Jörg Ellinger, Klinik und Poliklinik für Urologie, Universitätsklinikum Bonn, Sigmund-Freud-Strasse 25, 53105 Bonn, Germany. e-mail: firstname.lastname@example.org
Study Type – Prognosis (case series)
Level of Evidence 4
What’s known on the subject? and What does the study add?
Epigenetic alterations play an essential role during carcinogenesis. While DNA methylation has been extensively studied in bladder cancer, the relevance of histone modifications remains to be clarified. Earlier studies suggested that global histone modification levels are predictive for patients’ outcome in various tumour entities (e.g. prostate, lung, breast and kidney cancer). The possibility to determine global histone modification levels easily and inexpensively using immunohistochemistry increases a potential routine use in the future. Our aim was therefore to investigate the global levels of histone H3K4 and H4K20 mono-, di- and trimethylation. For this purpose we prepared tissue microarrays with non-muscle-invasive bladder cancer (NMIBC), muscle-invasive bladder cancer (MIBC), bladder cancer metastases (METS) and normal urothelium (NU) tissue to compare global H3K4 and H4K20 methylation in these tissues, as well to assess the prognostic value of histone modifications.
We show that global histone modification levels (H3K4me1, H3K4me3, H4K20me1, H4K20me2, H4K20me3) are lower in bladder cancer than in NU tissue. Furthermore, there was a decrease of histone modification levels (H3K4me1, H4K20me1, H4K20me2, H4K20me3) from NU over NMIBC and MIBCto METS. Histone modifications are correlated to advanced pathological stage in NMIBC and MIBC. Furthermore, H4K20me3 appeared to be a significant and independent prognostic predictor of bladder cancer-specific survival in patients with MIBC undergoing radical cystectomy. Our findings therefore provide a rationale for further investigation of histone modifications and their manipulation in bladder cancer.
• To determine the role of global histone methylation as a prognostic parameter in patients with bladder cancer.
PATIENTS AND METHODS
• We used a tissue microarray with samples from patients with non-muscle-invasive bladder cancer (NMIBC; n= 161), muscle-invasive bladder cancer (MIBC, n= 127), normal urothelium (NU; n= 31) and bladder cancer metastases (METS; n= 31) to determine global histone methylation (me) levels at histone H3 lysine 4 (H3K4) and H4K20.
• Global histone modification levels (H3K4me1, H3K4me3, H4K20me1, H4K20me2, and H4K20me3) were lower in bladder cancersamples than in NU tissue
• Global levels of H3K4me1, H4K20me1, H4K20me2 and H4K20me3 were decreasing from NU over NMIBC and MIBC to METS.
• H4K20me1 levels were increased in patients with NMIBC with advanced pTstage and less differentiated bladder cancer.
• In patients with MIBC, pTstage was negatively correlated with H3K4me1, H4K20me1 and H4K20me2 levels.
• H4K20me3 levels were significantly correlated in a univariate and multivariate model with bladder cancer-specific mortality after radical cystectomy in patients with MIBC.
• Global histone methylation levels may help to identify patients with bladder cancerwith poor prognosis after radical cystectomy.
papillary urothelial neoplasm of low malignant potential
non-muscle-invasive bladder cancer
normal urothelial tissue
bladder cancer metastases
muscle-invasive bladder cancer
carcinoma in situ
transurethral resection of the bladder
In 2009, 70 980 new cases of cancer of the urinary bladder were diagnosed in the USA and 14 330 patients died from bladder cancer . Predictors for clinical outcome of patients with bladder cancer with superficial and muscle-invasive tumours are:TNM stage, WHO grade, presence of carcinoma in situ, multifocality and lymphatic invasion [2,3]. Nevertheless, there are differences in the clinical course despite similar tumour characteristics. Accordingly, it would be desirableto subclassify bladder cancer to predict the outcome more accurately, to improve treatment and to find new therapy approaches, which are adapted to the tumour conditions.
It has become more and more apparent that epigenetic alterations play an essential role in bladder carcinogenesis. DNA methylation is the best-known epigenetic alteration and it was earlier shown that DNA hypermethylation of tumour suppressor genes plays an important role in the development, recurrence, progression and cancer-specific survival (CSS) in bladder cancer . Other epigenetic mechanisms, such as histone modifications have been far less comprehensively studied, but are also relevant: Histone acetylation and histone methylation are involved in the regulation of p16 in bladder cancercells. Histones are post-translationally modified at their N-terminal tails by acetylation (ac), methylation (me), phosphorylation, ubiquitination, sumoylation, ADP-ribosylation and deimination . These modifications either act by disrupting chromatin contacts or by recruiting non-histone proteins to the chromatin; they thereby influence transcription, replication, DNA repair and chromosome condensation. Histone lysine (HxKy) methylation appears in a mono-, di- or tri-methylated form, and the site of histone methylation is causative for its effect on transcriptional activity: Euchromatin and therefore transcriptional activation, is characterized by methylation at H3K4, H3K36 or H3K79, while methylation at H3K9, H3K27 and H4K20 is associated with transcriptional repression .
Several recent studies have shown that global levels of histone modifications are predictive of cancer patients’ outcome. Seligson et al.  were the first to show low levels of di-methylation(H3K4me2) and acetylated H3K18 (H3K18ac) in patients with increased risk of prostate cancer recurrence. Later, prognostic relevance independent from established cancer predictors (e.g. TNM staging, grading) was also shown in patients with breast cancer (low H3K4me2, H4K20me3, H3K9ac, H3K18ac; H4K12ac, H4R3me2 levels; ), lung cancer (low H3K4me2 and high H3K9ac levels; ) and RCC (low H3K4me levels ). Barbisan et al.  reported decreased global levels of H3K9ac in recurrent papillary urothelial neoplasm of low malignant potential (PUNLMP) compared to non-recurrent PUNLMP, but the study comprised only 20 cases. Another study in patients with bladder cancer showed that phosphorylation of histone H2AX at serine 139 was associated with lower risk of bladder cancer recurrence after transurethral resection .
Histone modifications seem to be a universal prognostic marker in patients with cancer, and the possibility to determine global histone modification levels easy and inexpensive using immunohistochemistry increases a potential routine use in the future. The aim of the present was therefore to investigate the global levels of histone H3K4 and H4K20 mono-, di- and tri-methylation. For this purpose we prepared tissue microarray with 350 specimens (non-muscle-invasive bladder cancer [NMIBC], muscle-invasive bladder cancer [MIBC], bladder cancermetastases [METS] and normal urothelial tissue [NU]) to compare global H3K4 and H4K20 methylation in these tissues, as well to assess the prognostic value of histone modifications.
PATIENTS AND METHODS
A tissue microarray (TMA) with samples of NMIBC (n= 161) and MIBC(n= 127) was prepared from formalin-fixed, paraffin-embedded tissue specimens (Table 1 for clinico-pathological characteristics). We also investigated histologically confirmed NU (n= 31) tissue from patients undergoing surgery for bladder cancer and METS (n= 31; one liver and 30 regional lymph node metastases). Three representative tissue cores per patient were arrayed using a manual device (Beecher Instruments, Sun Prairie, WI, USA) to obtain a representative image of the tumour. Samples were selected from the archival files of the Department of Pathology at the Universitätsklinikum Bonn according to tissue availability and were not stratified for any known preoperative or prognostic factor. The specimens were derived from patients undergoing transurethral resection of the bladder (TURB; NMIBC, n= 149; MIBC n= 18) or radical cystectomy (RC; NMIBC, n= 12; MIBC, n= 109) at the department of Urology at the Universitätsklinikum Bonn between 1991 and 2009. In 18 cases of MIBC the TURB material was used because of better suitability for TMA construction than in the RC material; all these patients underwent subsequently RC and the pathological staging from the date of RC was used for our study. In the NMIBC cohort, 12 patients underwent RC (i.e. for T1G3 bladder cancer) and the material from RC was more appropriate for TMA preparation and therefore used; follow-up information was missing for all 12 patients, thus recurrence/survival analyses are not biased. All cases were reviewed by a pathologist (L.C.H.) and staging/grading was performed according to the WHO classification. Follow-up information was available for all patients. Among patients with NMIBC, 97 had disease recurrence and seven died from bladder cancer. The mean (median; range) follow-up period was 72 (60; 1–401) months. In patients with MIBC the mean (median; range) follow-up period was 44 (25; 1–238) months; in 79 patients there was disease progression and cancer-related death occurred in 40 patients. The study was approved by the local ethic committee (ethic vote 289/08).
Table 1. Clinico-pathological characteristicsof patients with NMIBC, MIBC and NU
n.a., not applicable.
Mean(median; range)age, years
67.7 (70.0; 38–94)
Lymph node metastasis
Immunohistochemical staining of H3K4me1, H3K4me2 and H3K4me3 was described by our group . We also analyzed histone H4K20 methylation, and the staining procedure was adopted for these analyses. In brief, paraffin sections (5 µm) were cut from the TMA block. Deparaffinization was done using xylene and the sections were rehydrated with graded ethanol. Slides were placed in target retrieval solution (citrate buffer, pH 6.0) and heated for 10 min at boiling temperature using a microwave. After cooling for 15 min, endogenous peroxidase activity was blocked by treatment with 3% H2O2 for 10 min. The sections were washed with Tris-buffered saline. After a 15-min protein block with normal swine serum, the primary antibodies H3K4me1 (dilution: 1:250; Millipore, Lake Placid, NY, USA; catalog number 07-436; lot number 30218), H3K4me2 (dilution: 1:1000; Millipore; catalog number 07-030; lot number 26335), H3K4me3 (dilution: 1:25; Cell Signaling, Dnavers, MA, USA; catalog number 9727; lot number 1), H4K20me1 (dilution: 1:2000; Active Motif, Rixensart, Belgium; catalog number 39176; lot number 134), H4K20me2 (dilution: 1:250; Millipore; catalog number 07-747; lot number 30586) and H4K20me3 (dilution: 1:400; Millipore; catalog number 07-463; lot number 31392) were applied overnight at 4 °C. Immunohistochemical staining was performed using the streptavidine-biotin-peroxidase complex technique (LSAB+, DAKO Cytomation, Glostrup, Denmark). The biotin-conjugated secondary antibody was incubated for 30 min at room temperature and the avidin biotin enzyme reagent alike. The peroxidase was developed with the aminoethylcarbazole (AEC) system (DAKO). The sections were counterstained with haematoxylin and mounted. Negative control sections were identical TMA sections without the primary antibody. The sections were automatically scanned using the Mirax Scan (Carl Zeiss Microlmaging, Göttingen, Germany), and were virtually evaluated using the Mirax Viewer (Carl Zeiss Microlmaging).
The immunostaining results were semi-quantitatively recorded. The number of urothelial cells showing nuclear staining was estimated per core and scaled: 0, no positive cells; 1, 1–25% positive cells; 2, 26–50% positive cells; 3, 51–75% positive cells; and 4, 76–100% positive cells. These scores were multiplied with an intensity scale (0, negative; 1, weak; 2, moderate; and 3, intensive staining), and the mean staining for a patient was calculated; odd values were rounded up/down. The immunohistochemical staining was evaluated without knowledge of the specimen identity and the clinico-pathological parameters.
Clinico-pathological variables were correlated with the global histone methylation levels using the Spearman’s ρtest, the Mann–WhitneyU-test and the Kruskal–Wallistest, as appropriate. The Cox proportional hazard regression analysis was used to correlate the period of progression-free survival and CSS after surgery with global histone methylationlevels/clinical-pathological variables. P < 0.05 was considered to indicate statistical significance.
GLOBAL HISTONE METHYLATION LEVELS ARE DIFFERENT IN NU TISSUE AND BLADDER CANCER TISSUE
We first analyzed whether histone modification levels in bladder cancer and NU tissue were different. All investigated histone modification except of H3K4me2 (P > 0.1) were significantly less abundant in MIBC and NMIBC than in NU tissue (P < 0.001). METS showed also lower levels of histone modification (all P < 0.001) with exception of H3K4me2 (P= 0.853). Interestingly, global H3K4me1, H4K20me1, H4K20me2 and H4K20me3 levels were decreasing from NU over NMIBC and MIBC to METS (Fig. 1 and Fig. 2; Table 2).
Table 2. Global histone modifications at H3K4 and H4K20 are different in NU, NMIBC, MIBC and METS as determined using the Mann–Whitney U-test
P values – H3K4
P values – H4K20
NU vs NMIBC
NU vs MIBC
NU vs METS
NMBIC vs MIBC
NMIBC vs METS
MIBC vs METS
HISTONE MODIFICATIONS ARE ASSOCIATED WITH POOR PROGNOSIS
Mono-methylation of H3K4 and H4K20 seems to be of prognostic relevance in patients with bladder cancer because global levels of these modifications were correlated with staging and grading. For example, H4K20me1 levels were increased in patients with NMIBC with advanced pTstage (pTa vs pT1, P < 0.001; pTa vs carcinoma in situ(pTis), P < 0.001) and less differentiated bladder cancer (G1 vs G2, P= 0.005; G1 vs G3, P < 0.001; G2 vs G3, P= 0.019). Contrary to this, in patients with MIBC the pT stage was negatively correlated with H3K4me1 (pT2 vs pT3, P= 0.007; pT2 vs pT4, P < 0.001; pT3 vs pT4 P= 0.005), H4K20me1 (pT2 vs pT3 and pT2 vs pT4, P= 0.005) and H4K20me2 (pT2 vs pT3, P= 0.049; pT2 vs pT4, P < 0.036) levels. A detailed summary is provided in Table 3.
Table 3. Global histone modifications at H3K4 and H4K20 are correlated with pathological stage and grade in patients with NMIBC and MIBC
P values – H3K4
P values – H4K20
indicates positive correlations (P < 0.05),
respectively negative correlations as determined using the Mann–Whitney U-test.
Furthermore, we performed Cox proportional hazard models to analyse the prognostic value of global histone modification levels on patients’ outcome. There was no significant correlation of any histone modification level and cancer recurrence or CSS in patients with NMIBC (Table 4). However, increased levels of H4K20me3 were significantly correlated with bladder cancer-specific mortality afterRC (P= 0.047; hazard ratio [HR] 1.103) in patients with MIBC. In a multivariate analysis that also included pT stage, grade and metastasis, H4K20me3 levels were still predictive for bladder cancer-specific mortality (P= 0.006; HR1.172; Table 5 and Fig. 3).
Table 4. Global histone modifications at H3K4 and H4K20 are not correlated with bladder cancer recurrence and CSS in patients with NMIBC
Recurrence-free survival univariate analysis
CSS univariate analysis
HR (95% CI)
HR (95% CI)
Table 5. Global histone modifications at H4K20me3 are predictive for CSS in patients with MIBC
Recurrence-free survival univariate analysis
CSS univariate analysis
CSS multivariate analysis
HR (95% CI)
HR (95% CI)
HR (95% CI)
To the best of our knowledge, this is the first study investigating global histone lysine methylation levels in bladder cancer tissue. We show that global levels of H3K4 and H4K20 methylation are decreased in bladder cancer compared with NU tissue, and are lower in METSthan in the primary tumour; also, histone modification levels seem to be lower in MIBC than NMIBC. Global DNA methylation and histone H3K9ac levels have been investigated in bladder cancer tissue: Barbisan et al.  showed that global levels of 5-methylcytosine increase during bladder carcinogenesis. The number of 5-methylcytosine positive cells increased from NU over PUNLMP to high-grade urothelial carcinoma. Somewhat controversially Seifert et al.  reported lower levels of 5-methylcytosine in bladder cancer than in NU in urine cytologies. A correlation of H3K9ac and bladder cancergrade was not observed earlier . However, we assume that the epigenetic phenotype is related to bladder cancer aggressiveness.
This hypothesis is supported by the finding of an increased risk of bladder cancer- related death in patients with high levels of H4K20me3 (Fig. 2 and Table 5). A prognostic role was also reported in earlier studies: H4K20me3 levels were significantly decreased in patients with stage I lung adenocarcinoma and poor survival . In patients with breast cancer, low H4K20me3 levels were also correlated with clinico-pathological variables indicative for poor prognosis . Interestingly, low H4K20me3 levels were correlated with poor prognosis in lung and breast cancer, whereas in the present study high H4K20me3 levels were associated with a high risk of cancer death in patients with MIBC. This difference may be due to different expression profiles of histone modifying enzymes in various tumour entities.
To date, three H4K20 methyltransferases have been discovered. The bulk of H4K20me1 is created by the H4K20- mono-methyltransferase KMT5A (also termed as Pr-Set7) . However, loss of KMT5A in vivo did not only result in lower H4K20me1 levels, but also in lower levels of H4K20me2 and H4K20me3 . KMT5B (Suv4-20) and KMT5C (Suv4-20h2) are generating H4K20me2 and H4K20me3 . KMT5C levels were correlated with global H4K20me3 levels in breast cancer cell lines . Lower levels of KMT5C were also correlated with H4K20me3 levels in a murine model for liver cancer . To date, there is no known demethylase of H4K20, and it was suggested that H4K20 methylation is a stable mark. However, it was also assumed that histone demethylases do not exist until Shi et al.  reported the discovery of the nuclear amine oxidase homolog LSD1; since then numerous histone demethylating enzymes have been discovered  and it seems possible that H4K20-demethylases will also be discovered.
Although histone modifications influence the expression of specific genes, altered global histone modification levels are mostly a result of changes at repetitive DNA sequences. Lower global H3K4me3 levels in leukaemia cells were mainly attributable to loss at the juxtacentromeric satellite element Sat2, the subtelomeric repetitive element D4Z4 and the acrocentric chromosomal region NBL2, but not due to loss at specific histone-silenced genes orCpG islands . Notably, global 5-methylcytosine levels were also mainly decreased in these regions, and correlated to histone modification levels . Thereby, epigenetic alterations might contribute to chromosomal instability: an increased rate of loss of heterozygosity was due to hypomethylation at the centromeric and pericentric regions .
The discovery of high H4K20me3 levels in high-risk bladder cancer also offers the possibility of identifying those patients who benefit would from adjuvant therapy, although chemotherapy is not recommended for routine use to date . Chemotherapy for patients with metastatic bladder cancer is based on cisplatin (i.e. methotrexate, vinblastine, doxorubicin and cisplatin; gemcitabine and cisplatin); however, ‘epigenetic drugs’ (i.e. histone deacetylase (HDAC) inhibitors) are already being investigated in phase I clinical trials and may be effective in bladder cancer: suberoylanilidehydroxamic acid induced partial remission in two patients with metastatic bladder cancer ; suberoylanilidehydroxamic acid was approved for the treatment of cutaneous T-cell lymphoma by the USA Food and Drug Administration in 2006. Another HDAC inhibitor (CI-994) lead, combined with paclitaxel and carboplatin, to complete remission in a patient with metastatic bladder cancer . Therapy success was accompanied by increased global histone acetylation levels in post-treatment biopsy [24,25]. Growing knowledge on other histone modifications might also lead to inhibitors against other histone modifying enzymes. It should also be noted that global histone modification patterns (H3K4me2, H3K9me2, H3K18ac) were predictive of response to adjuvant 5-fluorouracil therapy in patients with pancreatic adenocarcinoma in the Radiation Therapy Oncology Group 9704 trial . Assessing the level of histone modifications before therapy may therefore help to identify patients who might benefit from targeted epigenetic therapies.
Some limitations of the present study should also be mentioned: The NU tissue was ‘normal’ tissue from patients with bladder cancer undergoing RC. This tissue may already contain molecular alterations because of the so-called ‘field effect in cancer’. However, the collection of NU tissue from healthy individuals is from an ethical point of view difficult, and thus the use of NU tissue from cancerous bladders is a reasonable alternative. It should be noted, that histone modification levels were different in NU from cancer-bearing bladders and bladder cancer tissue. Furthermore, in 18 patients with MIBC we used tissue derived from TURB rather than tissue from the RC specimen, because the material appeared better. We assume that this approach is feasible because the same tumour in the TURB and the RC specimen should have a similar molecular biology. It should also be noted that correlation analyses with the novel WHO grading (urothelial papilloma, PUNLMP, low-grade urothelial carcinoma, high-grade urothelial carcinoma) were not performed; instead the old, but still recommended for use, grading system (G1, G2, G3) was applied.
In conclusion, histone modifications are altered in bladder cancer. Immunohistochemical evaluation of H4K20me3 levels in patients with MIBC may help to identify patients with poor prognosis, as high levels of H4K20me3 increased the risk of bladder cancer-specific mortality.
We thank Mrs Doris Schmidt for technical assistance.
CONFLICT OF INTEREST
JörgEllinger has a research grant from the Deutsche Forschungsgemeinschaft (EL 623/1-1). The authors disclose any conflict of interest