Implications of polygenic risk for personalised colorectal cancer screening.

BACKGROUND: We modelled the utility of applying a personalised screening approach for colorectal cancer (CRC) when compared with standard age-based screening. In this personalised screening approach, eligibility is determined by absolute risk which is calculated from age and polygenic risk score (PRS), where the PRS is relative risk attributable to common genetic variation. In contrast, eligibility in age-based screening is determined only by age. DESIGN: We calculated absolute risks of CRC from UK population age structure, incidence and mortality rate data, and a PRS distribution which we derived for the 37 known CRC susceptibility variants. We compared the number of CRC cases potentially detectable by personalised and age-based screening. Using Genome-Wide Complex Trait Analysis to calculate the heritability attributable to common variation, we repeated the analysis assuming all common CRC risk variants were known. RESULTS: Based on the known CRC variants, individuals with a PRS in the top 1% have a 2.9-fold increased CRC risk over the population median. Compared with age-based screening (aged 60: 10-year absolute risk 1.96% in men, 1.19% in women, as per the UK NHS National Bowel Screening Programme), personalised screening of individuals aged 55-69 at the same risk would lead to 16% fewer men and 17% fewer women being eligible for screening with 10% and 8%, respectively, fewer screen-detected cases. If all susceptibility variants were known, individuals with a PRS in the top 1% would have an estimated 7.7-fold increased risk. Personalised screening would then result in 26% fewer men and women being eligible for screening with 7% and 5% fewer screen-detected cases. CONCLUSION: Personalised screening using PRS has the potential to optimise population screening for CRC and to define those likely to maximally benefit from chemoprevention. There are however significant technical and operational details to be addressed before any such programme is introduced.

Whole-genome sequencing of bladder cancers reveals somatic CDKN1A mutations and clinicopathological associations with mutation burden.

Bladder cancers are a leading cause of death from malignancy. Molecular markers might predict disease progression and behaviour more accurately than the available prognostic factors. Here we use whole-genome sequencing to identify somatic mutations and chromosomal changes in 14 bladder cancers of different grades and stages. As well as detecting the known bladder cancer driver mutations, we report the identification of recurrent protein-inactivating mutations in CDKN1A and FAT1. The former are not mutually exclusive with TP53 mutations or MDM2 amplification, showing that CDKN1A dysfunction is not simply an alternative mechanism for p53 pathway inactivation. We find strong positive associations between higher tumour stage/grade and greater clonal diversity, the number of somatic mutations and the burden of copy number changes. In principle, the identification of sub-clones with greater diversity and/or mutation burden within early-stage or low-grade tumours could identify lesions with a high risk of invasive progression.

Different APC genotypes in proximal and distal sporadic colorectal cancers suggest distinct WNT/ß-catenin signalling thresholds for tumourigenesis.

Biallelic protein-truncating mutations in the adenomatous polyposis coli (APC) gene are prevalent in sporadic colorectal cancer (CRC). Mutations may not be fully inactivating, instead producing WNT/ß-catenin signalling levels 'just-right' for tumourigenesis. However, the spectrum of optimal APC genotypes accounting for both hits, and the influence of clinicopathological features on genotype selection remain undefined. We analysed 630 sporadic CRCs for APC mutations and loss of heterozygosity (LOH) using sequencing and single-nucleotide polymorphism microarrays, respectively. Truncating APC mutations and/or LOH were detected in 75% of CRCs. Most truncating mutations occurred within a mutation cluster region (MCR; codons 1282-1581) leaving 1-3 intact 20 amino-acid repeats (20AARs) and abolishing all Ser-Ala-Met-Pro (SAMP) repeats. Cancers commonly had one MCR mutation plus either LOH or another mutation 5' to the MCR. LOH was associated with mutations leaving 1 intact 20AAR. MCR mutations leaving 1 vs 2-3 intact 20AARs were associated with 5' mutations disrupting or leaving intact the armadillo-repeat domain, respectively. Cancers with three hits had an over-representation of mutations upstream of codon 184, in the alternatively spliced region of exon 9, and 3' to the MCR. Microsatellite unstable cancers showed hyper-mutation at MCR mono- and di-nucleotide repeats, leaving 2-3 intact 20AARs. Proximal and distal cancers exhibited different preferred APC genotypes, leaving a total of 2 or 3 and 0 to 2 intact 20AARs, respectively. In conclusion, APC genotypes in sporadic CRCs demonstrate 'fine-tuned' interdependence of hits by type and location, consistent with selection for particular residual levels of WNT/ß-catenin signalling, with different 'optimal' thresholds for proximal and distal cancers.

A study of genomic instability in early preneoplastic colonic lesions.

It is difficult to explain the differential rates of progression of premalignant colonic lesions and differences in behaviour of morphologically similar lesions. Heterogeneity for microsatellite instability (MSI) and promoter methylation in driving these phenomena forward may explain this; however, no previous analysis has examined this in detail at the gland level, the smallest unit of colorectal premalignant lesions. We aimed to carry out an analysis of gland level genomic instability for MSI and promoter methylation. MSI occurred significantly more frequently (20%) in colonic glands than has previously been observed in whole colorectal polyps. Significant promoter methylation was seen in MLH1, PMS2, MLH3 and MSH3 as well as significant heterogeneity for both MSI and promoter methylation. Methylation and MSI may have a significant role in driving forward colorectal carcinogenesis, although in the case of MSI, this association is less clear as it occurs significantly more frequently than previously thought, and may simply be a passenger in the adenoma-carcinoma sequence. Promoter methylation in MLH1, MLH3, MSH3 and PMS2 was also found to be significantly associated with MSI and should be investigated further. A total of 273 colorectal glands (126 hyperplastic, 147 adenomatous) were isolated via laser capture microdissection (targeted at regions of MLH1 loss) from 93 colonic polyps and tested for MSI, and promoter methylation of the DNA mismatch repair genes MLH1, MSH2, MLH3, MSH6, PMS2, MGMT and MLH3 via methylation specific multiplex ligation-dependent probe amplification. Logistic regression modelling was then used to identify significant associations between promoter methylation and gland histological type and MSI status.

Common genetic variation contributes significantly to the risk of childhood B-cell precursor acute lymphoblastic leukemia.

Recent genome-wide association studies (GWAS) have provided the first unambiguous evidence that common genetic variation influences the risk of childhood B-cell precursor acute lymphoblastic leukemia (BCP-ALL), identifying risk single-nucleotide polymorphisms (SNPs) localizing to 7p12.2, 9p21.3, 10q21.2 and 14q11.2. The testing of SNPs individually for an association in GWA studies necessitates the imposition of a very stringent P-value to address the issue of multiple testing. While this reduces false positives, real associations may be missed and therefore any estimate of the total heritability will be negatively biased. Using GWAS data on 823 BCP-ALL cases by considering all typed SNPs simultaneously, we have calculated that 24% of the total variation in BCP-ALL risk is accounted for common genetic variation (95% confidence interval 6-42%). Our findings provide support for a polygenic basis for susceptibility to BCP-ALL and have wider implications for future searches for novel disease-causing risk variants.

TERC polymorphisms are associated both with susceptibility to colorectal cancer and with longer telomeres.

BACKGROUND AND AIMS: Shorter telomeres have been associated with increased risk of malignancy, including colorectal cancer (CRC). Telomere length is heritable and may be an intermediate phenotype linked to genetic susceptibility to CRC. METHODS: In a large sample, the study investigated whether candidate single nucleotide polymorphisms (SNP) in 'telomere biology' genes were associated with telomere length in leucocytes. SNP associated with an increased risk of CRC were searched for separately. RESULTS: Carriers of the common allele at SNP rs10936599, near the telomerase RNA component (TERC) locus, had significantly longer telomeres. It was independently found that the same rs10936599 allele was associated with increased risk of both CRC and colorectal adenomas. Neither telomere length nor CRC risk was associated with variation near telomerase reverse transcriptase or other telomere biology genes. In silico analysis showed that SNP rs2293607 was strongly correlated with rs10936599, mapped within TERC transcripts, had a predicted effect on messenger RNA folding and lay at a reported transcription factor binding site. TERC mRNA were expressed, differing only at the alleles of rs2293607, in CRC cell line HCT116. The long-telomere/CRC-risk allele was associated with higher levels of TERC mRNA and the formation of longer telomeres. CONCLUSIONS: Common genetic variation at TERC is associated with both longer telomeres and an increased risk of CRC, a potential mechanism being reduced levels of cell senescence or death. This finding is somewhat paradoxical, given retrospective studies reporting that CRC cases have shorter telomeres than controls. One possibility is that that association actually results from poorer survival in patients with longer telomeres.

A two-phase case-control study for colorectal cancer genetic susceptibility: candidate genes from chromosomal regions 9q22 and 3q22.

BACKGROUND: Colorectal cancer (CRC) is the second cause of cancer-related death in the Western world. Much of the CRC genetic risk remains unidentified and may be attributable to a large number of common, low-penetrance genetic variants. Genetic linkage studies in CRC families have reported additional association with regions 9q22-31, 3q21-24, 7q31, 11q, 14q and 22q. There are several plausible candidate genes for CRC susceptibility within the aforementioned linkage regions including PTCH1, XPA and TGFBR1 in 9q22-31, and EPHB1 and MRAS in 3q21-q24. METHODS: CRC cases and matched controls were from EPICOLON, a prospective, multicentre, nationwide Spanish initiative, composed of two independent phases. Phase 1 corresponded to 515 CRC cases and 515 controls, whereas phase 2 consisted of 901 CRC cases and 909 controls. Genotyping was performed for 172 single-nucleotide polymorphisms (SNPs) in 84 genes located within regions 9q22-31 and 3q21-q24. RESULTS: None of the 172 SNPs analysed in our study could be formally associated with CRC risk. However, rs1444601 (TOPBP1) and rs13088006 (CDV3) in region 3q22 showed interesting results and may have an effect on CRC risk. CONCLUSIONS: TOPBP1 and CDV3 genetic variants on region 3q22 may modulate CRC risk. Further validation and meta-analysis should be undertaken in larger CRC cohorts.

The role of cell proliferation and crypt fission in adenoma aggressiveness: a comparison of ileoanal pouch and rectal adenomas in familial adenomatous polyposis.

AIM: In patients with familial adenomatous polyposis (FAP), ileoanal pouch cancer is rare whereas rectal cancer is common, despite polyp initiation at the two sites being similar at the molecular level. This study investigated whether the disparity in adenoma aggressiveness reflects underlying differences in histogenesis. METHOD: Normal mucosal biopsies and 2-3 mm adenomas from patients with FAP were dissected into individual crypts. Crypt area, morphology, fission and mitoses were analysed for crypts from pouch, rectum and supra-anastomotic ileum. Immunohistochemistry of similar archival samples was performed for lysozyme, ß-catenin and TP53 expression. RESULTS: The morphology of normal crypts was similar at each site, although crypt area differed. The area of normal pouch crypts was intermediate between rectum and ileum. The area of adenomatous crypts of rectum and pouch was similar, but the latter had increased asymmetrical fission. Crypt mitoses were proportional to area in all tissues, but crypt fission was reduced in adenomatous crypts from the rectum compared with the pouch. Pouch adenomas retained lysozyme expression as seen in normal ileum. Nuclear ß-catenin accumulation was similar, but TP53 expression was increased in rectal adenomas. CONCLUSION: Diminutive polyps from rectum and pouch differ in morphology and proliferation. Aggressiveness in rectal polyps is not conferred by increased crypt proliferation, fission, or activation of the Wnt signalling pathway. Increased TP53 expression suggests other molecular mechanisms may be responsible. While crypt mitoses are proportional to crypt area, the threshold for fission may be site specific, indicating that tissue origin may influence histogenesis and thus malignant potential.

Replication study of SNP associations for colorectal cancer in Hong Kong Chinese.

BACKGROUND: Recent genome-wide association studies of colorectal cancer (CRC) have identified common single-nucleotide polymorphisms (SNPs) mapping to 10 independent loci that confer modest increased risk. These studies have been conducted in European populations and it is unclear whether these observations generalise to populations with different ethnicities and rates of CRC. METHODS: An association study was performed on 892 CRC cases and 890 controls recruited from the Hong Kong Chinese population, genotyping 32 SNPs, which were either associated with CRC in previous studies or are in close proximity to previously reported risk SNPs. RESULTS: Twelve of the SNPs showed evidence of an association. The strongest associations were provided by rs10795668 on 10p14, rs4779584 on 15q14 and rs12953717 on 18q21.2. There was significant linear association between CRC risk and the number of independent risk variants possessed by an individual (P=2.29 × 10(-5)). CONCLUSION: These results indicate that some previously reported SNP associations also impact on CRC risk in the Chinese population. Possible reasons for failure of replication for some loci include inadequate study power, differences in allele frequency, linkage disequilibrium structure or effect size between populations. Our results suggest that many associations for CRC are likely to generalise across populations.

A large-scale meta-analysis to refine colorectal cancer risk estimates associated with MUTYH variants.

BACKGROUND: defective DNA repair has a causal role in hereditary colorectal cancer (CRC). Defects in the base excision repair gene MUTYH are responsible for MUTYH-associated polyposis and CRC predisposition as an autosomal recessive trait. Numerous reports have suggested MUTYH mono-allelic variants to be low penetrance risk alleles. We report a large collaborative meta-analysis to assess and refine CRC risk estimates associated with bi-allelic and mono-allelic MUTYH variants and investigate age and sex influence on risk. METHODS: MUTYH genotype data were included from 20 565 cases and 15 524 controls. Three logistic regression models were tested: a crude model; adjusted for age and sex; adjusted for age, sex and study. RESULTS: all three models produced very similar results. MUTYH bi-allelic carriers demonstrated a 28-fold increase in risk (95% confidence interval (CI): 6.95-115). Significant bi-allelic effects were also observed for G396D and Y179C/G396D compound heterozygotes and a marginal mono-allelic effect for variant Y179C (odds ratio (OR)=1.34; 95% CI: 1.00-1.80). A pooled meta-analysis of all published and unpublished datasets submitted showed bi-allelic effects for MUTYH, G396D and Y179C (OR=10.8, 95% CI: 5.02-23.2; OR=6.47, 95% CI: 2.33-18.0; OR=3.35, 95% CI: 1.14-9.89) and marginal mono-allelic effect for variants MUTYH (OR=1.16, 95% CI: 1.00-1.34) and Y179C alone (OR=1.34, 95% CI: 1.01-1.77). CONCLUSIONS: overall, this large study refines estimates of disease risk associated with mono-allelic and bi-allelic MUTYH carriers.

Location in the large bowel influences the APC mutations observed in FAP adenomas.

The right colon differs from the left, in embryological origin, luminal environment, and function. In both sporadic colorectal cancer and Familial Adenomatous Polyposis (FAP), polyp density and cancer susceptibility vary markedly by colonic site. Adenomas in FAP have a different mutational spectrum in small intestine versus colon. This study aimed to investigate whether colonic location also influences the APC mutation spectrum in FAP. 127 1-2 mm mildly dysplastic adenomas from 5 patients with a codon 1309 germline mutation, and 41 from 3 patients with mutations proximal to codon 1265, were analysed to assess the frequency of loss of heterozygosity (LOH). We chose polyps from different locations in the colon. Immunohistochemistry for beta-catenin, caspase-3 and Ki-67 was performed to assess Wnt pathway activation, apoptosis and proliferation. In polyps from patients with a 1309 mutation, the frequency of LOH showed a gradient from rectum (highest) to caecum/ascending colon (lowest), but this was not present in patients with proximal germline APC mutations. Crypt-by-crypt analysis confirmed the LOH findings from whole polyps. Beta-catenin and caspase-3 expression showed no significant variation by colonic region, but Ki-67 expression decreased from ascending colon to rectum in tumours and normal tissue. Colonic site alters the mutational spectrum of APC, and crypt cell proliferation. The higher frequency of LOH in rectal polyps from patients with codon 1309 mutations may help to explain their increased polyp burden at this site compared with patients who have other germline APC mutations.

COGENT (COlorectal cancer GENeTics): an international consortium to study the role of polymorphic variation on the risk of colorectal cancer.

It is now recognised that a part of the inherited risk of colorectal cancer (CRC) can be explained by the co-inheritance of low-penetrance genetic variants. The accumulated experience to date in identifying these variants has served to highlight difficulties in conducting statistically and methodologically rigorous studies and follow-up analyses. The COGENT (COlorectal cancer GENeTics) consortium includes 20 research groups in Europe, Australia, the Americas, China and Japan. The overarching goal of COGENT is to identify and characterise low-penetrance susceptibility variants for CRC through association-based analyses. In this study, we review the rationale for identifying low-penetrance variants for CRC and our proposed strategy for establishing COGENT.

Evaluation of variants in the CHEK2, BRIP1 and PALB2 genes in an Irish breast cancer cohort.

It has been proposed that rare variants within the double strand break repair genes CHEK2, BRIP1 and PALB2 predispose to breast cancer. The aim of this study was to evaluate the prevalence of these variants in an Irish breast cancer cohort and determine their contribution to the development of breast cancer in the west of Ireland. We evaluated the presence of CHEK2_1100delC variant in 903 breast cancer cases and 1,016 controls. Six previously described variants within BRIP1 and five within PALB2 were screened in 192 patients with early-onset or familial breast cancer. Where a variant was evident, it was then examined in the remainder of our 711 unselected breast cancer cases. CHEK2_1100delC was found in 5/903 (0.5%) breast cancer cases compared to 1/1016 (0.1%) controls. One mutation at BRIP1 (2392 C>T) was identified in the early-onset/familial cohort. Examination of this variant in the remainder of our cohort (711 cases) failed to identify any additional cases. None of the previously described PALB2 variants were demonstrated in the early-onset/familial cohort. We show evidence of CHEK2_1100delC and BRIP1 2392 C>T within the Irish population. CHEK2_1100delC and BRIP1 mutations incidence in Ireland is similar to that found in other unselected breast cancer cohorts from northern European countries. We found no evidence to suggest that PALB2 mutation is an important breast cancer predisposition gene in this population.

APC and the three-hit hypothesis.

The seminal 'two-hit hypothesis' implicitly assumes that bi-allelic tumour suppressor gene (TSG) mutations cause loss of protein function. All subsequent events in that tumour therefore take place on an essentially null background for that TSG protein. We have shown that the two-hit model requires modification for the APC TSG, because mutant APC proteins probably retain some function and the two hits are co-selected to produce an optimal level of Wnt activation. We wondered whether the optimal Wnt level might change during tumour progression, leading to selection for more than two hits at the APC locus. Comprehensive screening of a panel of colorectal cancer (CRC) cell lines and primary CRCs showed that some had indeed acquired third hits at APC. These third hits were mostly copy number gains or deletions, but could be protein-truncating mutations. Third hits were significantly less common when the second hit at APC had arisen by copy-neutral loss of heterozygosity. Both polyploid and near-diploid CRCs had third hits, and the third hits did not simply arise as a result of acquiring a polyploid karyotype. The third hits affected mRNA and protein levels, with potential functional consequences for Wnt signalling and tumour growth. Although some third hits were probably secondary to genomic instability, others did appear specifically to target APC. Whilst it is generally believed that tumours develop and progress through stepwise accumulation of mutations in different functional pathways, it also seems that repeated targeting of the same pathway and/or gene is selected in some cancers.

Comprehensive analysis of common mitochondrial DNA variants and colorectal cancer risk.

Several lines of evidence implicate mitochondrial dysfunction in the development of cancer. To test the hypothesis that common mtDNA variation influences the risk of colorectal cancer (CRC), we genotyped 132 tagging mtDNA variants in a sample of 2854 CRC cases and 2822 controls. The variants examined capture approximately 80% of mtDNA common variation (excluding the hypervariable D-loop). We first tested for single marker associations; the strongest association detected was with A5657G (P=0.06). Overall the distribution of association P-values was consistent with a null distribution. Next, we classified individuals into the nine common European haplogroups and compared their distribution in cases and controls. This analysis also provided no evidence of an association between mitochondrial variation and CRC risk. In conclusion, our results provide little evidence that mitochondrial genetic background plays a role in modifying an individual's risk of developing CRC.

mRNA expression profiling of phyllodes tumours of the breast: identification of genes important in the development of borderline and malignant phyllodes tumours.

The aim of this study was to identify genes involved in the development of borderline and malignant phyllodes tumours of the breast (PTs). Expression profiling of 23 PTs (12 benign, 11 borderline/malignant) was performed using Affymetrix U133A GeneChips. mRNA expression in the borderline/malignant PTs was compared to the benign PTs. A group of 162 genes was over-expressed in the borderline/malignant group with a fold change > 2 and FDR < 0.1. Four of these genes were chosen for further investigation: PAX3, SIX1, TGFB2 and HMGA2. Over-expression was validated in a separate set of formalin-fixed, paraffin-embedded (FFPE) tumours, using either in situ hybridization or immunohistochemistry. This confirmed that expression of PAX3, SIX1, TGFB2 and HMGA2 in the stromal component of PTs was associated with the borderline/malignant phenotypes (p = 8.7 x 10(-5), p = 0.05, p = 0.009, p = 0.003, respectively; Fisher's exact test). The functional consequences of down-regulating these genes were studied using siRNA in short-term cultures and cell lines established from PTs. mRNA 'knock-down' of PAX3 resulted in significantly decreased cell proliferation in both a malignant and a borderline PT cell culture. mRNA 'knock-down' of SIX1 and HMGA2 resulted in decreased cell proliferation only in the malignant PT cell line, and 'knock-down' of TGFB2 resulted in decreased cell proliferation only in the borderline PT cell culture. This study shows that these four genes are involved in the development of borderline/malignant PTs. SIX1 over-expression was most marked in the highly malignant PTs, with particularly high expression in one case of metastatic PT. PAX3, TGFB2 and HMGA2 were expressed predominantly in borderline/malignant PTs, but showed some expression in benign tumours; they may be important in the transition from the benign to borderline/malignant phenotype.

Familial adenomatous polyposis and the small bowel: a loco-regional review and current management strategies.

Small-bowel tumours are an important cause of morbidity and death in patients with familial adenomatous polyposis. Intensive endoscopic surveillance is now standard in the long-term management of this condition. Thus, lesions occurring throughout the small bowel are increasingly noted by oesophagogastroduodenoscopy and flexible pouchoscopy. Some occur commonly de novo (in stomach, duodenum and ampulla), while others may occur following surgery (polyps of the ileostomy, ileoanal pouch, or small bowel above an anastomosis). These differ widely in incidence, natural history and management. This review provides a regional overview of these lesions, in terms of current research findings and management protocols.

Most low-level microsatellite instability in colorectal cancers can be explained without an elevated slippage rate.

Many cancers show a low level of microsatellite slippage and are labelled MSI-L (microsatellite instability--low). However, it is unclear whether this slippage can be attributed to some underlying genetic change that results in a mutator phenotype, analogous to mismatch repair deficiency in MSI-H cancers, or whether the apparent instability is the result of relatively frequent normal somatic slippage. Here, we have used a mathematical model of microsatellite slippage during cancer growth to estimate the degree of microsatellite slippage expected in a cancer due to normal somatic slippage. We compared the model to the slippage observed in 42 non-MSI-H cancers that were macro-dissected into four distinct regions and genotyped at N = 9 microsatellite loci. When the slippage rate was set at mu = 10(-5) per locus per division, ten cancers showed a level of slippage in at least one region that was too severe to be expected from normal somatic slippage alone, suggesting that these cancers had acquired MSI-L. Only one of these ten cancers had putative MSI-L in all four regions. When we considered a slightly higher slippage rate of mu = 5 x 10(-5), none of the cancers showed a degree of slippage that could not be reasonably explained by normal somatic slippage. Counting the number of 'unstable' loci was a poor indicator of putative MSI-L status. We conclude that most low-level microsatellite instability in colorectal cancers can be explained without requiring an elevated slippage rate during neoplastic development, and hence there is little evidence for a discrete MSI-L group of cancers. Putative MSI-L status is indicated by the presence of at least one locus that has multiple alleles that differ by at least five motif repeats from the germline. If an underlying genetic change does cause MSI-L, it appears to be a relatively uncommon event that occurs late in oncogenesis.

A comprehensive genetic profile of phyllodes tumours of the breast detects important mutations, intra-tumoral genetic heterogeneity and new genetic changes on recurrence.

The aims of this study were to identify genetic changes associated with malignant progression of the fibroepithelial neoplasms, phyllodes tumours of the breast (PTs), and to ascertain whether genetic progression occurs when PTs recur locally. A further aim was to assess whether the genetic data support the classification of these tumours into three subtypes, benign, borderline and malignant. 126 PTs (37 benign, 41 borderline, 48 malignant) were analysed by either array-CGH or the Illumina Goldengate assay. The large-scale genetic changes associated with malignant/borderline phenotypes were +1q, +5p, +7, +8, -6, -9p, -10p and -13. Cluster analysis of the array-CGH data supported the division of malignant and borderline PTs into two separate groups, one comprising almost all malignant lesions and the other, benign and borderline tumours. Interstitial deletions of 9p21 that involved the p16INK4a locus were present in many malignant/borderline PTs, and some of these appeared to cause homozygous loss. Loss of expression of p16INK4a was found frequently and this was associated with 9p deletion; we also identified one p16INK4a mutation and evidence of methylation of p16INK4a in malignant PTs. Our evidence shows that inactivation of this gene is important in the development of malignant PTs. In selected PTs, multiple areas of stroma were isolated and analysed separately by array-CGH. We found considerable intra-tumoral genetic heterogeneity. Analysis of paired primary and recurrent tumours showed that recurrent tumours often acquired new genetic changes; in particular, benign tumours tended to acquire changes characteristic of the malignant/borderline phenotype. We believe it likely that unfavourable sub-clones not easily identified by histology account for the unpredictable clinical behaviour of these tumours.