The CBP gene (OMIM #600140) encodes the 2441 amino acid cyclic AMP response element binding protein (CREB) binding protein (Mr 265 000) that functions as an essential co-activator for a number of transcription factors.¹ In particular, CBP, like its homologue EP300, enhances gene transcription by linking sequence specific transcription factors to transcription factor IIB and the RNA polymerase II holoenzyme.² CBP can also promote gene transcription by acetylation of histones through its histone acetyltransferase activity (HAT)³ and by acetylation of specific transcription factors such as p53.^4–6 Several lines of evidence suggest the involvement of CBP in tumour formation. CBP interacts with the human T cell leukaemia virus (HTLV) Tax protein and v-myb to induce viral and cellular genes that promote cell transformation, while adenovirus E1A, simian virus SV40 Tag, and human papillomavirus (HPV) E6 bind to CBP and inhibit its co-activator function. CBP is also known to modulate the actions of the hMDM2 and AML1 cellular proto-oncogenes and the BRCA1 and p53 tumour suppressors.⁷ More direct evidence comes from genetic studies of translocation events involving CBP in leukaemia. In particular, the t(8;16)(p11;p13) translocation in AML involves a fusion of the CBP and MOZ genes,⁸ a novel t(10;16)(q22;p13) translocation in a childhood acute myelogenous leukaemia (AML-M5a) generates a MORF-CBP chimera,⁹ and the t(11;16)(q23;p13.3) translocation in topoisomerase II treated acute leukaemia or myelodysplasia results in a MLL-CBP fusion.¹⁰ In addition, haploinsufficiency of CBP in humans due to chromosomal rearrangements, microdeletions, and point mutations results in Rubenstein-Taybi Syndrome (RTS), and is associated with an increased risk for a variety of tumours.¹¹

As a result of these observations, a number of studies have evaluated whether CBP is a direct mutational target in cancer and has tumour suppressor activity. One recent mutation screening study of CBP in breast, ovarian, and colorectal tumours and cell lines identified two CBP truncating mutations in ovarian cancer cell lines, but failed to detect mutations in the tumours.¹² Haematological malignances displaying loss of heterozygosity (LOH) at the CBP locus have been detected in aged CBP heterozygous mice,¹³ and in chimeric mice derived from CBP^–/– embryonic stem cells.¹⁴ In addition, recent studies using an MMTV driven conditional knockout mouse model demonstrated that truncation of the CBP protein in the thymus results in development of T cell lymphomas,¹⁵ suggesting that targeted mutations in CBP can predispose to cancer. To determine whether CBP is a mutational target in human epithelial cancers, and specifically to evaluate the presence of CBP mutations in breast and ovarian tumours, we conducted a mutational analysis of the CBP gene.

MATERIALS AND METHODS

Sample selection

A total of 75 unselected surgically resected fresh frozen ovarian tumour specimens and 61 unselected fresh frozen breast tumour specimens were obtained from the ovarian and breast tumour banks at the Mayo Clinic. Use of these samples was approved by the Mayo Clinic institutional review board. Each tumour specimen was defined by haematoxylin and eosin analysis as containing at least 30% tumour cells. Genomic DNA was prepared from sections of each tumour using the PureGene DNA extraction kit (Gentra). Genomic DNA was also prepared from 1×10⁶ cells from each of 8 ovarian cancer cell lines (SKOV3, OVCAR5, OVCAR8, OV17.17, Ov167, Ov177, Ov202, Ov207) and eight breast cancer cell lines (MCF7, BT474, ZR75-1, UACC893, UACC812, MB-MDA468, MB-MDA157, MB-MDA361).

Mutational analysis

Selection of primer pairs for PCR amplification of the CBP coding sequence and splice sites from genomic DNA was based on previous studies¹⁶ and on the genomic structure of CBP.¹⁷ The sequences of the 36 primer pairs used for PCR and the associated PCR conditions are available upon request. Briefly, 20 ng of tumour genomic DNA was used as template for PCR. PCR products were heteroduplexed, loaded onto a Transgenomic WAVE dHPLC system, and analysed using product specific melting temperatures. When altered peaks were identified on the dHPLC, the original tumour DNA was re-amplified by PCR and sequenced in both directions.

Germline DNA analysis

Genomic DNA was extracted from 5 μm sections of paraffin blocks of normal ovaries and/or endometrial tissue. Briefly, the sections and cores were deparaffinised with xylene and treated with proteinase K at 55°C overnight. Genomic DNA was isopropanol precipitated. These DNAs were used as templates for PCR of the exons containing the CBP mutations.

Biallelic inactivation of CBP

The complete inactivation of the CBP locus was established by immunohistochemistry of sections from paraffin blocks of tumour specimens using anti-CBP antibodies A22 (sc-369) and C20 (sc-583). The A22 antibody detects an epitope in the N terminus of CBP and the C20 antibody detects an epitope in the C terminus. Briefly, 5 μm sections were deparaffinised and stained with A22 (1:1000) or C20 (1:800) and the EnVision+ HRP labelled polymer (Dako, Glostrup, Denmark) on a Dako Autostainer. LOH analysis at the CBP locus was also performed using PCR techniques. Briefly, genomic DNA was extracted from 0.6 mm cores of ovarian tumour paraffin blocks as described above. Cores were used to enrich for tumour cells over infiltrating normal cells. The tumour and normal DNA were used as templates for PCR of the exons containing the CBP mutations, and the products were evaluated by dHPLC. Tumour DNAs displaying little or no mutant heteroduplex formation were considered to have LOH of the CBP locus.

Immunohistochemistry of p53 and p27^KIP1

Sections from the paraffin block tumour specimens were also stained for the presence of stabilised p53 and p27^KIP1 using anti-p53 clone DO-7 (Dako) and anti-p27 clone SX53G8 (Dako) antibodies as described above.

RESULTS

Determination of the intron–exon structure of CBP

The detailed genomic structure of the CBP gene was analysed by comparing the genomic sequence of the two bacterial artificial chromosomes and two cosmid clones RP11-420F6, LA16c-RT191, LA16c-RT102, and RP11-316H7 (GenBank AC005564, AC004509, AC004651, AC004760) with the known CBP cDNA sequence. The sequence alignment revealed that CBP is comprised of 31 exons ranging in size from 45 to 3306 bp. This predicted gene structure matched exactly with the observations from another recent study.¹⁷ The N terminal transactivating domain of CBP is encoded by exons 2–4, the three cysteine histidine rich domains by exons 4–5, 19–28, and 30–31, the KIX domain by exons 6–10, the bromodomain by exons 17–18, the HAT domain by exons 18–30, and the C terminal transactivating domain by exon 31.

CBP mutation screening in primary tumours

To determine whether mutations in CBP were associated with human breast or ovarian cancer, we conducted a mutation screen of all CBP coding regions and splice acceptor and donor sites in genomic DNA from 75 unselected fresh frozen ovarian tumour specimens, 61 unselected fresh frozen breast tumour specimens, and 16 breast and ovarian cancer cell lines. Each genomic DNA sample was PCR amplified using 36 independent PCR primer sets. PCR products were heteroduplexed, or in the case of cell line products were mixed with PCR products from normal lymphoblastoid cell DNA and subsequently heteroduplexed, and subjected to dHPLC analysis (Transgenomics Inc.). Samples with altered peak structure on the dHPLC were re-amplified from the original genomic DNA and sequenced to identify the specific sequence alterations.

Two unique frameshift mutations, six missense mutations, twelve intronic sequence alterations, four alterations in the 3′ untranslated region, and eight silent mutations were identified in the tumours and cell lines (table 1). The frameshift mutations (765delC and 4675delA) were found in two ovarian tumours. These mutations result in truncation of CBP at residue 297 in the N terminal transactivation domain and residue 1548 in the histone acetyltransferase domain (HATD), respectively. Both mutations are expected to inhibit CBP function.

Three of the missense mutations were detected in ovarian tumours, two were found in breast tumours, and one was identified in the OVCAR5 ovarian cancer cell line. The A467T mutation was detected in the ovarian tumour (ov519) that also carries the 4675delA frameshift. It was unclear whether this missense mutation was in cis or trans with the frameshift variant or whether this variant in the CREB binding domain could influence CBP function. An A1354V variant in the HAT domain of CBP and an N1978D variant in the Gln rich region and C terminal transactivation domains were also identified in ovarian tumours, while the OVCAR5 cell line contained the A1081T variant. Interestingly, the N1978D alteration was also observed in an unrelated breast tumour (B110). A separate S893L variant was detected in another breast tumour (BT0167). It is not known whether these missense mutations have any influence on CBP activity.

CBP mutation analysis in germline DNA

Genomic DNA was extracted from paraffin blocks of normal ovaries and/or endometrial tissue from the individuals with missense or frameshift CBP mutations in their ovarian tumours. The exons containing these mutations were PCR amplified using the normal genomic DNA as template as described above and the PCR products were directly sequenced. The 765delC, 4675delA, 1597G→A (A467T), and 4259C→T (A1354V) mutations were not detected in the germline DNA, suggesting that these mutations occurred somatically. However, the 6130A→G (N1978D) variant from ov735 was detected in germline DNA, indicating that this mutation was inherited. These studies were not possible for the mutations identified in breast tumours because the tumours had been de-identified prior to use and we could not identify the relevant patients in order to obtain normal DNA from blood or paraffin blocks of normal tissue.

Biallelic disruption of CBP

To determine if all alleles of CBP were disrupted in the tumour tissue containing CBP mutations we obtained paraffin block specimens of the ov623, ov519, ov51, and ov735 ovarian tumours and evaluated the presence of CBP by immunostaining. Using the A22 antibody against an N terminal CBP epitope we detected CBP protein in all tumour cells (fig 1). However, when the C-20 antibody, which recognises a C terminal CBP epitope, was used, no CBP protein was detected in tumour cells from ov623 and ov519, which contain truncated forms of CBP, although staining was evident in normal stromal cells in these samples (fig 1). This result suggests that these tumour cells express only the truncated form of the CBP protein and therefore display either LOH at the CBP locus or inactivation of the second allele by other genetic or epigenetic mechanisms. Immunohistochemistry analysis of ov51 and ov735 that contain CBP missense mutations detected CBP protein in tumour and normal cells. LOH analysis at the CBP locus in these four tumours was also performed using dHPLC analysis of genomic DNA from tumour cells isolated from paraffin block tumour specimens. All tumours displayed LOH in this region.

p53 and p27^KIP1 status of CBP mutant tumours

CBP has been shown to interact with and regulate the transcriptional activity of p53,¹⁸ suggesting that CBP loss might contribute to tumorigenesis through failure to correctly induce p53 expression and activity. To test this possibility, we evaluated p53 expression in ovarian tumours with truncating and missense mutations of CBP by immunostaining. Stabilisation or overexpression of p53 was observed in ov735, ov623, and ov519, which harbour the N1978D, L297X, and L1548X mutations, respectively (fig 1), while ov51, containing the A1354V missense mutation, showed no p53 staining (data not shown). We noted that p53 was overexpressed and probably mutated in tumours that also contained CBP inactivating mutations. As inactivation of both proteins is needed for formation of some ovarian tumours, it appears that these proteins are not part of the same signalling pathway, and that inactivated CBP contributes to ovarian cancer in a p53 independent manner.

We also evaluated p27^KIP1 levels by immunostaining in these tumours because previous work showed that p27^KIP1 downregulation strongly synergises with CBP loss in T cell lymphomagenesis.¹⁵ Staining for p27^KIP1 was evident in the ov623 tumour with the L297X mutation but not in the ov519 tumour with the L1548X mutation. Thus, while p27^KIP1 downregulation is not a requirement for development of all CBP deficient human ovarian cancers, in contrast to its uniform contribution to CBP deficient lymphomagenesis in the mouse, it may contribute to development of a subset of CBP mutant ovarian tumours.

DISCUSSION

In this study, we report that two frameshift mutations of CBP were detected in a mutation screen of 75 ovarian epithelial tumours. Both mutations truncate the CBP protein N terminal of the HAT domain, and are expected to result in expression of inactive partial CBP proteins. We demonstrated by immunostaining with CBP antibodies that the second CBP allele was inactivated in both tumours with frameshift/truncating mutations. Thus, we have established for the first time that CBP is a biallelically inactivated tumour suppressor in at least 2/75 (3%) of ovarian tumours. This finding is in strong agreement with mouse model studies that have shown tumour development in animals deficient in CBP function,^13,15 and with the detection of truncating mutations in two ovarian cancer cell lines.¹² The discovery of inactivating CBP mutations identifies CBP as one of a small number of genes including p53, BRCA1, and BRCA2 that are mutated in ovarian tumours. Indeed, the frequency of truncating mutations in CBP in ovarian tumours is at least as high as that of BRCA2.

Although two truncating mutations were found, it is difficult to accurately establish the frequency of mutations in this gene based on this study. The recent description of internal CBP tandem duplications in human oesophageal cancer¹⁹ suggests that other mutations, either in the form of large genomic rearrangements or mutations in regulatory regions of the gene, undetected by the dHPLC approach described here, may be present in CBP in ovarian and/or breast tumours, and that the actual frequency of CBP mutations in ovarian cancer is higher.

We also report on the identification of three CBP missense mutations in ovarian tumours, one missense mutation in an ovarian cancer cell line, and two missense mutations in breast tumours.

Of these, the 1597G→A (A467T) variant was detected in the same ovarian tumour as the 4675delA variant. It is not known whether this mutation can alter CBP function or CBP mRNA and/or protein stability, but we do know that both alleles of CBP in this tumour fail to encode a full length CBP protein. The absence of CBP protein encoded by the second allele appears to be due to LOH within the tumour. The 6130A→G (N1978D) variant was found in an ovarian and a breast tumour from unrelated individuals and was also detected in germline DNA from one of the individuals with the mutation in their tumour DNA. LOH at the CBP locus was found in the ovarian tumour containing this mutation, and the residue is conserved throughout evolution at least as far as chicken, suggesting that it might be functionally important. However, the individual carrying the mutation in the germline does not have Rubenstein-Taybi syndrome, as would be expected of someone with a CBP inactivating mutation, suggesting that the mutation does not disrupt CBP function sufficiently to cause this syndrome and to predispose to breast and/or ovarian cancer. Interestingly, a missense mutation of the same residue (N1978S) has been detected in an individual with a mild form of Rubenstein-Taybi syndrome. Thus, different alterations in the same residue of CBP may have significantly different effects on CBP function. We were unable to derive any useful information about the disease relatedness of the 4259C→T (A1354V) missense mutation from studies of germline DNA and immunostaining of the mutant tumour for CBP, p53, and p27^KIP1. Similarly, we were unable to determine the relevance of the S893L and A1081T variants to development of the BT0167 breast tumour and OVCAR5 cell line, respectively. Additional functional studies of these variants may help to clarify the effects of the various missense mutations on CBP activity and on ovarian tumorigenesis.

Eleven unique intronic variants in the CBP gene in breast and ovarian tumours and in cell lines were detected. RT-PCR analysis of RNA from the tumours with these mutations failed to detect any splicing alterations in the CBP transcript (data not shown), suggesting that these mutations have no influence on the CBP gene or protein and make no contribution to tumour development.

Of considerable interest is the finding that the mutations detected in ovarian tumours were found in four histologically different forms of the disease (table 2). This suggests that CBP mutations are not restricted to specific histological subtypes of ovarian cancer. In addition, CBP mutations appear to be present in both early and late stage ovarian tumours. Interestingly, truncating CBP mutations have now been found in ovarian but not breast tumours, while p300 mutations have been detected in breast tumours but not in ovarian tumours,²⁰ suggesting a level of specificity in the relative contributions of the CBP and p300 paralogues to hormonally dependent epithelial tumours. However, CBP and p300 mutations have been found in breast and ovarian cancer cell lines, respectively.¹² Thus, further studies are needed to establish the relevance of CBP and p300 mutations to ovarian and breast cancer development.

Using immunostaining approaches, we were able to demonstrate that both of the tumours containing truncated forms of CBP also expressed stabilised forms of p53. The identification of CBP mutations in association with p53 mutation/overexpression suggests that disruption of CBP does not inactivate the p53 signalling pathway and that certain tumours need additional genetic events to further disrupt p53. Thus, it appears that mutant CBP contributes to ovarian tumorigenesis in a currently undefined p53 independent manner. Furthermore, the finding of CBP and p53 mutations in the same ovarian tumours suggests that p53 mutation/deregulation can synergise with mutant CBP during ovarian tumorigenesis. These findings are in full agreement with recent studies in conditional knockout models of the CBP gene in which it was found that disruption of p53 accelerates tumorigenesis in CBP deficient animals.¹⁵

In summary, the identification of CBP inactivating mutations in conjunction with disruption of the second CBP allele strongly suggests that CBP is a tumour suppressor gene that contributes to ovarian tumorigenesis in humans.