocyte by a normal sperm. In one-third of such cases the loss of one chromosome results in mUPD. (C) Monosomy duplication and (D) compensatory UPD with monosomic duplication, the replacement of a structurally abnormal chromosome, i.e. marker or ring chromosome, by means of somatic duplication. (E) UPD resulting from reciprocal exchange during mitosis, with two cells with maternal and paternal UPD from the recombination site of the telomere. A selection advantage results in somatic UPD of one parental origin. (F) Genome-wide pUPD, which is the result of a normal fertilization followed by failure of maternal DNA replication and paternal genome endoreplication. The overexpression of paternally expressed imprinted genes is suspected to result in the genome-wide pUPD cell line possessing a growth advantage over its normal counterpart.

In monosomy rescue, monosomic and nullisomic gametes unite, resulting in a monosomic zygote. Mitotic duplication of the only chromosome present leads to complete isodisomy and normal diploidy.

A very rare mechanism leading to compensatory UPD represents a mitotic attempt to correct aneuploidy, in which a structurally abnormal chromosome, such as a marker or ring chromosome, vanishes from the cell, being replaced by a second isodisomic chromosome through duplication (Miyoshi et al., 1999). Marker chromosomes associated with UPD have been described in ∼50 patients, mostly in the case of chromosomes 7, 14, 15 and 20, but they have also been described sporadically with regard to other chromosomes (Liehr, 2010). Most cases are maternal and only one-quarter of these anomalies was non-mosaic (Liehr, 2010).

Post-fertilization errors at early embryonic stages can also give rise to UPD for the complete or part of a chromosome. As described earlier, mitotic non-disjunction and the subsequent duplication of the remaining chromosome lead to complete isodisomy. However, UPD that occurs only in a part of a chromosome may arise from somatic recombination, which leads to segmental UPD. In this case reciprocal exchange would lead to a region of isodisomy between the point of recombination and the telomere in the daughter cells. Gene conversion is a similar mechanism that could also lead to small isolated regions of isodisomy anywhere along the chromosome (Thiagalingam et al., 2000). These somatic UPD events in early mitotic divisions, followed by the loss of one of the cell lines, will produce a segmental UPD without mosaicism. However, if both lines persist, mosaicism will result, with the daughter cell containing either mUPD or pUPD, which would not be detectable when using conventional genetic methods (Henry et al., 1993; Robinson, 2000).

Cytogenetic abnormalities associated with UPD

Many forms of structural chromosome abnormalities have been described in association with UPD in addition to the marker and ring chromosomes described above. The most prevalent are Robertsonian translocations, whole arm rearrangements between acrocentric chromosomes (chromosomes 13, 14, 15, 21 and 22) and isochromsomes. Isochromosomes are cytogenetically indistinguishable from homologous Robertsonian translocations and usually result from the misdivision of the centromere, so that both long arms move to one pole, and both short arms travel to the opposite pole. The association of acrocentric rearrangements and UPD is likely to be due to an increase in aneuploidy and the subsequent rescue, but the meiotic transmission of a Robertsonian translocation can associate with UPD if the corresponding homologue from the other parent is lost or not transmitted (Berend et al., 1999, 2000).

Developmental consequences of UPD

The severity of the resulting phenotype of each UPD varies greatly. Some UPDs are expected to be lethal early in zygotic development and will go undetected. UPDs without discernable phenotypes will also go undetected. Disease phenotypes associated with UPD in humans may be due to three distinct factors (Table 1).

Table 1.  Developmental consequence of UPD for each formation mechanism
UPD mechanismConsequence
Trisomy rescue– Heterodisomy
 – Mosaicism with trisomic cells
 – Unmasking mutant recessive mutations in isodisomic regions
 – Altered imprinted gene dosage
Gamete complementation– Heterodisomy
 – Unmasking mutant recessive mutations in isodisomic regions
 – Altered imprinted gene dosage
Post-fertilization error– Isodisomy
 – Mosaicism with normal cells
 – Unmasking mutant recessive mutations in isodisomic regions
 – Altered imprinted gene dosage
Monosomy rescue– Isodisomy
 – Unmasking mutant recessive mutations in isodisomic regions
 – Altered imprinted gene dosage
Somatic recombination– Segmental isodisomy between the point of mitotic recombination and the telomere
 – Mosaicism with normal cells
 – Unmasking mutant recessive mutations in isodisomic regions
 – Altered imprinted gene dosage
Gene conversion– Small regions of isodisomy within chromosome
 – Mosaicism with normal cells
 – Unmasking mutant recessive mutations in isodisomic regions
 – Altered imprinted gene dosage

Mosacisim

The viability of UPDs can also depend upon the degree of mosaicism, which can affect the entire fetus or be tissue-specific. CPM (confined placental mosaicism) is seen in 1–2% of CVS (chronic villus samples) used in prenatal diagnosis, and it describes the situation in which trisomic cells can be found exclusively in the placenta, with undetectable trisomy in the fetus (Kalousek and Vekemans, 1996). The most reported example of UPD with CPM is associated with trisomy 16, which is commonly associated with early pregnancy loss and is the most common form of CPM seen in IUGR (intrauterine growth restriction). In cases of mUPD16, the associated IUGR is related to the level of trisomy mosaicism in the placenta as opposed to UPD, since fetuses with biparental rescue are also growth-restricted (Kalousek, 1994; Moore et al., 1997).

Imbalance of imprinted gene expression

Heterodisomy is not expected to cause abnormalities unless genes on affected chromosomes are subject to genomic imprinting. Genomic imprinting is the monoallelic expression that is dependent on the allele's parental origin. This monoallelic gene expression is dictated by allelic DNA methylation originating in either the oocyte or sperm. When UPD occurs in a region containing an imprinted domain, cells will inherit either two active or two repressed copies, leading to an abnormal dosage of these imprinted gene products. This is different from the commonly termed loss-of-imprinting that is associated with aberrant imprinted gene expression due to epimutations, since UPD chromosomes possess the correct epigenetic profile of either parent-of-origin.

Abnormal dosage of imprinted genes has been associated with several imprinting disorders including the PWS (Prader—Willi syndrome; mUPD15), AS (Angelman syndrome; pUPD15), BWS (Beckwith Wiedemann syndrome; pUPD11p15.5), TDNM (transient neonatal diabetes mellitus; pUPD6), SRS (Silver Russell Syndrome) [mUPD7 (maternal UPD7-like syndrome)], mUPD14 (maternal UPD14-like syndrome), pUPD14 (paternal UPD14) and Type I PHP-1A (pseudo-hypoparathyroidism 1A; pUPD20) (Table 2). In some cases of imprinting syndromes caused by UPD, the phenotype is often less severe than that of individuals with epimutations or deletions. This is especially noticeable in AS and SRS (Bottani et al., 1994; Moncla et al., 1999; Hannula et al., 2001); however, phenotype/genotype correlations between SRS patients may indicate differences between individuals with mUPD7 and 11p11.5 H19 methylation defects. In cases of PHP-1A, the zygotic recombination associated with partial chromosome 20q isodisomy coincides with a ‘breakpoint region’ associated with PHP-1A patients with 20q12 deletions (Aldred et al., 2002). This suggests that in 20q12 there are regions of recombination that could lead to either deletions or partial isodisomies (Fernández-Rebollo et al., 2010). However, a similar analysis of nine BWS patients with UPD11p isodisomy failed to identify a common mitotic recombination region (Romanelli et al., 2011a). Further studies are required to confirm if this is also true with regard to other imprinting syndromes.

Table 2.  The frequency of UPD in human imprinting syndromes
ChromosomeParental originPhenotypeGenesFrequency of UPD (%)Reference
6PaternalTransient neonatal diabetes mellitusPLAGLI/HYMAI∼40Gardner et al. (1998)
7MaternalSilver—Russell syndromeGRB10/MEST?6-10Kotzot et al. (1995)
11PaternalBeckwith—Wiedemann syndromeIGF2/miR483/miR675/CDKN1C∼20Dutly et al. (1998)
14PaternalPolyhydramnios/pre—postnatal growth restriction/bell-shaped thorax/muscular hypotoniaDLK1/RTL1∼30 reported casesPapenhausen et al. (1995)
 MaternalPre—postnatal growth restriction/facial dysmorphisms/early puberty/developmental delayDLK1/RTL1∼50 reported casesMiyoshi et al. (1998)
15PaternalAngelman syndromeUBE3A∼2Malcolm et al. (1991)
 MaternalPrader—Wili syndromeSNRPN/snoR116∼25Nicholls et al. (1989)
16MaternalVariable outcomes involving growth and developmental delayUnknownUnknownVaughan et al. (1994)
20PaternalPseudo-hyoparathyroidism type IbGNASUnknownFernández-Rebollo et al. (2010)
 MaternalPre- and post-natal growth restriction/microcephaly/psychomotor developmental delayGNAS clusterUnknownSalafsky et al. (2001)

For neurological disorders such as AS and PWS, opposite parental UPDs are associated with different imprinting phenotypes; however, this is not always the case. One parental UPD may be related to a specific phenotype, whereas the opposite UPD is undetectable. For example, the lack of paternally expressed genes within PEG10, MEST and GRB10 domains on chromosome 7 is associated with mUPD7 and an SRS phenotype (Kotzot et al., 1995; Preece et al., 1997), whereas there is no obvious imprinting phenotype associated with pUPD7 since all known cases were identified due to the unmasking of recessive mutations (Table 3). Some chromosomes with reciprocal UPDs are not associated with consistent phenotypes, only being identified by the unmasking of recessive mutations, this suggests that these chromosomes do not harbour yet to be identified imprinted genes. The exceptions are chromosomes 1 and 13, where the altered dosage of the imprinted gene, DIRAS3 and RB1, is not associated with a recognizable developmental phenotype (Slater et al., 1995).

Table 3.  Phenotypes associated with isodisomy and unmasking recessive mutationsBold type indicates opposite parental UPDs associated with unmasking mutations for the same recessive disease.
ChromosomeParental originPhenotypeMechanism/mutationReference
1PaternalPycnodysostosisPaternal isodisomy (Cathepsin K mutation?)Gelb et al. (1998)
  RhD blood groupPaternal isodisomyMiyoshi et al. (2001)
  Congenital insensitivity to painUnmasking NTRK1 mutationsMiura et al. (2000)
  Retinitis pigmentosaUnmasking USH2A mutationsRivolta et al. (2002)
  SchizophreniaPaternal isodisomyAbecasis et al. (2004)
  Herlitz Junctional Epidermolysis BullosaPaternal isodisomy (LAMB3 mutation?)Takizawa et al. (2000)
  Chediak—Higashi syndromeUnmasking CHS1 mutationManoli et al. (2010)
 MaternalChediak—Higashi syndromeMaternal isodisomy (CHS1 mutation?)Dufourcq-Lagelouse et al. (1999)
  Herlitz Junctional Epidermolysis BullosaUnmasking LAMB3 mutationPulkkinen et al. (1997)
  Zellweder syndromeUnmasking PEX10 mutationTurner et al. (2007)
2PaternalRetinal dystrophyUnmasking MERTK mutationThompson et al. (2002)
  Crigler—Najjar type 1 syndromeUnmasking UGT1A1 mutationPetit et al. (2005)
  Donnai—Barrow syndromeUnmasking LRP2 mutationKantarci et al. (2008)
 MaternalLethal respiratory failureUmasking SFTPB mutationHamvas et al. (2009)
  Infantile-onset ascending spastic paralysisUnmasking ALS2 mutationHerzfeld et al. (2009)
  Primary hyperoxaluria type 1Unmasking AGXT mutationChevalier-Prost et al. (2005)
  Lethal trifunctional protein deficiencyUnmasking alpha-subunit mutationSpiekerkoetter et al. (2002)
3PaternalNo phenotype Xiao et al. (2006)
 MaternalDystrophic Epidermolysis BullosaUnmasking COL7A1 mutationFassihi et al. (2006)
  Fanconi—Bickel syndromeUnmasking ABCC8 mutationHoffman et al. (2007)
  Glycosylation type 1d disorderUnmasking ALG3 mutationSchollen et al. (2005)
4MaternalAbetalipoproteinemiaUnmasking MTP mutationYang et al. (1999)
  Ellis—van Creveld syndromeUnmasking EVC mutationTompson et al. (2001)
  DepressionMaternal isodisomyMiddleton et al. (2006)
  Congenital afibrinogenaemiaUnmasking FGA deletionSpena et al. (2004)
5PaternalSpinal muscular atrophyPaternal isodisomy (SMN1 mutation?)Brzustowics et al. (1994)
6PaternalIFN-gamma receptor 1 deficicencyUnmasking IFNGR1Prando et al. (2010)
 MaternalIsolated cleft lip and palateMaternal heterodisomySalahshourifar et al. (2010)
  Congenital adrenal hyperplasia+IUGRUnmasking mutation (CYP21A2 mutation?)Spiro et al. (1999)
  Congenital adrenal hyperplasiaUnmasking CYP21A2 mutationParker et al. 2006
7PaternalCongenital chloride diarrhoea Höglund et al. (1994)
  Complete situs inversus and immotile cilla(DNAH11 mutation?)Pan et al. (1998)
  Cystic fibrosisUnmasking CFTR mutationLe Caignec et al. (2007), Fares et al. (2006)
8PaternalLipoprotein lipase deficiencyUnmasking LPL mutationBenlian et al. (1996)
 MaternalNo phenotype Karanjawala et al. (2000)
9MaternalSyndromic congenital hypothyroidismUnmasking FOXE1 mutationCastanet et al. (2010)
  Cartilage-hair hypoplasia Sulisalo et al. (1997)
  Leigh syndromeUnmasking SURF1 mutationTiranti et al. (1999)
10PaternalMultiple endocrine neoplasia type 2A(RET mutation?)Kousseff et al. (1992)
 MaternalFamilial hemophagocytic lymphohistiocytosisUnmasking PRF1 mutationAl-Jasmi et al. (2008)
12MaternalNo phenotype Von Eggeling et al. (2002)
13PaternalNo phenotype Slater et al. (1995)
  Hearing lossUnmasking GJB2 mutationYan et al. (2007)
 MaternalHearing lossUnmasking GJB2 mutationAlvarez et al. (2003)
  No phenotype Slater et al. (1995)
16PaternalLethal respiratory failureUmasking ABCA3 mutationHamvas et al., 2009
  No phenotype Kohlhase et al. (2000)
  Hydrops fetalis alpha-thalassemia Ngo et al. (1993)
 MaternalHydrops fetalis alpha-thalassemia Wattanasirichaigoon et al. (2008)
  Malonic aciduriaUnmasking MLYCD mutationMalvagia et al. (2007)
17PaternalJunctional epidermolysis with pyloric atresia Natsuga et al. (2010)
 MaternalNephropathic cystinosisUnmasking CTNS deletionLebre et al. (2009)
21PaternalNo phenotype Blouin et al. (1993)
 MaternalNo phenotype Rogan et al. (1999)
22PaternalNo phenotype (infertility?) Ouldim et al. (2008)
 MaternalNo phenotype (infertility?) Schinzel et al. (1994)
XMaternalDuchenne muscular dystrophyUnmasking DMD deletionQuan et al. (1997)

Reduction to homozygosity

In regions of isodisomy, the inheritance of two mutant copies of a recessive gene mutation from a carrier parent can lead to abnormal phenotypes. The first reported case of isodisomy unmasking a mutant recessive allele was described by Spence and co-workers in 1988 (Spence et al., 1988). They described a small child with CF (cystic fibrosis) who had received two copies of the same chromosome 7 with a CFTR (cystic fibrosis transmembrane conductance regulator) mutation from her carrier mother, and no contribution from her non-carrier father. Notably, this patient had short stature and body asymmetry suggestive of imprinting disorder SRS. Since this initial report, more than 50 cases of recessive disorders have been found to be associated with UPD (Kotzot, 2004; Engel, 2006) (Table 3). In extremely rare cases, UPD for opposite parental chromosomes have unmasked different recessive mutations for the same gene in individuals with the same syndrome. For example both paternal and maternal isodisomy for chromosome 1 have been identified due to mutations of LAMB3 in Herlitz Junctional Epidermolysis Bullosa patients (Pulkkinen et al., 1997; Takizawa et al., 2000) and mutations of CHS1 in the case of the Chediak—Higashi syndrome (Dufourcq-Lagelouse et al., 1999; Manoli et al., 2010). These rare recessive disorders associated with reciprocal UPDs suggest that these chromosomes are not associated with a severe phenotype due to altered imprinted gene dosage.

Using the frequency of imprinting syndromes to extrapolate the rate of UPD formation

Very little is known about the frequency of UPD formation, but one of the best estimates can be drawn from the incidence of UPD in imprinting disorders. For example, the prevalence of PWS and AS is approx. 1/20000 and 1/18000 live births respectively, with UPD accounting for 25% of PWS and 2% of AS cases. Extrapolation of these figures reveals that mUPD15 occurs once every 1/80000 live births, whereas pUPD15 occurs every 1/900000 births. The same calculations can be performed for BWS/pUPD11p15.5, TDNM/pUPD6 and SRS/mUPD7 that give frequencies of 1/650000, 1/1250000 and 1/980000 live births. However, it must be remembered that these figures will probably indicate an ascertainment bias since these are the most frequently observed viable UPDs.

Genome-wide UPD mosaics associated with imprinting phenotypes

Recently a number of mosaic genome-wide pUPDs were described with BWS-like phenotypes (Morales et al., 2009; Yamazawa et al., 2011; Romanelli et al., 2011b). This is remarkable, since non-mosaic genome-wide pUPD is incompatible with fetal development and results in a hydatidiform mole (Devriendt 2005). Two of these rare BWS-like cases were also reported to have developed multiple tumours (Morales et al., 2009; Romanelli et al., 2011b). The most likely mechanism of mosaic genome-wide UPD is normal fertilization, followed by a failure of replication and condensation of the maternal pronuclei, followed by paternal genome endoreduplication, resulting in androgenetic/biparental mosaicism, with over-representation of genome-wide pUPD cell lines (Figure 2F). Two of the BWS-like genome-wide pUPD cases do not show strong features of TNDM or AS (Romanelli et al., 2011b; Yamazawa et al., 2010), suggesting either that the complete overexpression of PLAGL1 and UBE3A is required for phenotypic features or that the mosaic cellular contribution differs in various tissues with the ‘normal cell line’ highly represented in tissues associated with those phenotypes. This type of tissue-specific mosaicism is well documented in mouse models, where temporal and spatial selection occurs during parthenogenetic-normal fetal chimaera development (Nagy et al., 1989; Fundele et al., 1990).

Interestingly, a single reported case of mosaic genome-wide mUPD has been described, being associated with an SRS-like phenotype (Yamazawa et al., 2010). The prominent imprinting phenotypes that are associated with these opposing genome-wide UPDs are also caused by epimutations of H19-DMD (Weksberg et al., 1993; Gicquel et al., 2005), suggesting that altered expression of IGF2, miR-483, H19 and miR-675 is strongly associated with the resulting phenotypes.

Cancer associated copy number neutral LOH (loss-of-heterozygosity)

The observation that the BWS-like genome-wide pUPD cases developed multiple cancers (kidney hyperplasmia/Wilms tumour, adrenal adenoma, hepatoblastoma and pancreatic tumour) suggests that the overexpression of paternally expressed genes/lack of maternally expressed genes contributes to cancer initiation and growth. However, most reports in the literature of UPD in cancer to date have focused upon LOH and the unmasking of cancer driver mutations as a principal mechanism of carcinogenesis, without taking into account the effects of imprinted genes.

With recent advances in molecular techniques, it is now possible to analyse the complete genome at a 500000 SNP resolution. This facilitates the simultaneous evaluations of the copy number generated by deletions and duplications and copy number-neutral changes associated with somatic UPD. Regions of homozygosity and heterozygosity can be precisely mapped along the chromosomes based on SNP genotypes. Incidences of isodisomy can therefore be detected reliably, but not those of heterodisomy as this would require parental DNA samples to be analysed simultaneously.

Mechanisms of copy number neutral LOH in cancer

UPD and the unmasking of recessive mutations were first reported with regard to developmental disorders, but similar mechanisms have been shown to be of importance in the inactivation of tumour-suppressor genes or the activation of oncogenes. In addition to the genome-wide UPD and altered imprinted gene expression described above, other mechanisms leading to UPD can occur in cancer. If the entire chromosome is subject to isodisomy in somatic cells, then the UPD probably will arise from a chromosomal segregation error in mitosis. This doubtless occurs at high rates due to the dysregulated cell cycle and replication associated tumour progression, in which one allele is lost in anaphase lag and the remaining allele is reduplicated. Where the UPD occurs in only a segment of the chromosome, it does so probably through either gene conversion or mitotic recombination. Mitotic recombination between identical low copy repeats in the G2 phase of the cell cycle (Reliene et al., 2007) is the most likely cause, since gene conversion, where small 300 bp-1 kb regions of DNA sequence are transferred from one allele to another allele, is unlikely, as most detectable segmental UPDs extend beyond 1 kb.

Common copy number neutral LOH regions in cancer

Recently, studies comparing constitutional and tumour DNA genotypes revealed that UPD occurs in a variety of tumour types including neuroblastomas (Carén et al., 2008), clear cell renal carcinomas (Toma et al., 2008), pancreatic adenocarcinomas (Harada et al., 2008) and breast cancer (Murthy et al., 2002; Tuna et al., 2010) with most studies being performed on haematological malignancies (Fitzgibbon et al., 2005; Flotho et al., 2007; Mullighan et al., 2007; Radtke et al., 2009; Walter et al., 2009; Bullinger et al., 2010) (Figure 3). These studies have revealed that UPD can occur in almost any chromosome, but it is becoming evident that UPDs are non-randomly distributed with cooperation occurring between the acquired UPD and gene mutations. For example, studies on cytogenetically normal acute myeloid leukaemias found that all segmental UPD 13q harbour FLT3 mutations, that all UPD 11p cases have WT1 mutations and that UPD 17q occurs with NF-1 mutations (Flotho et al., 2007; Grand et al., 2009; Bullinger et al., 2010). Large-scale studies of acute childhood and chronic lymphoblastic leukaemias have also shown that UPD can lead to homozygous deletions. The UPD of chromosome 9p has revealed focal homozygous deletions containing CDKN2A, whereas UPD 13 can result in homozygous deletions of miR-15a and miR-16-1 (Mullighan et al., 2007; Lehmann et al., 2008). The presence of acquired UPD in haematological malignancies has also been shown to correlate with clinical outcomes, with UPD 11p being favourable when compared with UPD 13q (Bullinger et al., 2010). Correlations between somatically acquired UPDs and response to treatment have been reported with regard to solid tumours. Breast cancer patients with triple negative tumours [ER (oestrogen receptor)-negative, PR-negative and HER2/neu-negative] are associated with recurrent UPDs of 17q and 13q, whereas HER2/neu-positive tumours are associated with recurrent UPDs of 1q and 16q UPDs (Tuna et al., 2010). These subgroups of patients have different responses to treatments/outcomes, with HER2/neu-positive (1q and 16q UPD) patients responding to trantuzumab treatment, while triple negative cancers (17q and 13q UPD) show poorer prognoses, as there is currently no available targeted therapy.

Figure 3.

Distribution of acquired UPD/copy number neutral LOH in colorectal (red), breast (blue), hepatocarcinoma (green) and haematological cancers (black)

Each colour bar represents a region of recurrent UPD. Known homozygous mutated tumour suppressor or oncogenes are labelled in black and imprinted genes mapping to common regions of UPD are labelled in red. (Data taken from Lips et al., 2007; Tuna et al., 2010; Midorikawa et al., 2006; Walter et al., 2009; Fitzgibbon et al., 2005; Bullinger et al., 2010.)

Cancer, imprinting and copy number neutral LOH

LOI (loss-of-imprinting) is a common event in cancer, but it is currently unknown whether abnormal imprinted gene dosage is a driver of tumorigenesis or whether it occurs as a consequence of the epigenetic disruption that occurs in cancer progression. The recently acquired ability to identify small regions of somatic UPD has helped identify many regions of copy number neutral LOH that harbour mutated tumour suppressor genes or oncogenes. The mapping of these regions has revealed that many of the same contain imprinted gene clusters that will be dysregulated. For example, UPD 1p containing c-MPL or NRAS mutations will also possess an altered expression of DIRAS3, and UPD 11p with homozygous WT1 mutations will have an abnormal dosage of both H19-IGF2 and KCNQ1OT1-CDKN1C domains. Lastly, UPD 13 that has been reported in many tumour types, and is associated with homozygous mutations of RB1, will also be subject to parent-of-origin effect, as this critical cancer gene has recently been shown to be imprinted (Kanber et al., 2009; Beà et al., 2009).

The majority of cancer-associated UPDs is isodisomic and is identified without parental DNA samples, meaning that the parental origin of the UPD cannot be assigned. Without analysis of allelic DNA methylation within each imprinted locus, one cannot determine whether or not reactivation of the normally silent allele of a growth-promoting gene and/or silencing of the normally active allele of a tumour-suppressor has a cumulative effect along with homozygous mutations on cancer initiation and progression.

Summary and future directions

The number of UPD cases reported in the literature is continually increasing. This is mainly due to advances in genomic technologies that have allowed for high-resolution SNP analysis and more efficient mutation detection. Constitutional UPD is associated with developmental diseases as a consequence of meiotic errors, which can result in either hetero- or iso-disomy depending on whether the error occurred in meiosis I or II. Acquired UPD probably occurs as a result of mitotic error in somatic cells, which can be an important initiation step in tumorigenesis. It will be vital to determine whether somatic UPD is a contributing factor in additional late onset diseases other than cancer. Copy number maps from SNP array datasets have identified not only deletions and duplication but also recurrent chromosomal regions subject to UPD in many cancer types. Often these regions of UPD are associated with mutations in tumour-suppressor genes or oncogenes; they also alter imprinted gene expression. Future work that assesses these vast datasets will allow for the identification of not only new tumour suppressor genes but also of additional imprinted loci. Importantly, additional research is need to determine whether imprinted gene expression exacerbates the tumourigenic affects of cancer-associated mutations within the region of UPD and whether UPD mapping can reveal information critical for prediction of the severity and treatment response of various cancer types.

Acknowledgement

We thank members of the Monk laboratory for their helpful comments on the manuscript.

Funding

D.M. is a Ramon y Cajal Research Fellow and is funded by Spanish Ministerio de Educación y Ciencia (SAF2008-1578) and the Asociación Española Contra el Cancer (AECC). D.M. and P.L. are co-funded by the Fundació La Marató de TV3.

Footnotes

  1. Uniparental disomy: The inheritance of a complete chromosome/chromosome segment from only one parent.

  2. Copy number neutral LOH: Small chromosomal regions inherited from one parent, normal as a result of somatic recombination. The majority of these small regions is identified by means of genome SNP-arrays.

  3. Heterodisomy: The inheritance of both parental homologues by a chromosome from only one parent.

  4. Isodisomy: The inheritance of identical chromosomes/segments from only one parental homologue.

  5. Imprinting: Genomic imprinting is an epigenetic process, involving germline derived DNA methylation that ensures parent-of-origin monoallelic gene expression in somatic tissues. There are approx. 50 imprinted genes in humans, whose disruption is associated with specific syndromes and cancer.

  6. Mosaicism: Denotes the presence of two populations of cells with different genotypes in an individual.

Ancillary