Histone acetylation is one of the best characterized epigenetic modifications and is controlled by histone acetyltransferases (HATs) and histone deacetylases (HDACs).1 The histone H4 tail contains 4 closely spaced lysines (K5, K8, K12, K16) that are subject to acetylation by various HATs.2 Histone acetylation is generally regarded as an activating modification but has also been shown to play a role in gene repression, DNA repair, DNA replication and recombination.3, 4 Considering the involvement of histone acetylation in such fundamental processes, it is not surprising that perturbation of the physiological acetylation status has been implicated in human cancer. Mutations in the p300 acetyltransferase, which acetylates many lysine residues in the histone tails, were found in primary tumors and cell lines.5 Analysis of global histone modifications, including several acetylated lysines, was shown to have prognostic value in predicting recurrence of prostate cancer after surgery.6 It was observed that HDACs have increased activity in cancer cells,7 and HDAC inhibitors have shown promise as antitumor agents in several clinical trials.8 However, it is not clear, which particular enzyme or affected lysine residue is responsible for tumorigenesis in these scenarios. It is therefore important to identify the enzymes that modify specific residues, as this will help to direct the search for potential drug targets.
Interestingly, about 60% of total histone H4 is monoacetylated, mostly at lysine 16,9 in normal human cells, whereas this acetylation is frequently lost in cancer.10 Mendjan et al.11 and Smith et al.12 recently purified the human MOF (hMOF) complex and found that similar to its Drosophila orthologue, hMOF specifically acetylates H4K16 in mammalian cells, and that depletion of hMOF leads to global reduction of H4K16 acetylation in HeLa cells.13 In addition, hMOF depleted cells showed an impaired DNA repair response following ionizing radiation,13 and hMOF was also shown to be involved in ATM function.14 More recently, hMOF was found to be able to acetylate the tumor suppressor protein p53, and this acetylation is able to influence the behavior of p53 in response to DNA damage.15 Together this suggests a role for hMOF in transcriptional regulation, cell proliferation, differentiation and the DNA repair response.12, 13
Since hMOF is involved in these important cellular processes with obvious links to cancer, we were interested in understanding whether hMOF expression is lost in tumors and whether loss of hMOF is the reason for the decrease in the levels of H4K16Ac in tumors. We therefore investigated hMOF protein expression and H4K16 acetylation status by immunohistochemistry (IHC) on tissue microarrays (TMAs) in 298 primary breast carcinomas and 180 medulloblastomas, and corresponding normal tissues.
CI, confidence interval; HAT, histone acetyltransferase; HDAC, histone deacetylase; HE, hematoxyline and eosine; H4K16, histone 4 lysine 16; IHC, immunohistochemistry; QRT-PCR, quantitative real-time PCR; TMA, tissue microarray; WHO, World Health Organization.
Material and Methods
Tumor material, patient characteristics and preparation of TMAs
A commercially available human breast cancer TMA (CBA-5-SBC) including 40 primary samples of breast carcinomas and 8 controls of normal mammary gland was purchased from Biocat (Heidelberg, Germany). Details on collection and preparation of specimens can be obtained at www.biocat.de.
Two custom TMAs, one each for the investigation of breast carcinomas and medulloblastomas, were constructed. Hematoxyline and eosine stained sections from all paraffin blocks were prepared to define representative tumor regions as previously described.16 All tumors were arrayed in duplicate.
For the breast TMA consisting of 258 breast cancer samples, 46 mastopathies and 53 normal glandular breast samples, specimens were consecutively collected at the Department of Gynecology, Heidelberg University Hospital, Germany, between 2001 and 2003. Histological analysis and standard IHC stains were carried out by the Department of Pathology, University of Heidelberg. Diagnoses were confirmed by a second pathologist. None of the patients had received irradiation or chemotherapy prior to surgery. Patient characteristics of all samples from breast cancer patients and healthy donors contained on the commercially available as well as the self-constructed TMAs are summarized in Supplementary Table I (numbers are added up from both arrays).
Breast cancer patients (n = 100) in the mRNA expression profiling study were part of two phase I/II trials which have been previously reported.17, 18 Tumor sizes ranged from 2 to 10 cm in diameter. All samples were taken at time of diagnosis, before applying chemotherapy and/or radiotherapy, by core biopsy from the primary tumor and frozen immediately.
Medulloblastoma samples (n = 180) were randomly collected at the Department of Neuropathology, Burdenko Neurosurgical Institute, Moscow, Russia, between 1993 and 2003. Diagnoses were confirmed by 2 neuropathologists according to the 2000 WHO classification. Approval to link laboratory data to clinical data was obtained by the Institutional Review Board. None of the patients had received irradiation or chemotherapy before collection of specimens. Metastatic state (M-stage) was determined by magnetic resonance imaging and cerebrospinal fluid cytopathology at diagnosis. Clinical and histopathological data are summarized in Supplementary Table II.
IHC and antibodies
Staining of TMAs was performed according to standard protocols using the Vectastain Elite ABC Kit (Linaris, Germany) and DAB substrate Kit (Linaris, Germany). Paraffin sections were de-waxed with xylene and passed through ethanol series (100%, 96%, 70%; 2 min each). For antigen retrieval, sections were treated by microwave irradiation (750 W, 15 min) in citrate buffer (pH 6.0) and then incubated with the primary antibody. Polyclonal rat antibody against hMOF (previously described in13) and monoclonal antibody against p300 (US Biological, USA) were used at 1:100 dilution, while polyclonal H4K16Ac (Chemicon, USA) antibody was used at 1:1,000 dilution. For detection, TMAs were incubated for 30 min with biotinylated secondary antibody, and for 30 min with ABC reagent. The slides were stained with 3, 3′-diaminobenzidine, counterstained with haematoxyline, and mounted. As negative controls, primary antibodies were substituted with commercially available control reagents according to the manufacturer's instructions. IHC scoring was performed essentially as described by Brabletz et al.19 Briefly, all individual sections were scored for intensity (0, no staining; 1, weak; 2, moderate; 3, strong) and percentage of staining cells (0, 0%; 1, 1–2%; 2, 3–10%; 3, 11–50%; 4, 51–100%). A final score was achieved by multiplying both scores and averaging the scores of the 2 replicates (range: 0–12). For classification into low and high expressing samples, final scores of 0–6 were regarded as low, 7–12 as high staining. All analyses were carried out for 2 replicates of each tumor by 2 investigators (B. Straub and S. Pfister) who were blinded to clinical and molecular variables. Tumors were subjected to statistical analysis if at least 1 replicate of a tumor was left on the TMA after immunostaining (referred to as informative cases in the text).
For investigation of pairwise relationships of prognostic factors, Spearman's correlation coefficient was used together with its 95% confidence interval (CI). Estimation of the survival time distribution of patients with primary tumors was carried out by the method of Kaplan and Meier. Multivariable analysis of the dependence of survival times since the date of diagnosis on hMOF protein levels (categorized 0–6 vs. 7–12), H4K16Ac and clinical parameters (age (years), gender, administration of chemotherapy, M-stage, histology, level of resection) in medulloblastoma was performed by Cox proportional hazards regression. To provide quantitative information of the relevance of results, hazard ratios (HR) and their corresponding 95% CI s were computed. All statistical analyses were performed using the statistical software environment R, 2.1.1.,20 and StatXact 6.0 (Cytel Software Corp).
For measuring hMOF mRNA abundance in medulloblastoma, a reference of total RNA obtained from normal human cerebellum of 24 individuals (age 16–70; BD Bioscience, USA) was used for tissue-specific normalization. Reference for breast carcinomas consisted of pooled total RNA from normal glandular breast tissue (BD Bioscience and Biocat, Germany) obtained from 3 healthy individuals (age 26–36). Each cDNA sample was analyzed in triplicate with ABI PRISM 7700 (Applied Biosystems, USA) using Absolute SYBR Green ROX Mix (ABgene, UK) according to the manufacturer's instructions. Two endogenous housekeeping genes (PGK1, LMNB1) were used as standards. All primers were tested to exclude amplification from genomic DNA. The relative quantification of the RNA of interest in comparison to the housekeeping genes was calculated according to a previously published algorithm.21 Oligonucleotide sequences are available upon request.
Frequent downregulation of hMOF mRNA in primary breast carcinoma
Additional analysis of a recently performed mRNA expression profiling study of 100 breast cancer samples22 revealed significant (>2-fold) downregulation of hMOF mRNA in 41% (40/98) of patients, whereas none of the patients showed significant (>2-fold) upregulation of hMOF (Fig. 1a). In two cases, information on hMOF expression was filtered out during quality assessment. To further validate the frequent downregulation of hMOF RNA expression in primary breast carcinoma, mRNA expression of hMOF in 9 random cases of newly diagnosed invasive breast carcinomas was examined by quantitative real-time PCR (QRT-PCR). Consistent with the results obtained by mRNA expression profiling, we found significantly (>2-fold) reduced hMOF mRNA expression in 7/9 cases (Fig. 1b) using mRNA from normal glandular breast tissue as reference (pool of 3 donors).
Reduction of hMOF protein in primary breast carcinoma
The results above clearly show reduction of hMOF expression at the RNA level. We were interested to study whether or not this resulted in the reduction of hMOF protein levels and if so, to what extent. We therefore screened 298 primary breast carcinomas for hMOF protein expression and H4K16 acetylation by immunohistochemistry (IHC). For the investigation of breast carcinomas, we combined the results of a commercially available TMA containing sections from 40 breast carcinomas and 8 controls of normal mammary gland with a custom-made array consisting of sections from 258 primary breast carcinomas, 46 mastopathies and 53 normal glandular breast samples. The custom-made TMA was constructed from freshly frozen and clinically as well as histologically well-characterized tumor samples and respective control tissues. All tumor samples were arrayed in duplicate. Clinical information is given in Supplementary Table I. After immunostaining, individual sections were scored as outlined in the Methods section. Tumors were included in the statistical analysis if at least 1 replicate was left on the TMA after immunostaining (referred to as informative cases from here onwards).
In a total of 280 informative breast cancer samples, we observed markedly reduced or undetectable hMOF protein expression in 18% of cases (49/280; Table I and Fig. 2). In contrast, all informative 54 normal tissues (glandular breast tissue from healthy donors) and 32 informative sections obtained from patients with mastopathies displayed abundant hMOF protein expression. Reduced hMOF expression was more frequently found in infiltrating ductal carcinomas (43/206) than in infiltrating lobular carcinomas (2/41). Interestingly, none of the low-risk tumors (mucinous, tubular and in situ carcinomas) included in our series showed loss of hMOF expression (0/20). Consistent with hMOF being the H4K16 specific acetyltransferase, we found that H4K16 acetylation of all samples closely correlated with hMOF protein expression (Spearman rank correlation ρ = 0.52; Table I and Fig. 2). In contrast, the p300 acetyltransferase protein was abundantly expressed in tissue sections from the same tumors (Fig. 2). Taking routine markers into account, we observed positive correlation with estrogen receptor (ρ = 0.17) and progesterone receptor status (ρ = 0.19) and negative correlation with tumor grading (ρ = −0.20) and expression of the proliferation marker Ki67 (ρ = −0.16) in breast carcinomas. No significant correlation was found between hMOF expression and p53 expression, Her2/neu expression, metastatic stage or patients' age upon diagnosis (Supplementary Table I).
Table I. Expression of hMOF Protein and Correlation with H4K16 Acetylation
Informative cases: Tumors were subjected to statistical analysis if at least one replicate of a tumor was left on the TMA after immunostaining. IHC scoring was performed as outlined in the Materials and Methods section. CI, Confidence Interval.
DNA copy-number losses at the hMOF locus in medulloblastoma
In a recent effort for the genome-wide detection of copy-number aberrations in a total of 102 medulloblastomas by array-based Comparative Genomic Hybridization (Ref.23 and unpublished data), we detected heterozygous deletions of chromosome band 16p11.2 in 11/102 (11%) of cases. In one of these cases, we found a distinct deletion of 2.4 megabases spanning the hMOF locus (Fig. 3a). In most cases we detected heterozygous loss of chromosome arm 16p or monosomy 16.
Frequent downregulation of hMOF mRNA expression in medulloblastoma
Because of the frequent deletion of the hMOF genomic locus in medulloblastoma, we were interested in hMOF mRNA expression in these tumors. Therefore, we investigated 14 randomly chosen medulloblastoma samples by QRT-PCR. In these tumors, we found significant (>2-fold) downregulation in 11/14 cases (79%) using normal cerebellum (pool of 24 unaffected donors) as a reference, whereas none of the tumors showed significant upregulation (Fig. 3b).
Loss or reduction of hMOF protein in primary medulloblastoma
For the analysis of hMOF protein expression in medulloblastoma, we used a TMA containing 180 histologically and clinically well-characterized tumor samples (Supplementary Table II). In this series, hMOF protein expression could be analyzed for 164 informative samples and was reduced or absent in 40% of cases (66/164). hMOF expression was more frequently lost in anaplastic (17/41) and classic (44/105) medulloblastomas than in desmoplastic tumors (5/18). As observed for breast carcinoma and control tissues, hMOF expression significantly correlated with H4K16 acetylation (Table I and Fig. 4). In addition, we performed IHC on slices of adult and embryonic cerebellum as a reference for medulloblastoma. We found abundant nuclear hMOF protein expression and H4K16 acetylation in both embryonal and adult cerebellum (Fig. 4).
hMOF protein expression is a prognostic marker for medulloblastoma
As clinical follow-up data was available for the medulloblastoma patients, we studied the potential prognostic value of hMOF protein expression for clinical outcome. We identified hMOF protein expression as a prognostic marker for overall survival when stratifying patients into 2 risk groups: low expression (IHC score 0–6) versus high expression (IHC score 7–12) in a univariate analysis (Log rank test: p = 0.03). Patients harboring tumors with low hMOF expression show a significantly worse overall survival (Fig. 5a). Moreover, multivariate Cox regression analysis including categorized hMOF expression (hMOF cat), H4K16 acetylation and clinical parameters such as age, gender, histology, M-stage, chemotherapy and extent of resection identified hMOF expression as an independent prognostic marker for clinical outcome (p = 0.02; Fig. 5b). Importantly, H4K16 acetylation also correlated with prognosis in this multivariate analysis (p = 0.05). Established prognostic factors, such as administration of chemotherapy (yes:no; p = 0.0001) and M-stage (M1:M0; p = 0.01) were confirmed in our set. The effect of age (p = 0.70), gender (female:male; p = 0.42) and level of resection (R1:R0; p = 0.82) was not statistically significant.
We here report the first extensive survey on the expression of the HAT hMOF and its corresponding modification in tumor tissues.
In this study, we investigated hMOF protein expression and H4K16 acetylation in a large series of primary breast carcinomas by immunohistochemistry. In contrast to all informative control sections (54 samples of glandular breast tissue, 32 mastopathies), we found reduction or complete loss of hMOF expression in a large subset of primary breast carcinomas (18%). For all investigated tissues, hMOF expression strictly correlated with H4K16 acetylation and loss of this enzyme correlated with hypoacetylation of H4K16. Reduced hMOF expression was found to be much more frequent in infiltrating carcinomas than in low-risk tumors, such as mucinous, tubular and in situ carcinomas, and was much more frequent in invasive ductal carcinomas than in lobular carcinomas.24 Correlation of hMOF expression with well-established prognostic markers in breast cancer, such as estrogen and progesterone receptor status and tumor grading, suggests that hMOF might have an important role in breast cancer development and progression. However, clinical observation time in our cohort was too short for analyses of overall survival or event free survival in these patients. Prospective trials will therefore be necessary to test the potential of hMOF staining for routine applications in the management of breast carcinomas.
For medulloblastoma, we found significant downregulation of hMOF mRNA expression in 11/14 (79%) of tested samples. Applying immunohistochemsitry to a large series of medulloblastomas, we found significant reduction or loss of protein expression and concomitant loss of H4K16 acetylation in 40% of tumors. Moreover, the value of hMOF as a prognostic marker is illustrated by our findings in this retrospective study investigating 180 adult and pediatric medulloblastomas of all 3 histological subtypes (desmoplastic, classic, anaplastic). Loss of hMOF was more frequent in the more aggressive anaplastic and classic medulloblastomas than in desmoplastic tumors that are known to have a much better overall survival.25 Most importantly, hMOF expression can be used as a molecular prognostic marker in medulloblastoma, as it significantly distinguishes two risk groups in univariate (p = 0.03) and multivariate analyses (p = 0.02). When only looking at patients, who had received maintainance chemotherapy after surgical resection and radiotherapy, hMOF expression again distinguishes 2 subgroups of patients with significantly different outcome (data not shown). As expected for the major hMOF target, loss of H4K16 acetylation in the multivariate analysis also showed association with poor prognosis (p = 0.05). Having identified hMOF as a novel independent prognostic marker for medulloblastoma outcome, it will be of great interest for clinical routine to test the impact of hMOF expression on survival of medulloblastoma patients in prospective trials.
The molecular mechanism linking loss of hMOF expression to cancer will be an exciting avenue for future research. There are several possible mechanisms that could explain such a scenario: Global loss of H4K16 acetylation in the absence of hMOF affects transcription, leading to aberrant expression of key genes in carcinogenesis. Alternatively, loss of hMOF and H4K16 acetylation could alter global chromatin structure making it more susceptible to DNA damage26 or hMOF may play a more direct role in regulating the response to DNA damage.15 It is important to emphasize that these hypotheses are not mutually exclusive and there may be combinations of different mechanisms. Additionally, it will be of great interest to compare hMOF protein expression and H4K16 acetylation in cancer tissues and healthy tissues of the same patients to determine, if a generally low hMOF activity might predispose patients for disease.
In summary, the results presented here indicate that hMOF is the major H4K16 acetyltransferase in human cells and downregulation of hMOF is a frequent event in tumors.
We thank the members of the Akhtar Lab and Dr. Bernhard. Radlwimmer from Dr. P. Lichter's Lab as well as Dr. A. Kulozik, Dr. A. Schneeweiss and Dr. S. Kass for discussions, Mr. G. Toedt and Dr. M. Hahn for their help in the analysis of microarray results, and Mr. A. Benner for invaluable help with statistical analysis. Ms. A. Wittmann, Ms. S. Hofmann, Ms. M. Schlotter and Ms. F. Devens are acknowledged for excellent technical assistance and Dr. G. Reifenberger for providing tissue sections from adult and embryonal cerebellum. Dr. A. Korshunov (Burdenko Neurosurgical Institute, Moscow, Russia) has generously provided tumor samples and clinical data from all medulloblastoma patients included in this study. Dr. S.P. is supported by a Young Investigator Award of the Medical Faculty of Heidelberg (Germany). Dr. S.R. is funded by an International Human Frontiers Science Program Organization long term fellowship.