Lars Dyrskjøt, Department of Molecular Medicine, Aarhus University Hospital, Skejby, Brendstrupgaardsvej 100, 8200 Aarhus N, Denmark. e-mail: email@example.com
What's known on the subject? and What does the study add?
Several studies have shown that defects in DNA-damage response are associated with good survival after chemotherapy and radiotherapy. Furthermore, loss of cell cycle regulators may be prognostic indicators of poor survival after cystectomy. However, the potential clinical impact of previous findings is hampered by insufficient validation of significant results in suitable cystectomy and radiotherapy cohorts.
Here we use a large cohort of patients receiving radiotherapy to successfully validate the importance of MRE11 as a predictive marker of disease-specific survival (DSS). Furthermore, using two independent patient cohorts we show for the first time that TIP60 is a predictive marker of DSS after cystectomy. We show that combined use of TIP60 and MRE11 may hold the potential to guide treatment decisions.
• To determine the association between the proteins: tat-interactive protein 60 kDa (TIP60), p16, meiotic recombination 11 homolog (MRE11), phosphorylated ataxia telangiectasia mutated (ATM), retinoblastoma protein (Rb), Ki67, and p53 and clinical outcome in invasive lymph node-negative bladder cancer.
PATIENTS AND METHODS
• Protein expression was measured by immunohistochemistry in cancer specimens from two independent cohorts of patients with bladder cancer treated with cystectomy (162 patients and 273) and one cohort of patients receiving radiotherapy (148).
• Disease-specific survival (DSS) was used as the outcome measure, and patients with no disease-specific death were followed for a minimum of 36 months.
• TIP60 was significantly correlated with DSS in both cystectomy cohorts (hazard ratio [HR] 0.42, 95% confidence interval [CI] 0.26–0.68, P < 0.001 and HR 0.45, 95% CI 0.28–0.72, P= 0.001).
• MRE11 was significantly correlated with DSS in the cohort receiving radiotherapy (HR 0.64, 95% CI 0.47–0.86, P= 0.005).
• P16 was significantly correlated with DSS in all three cohorts (HR 0.46, 95% CI 0.30–0.75, P= 0.032; HR 0.60, 95% CI 0.37–0.97, P= 0.032; HR 0.52, 95% CI 0.28–0.96, P= 0.001).
• Rb was significantly correlated with DSS in one cystectomy cohort (HR 1.71, 95% CI 1.13–2.75, P= 0.017).
• Ki67, p53, and pATM were not significantly correlated with DSS in any of the cohorts.
• TIP60 protein expression was a predictive marker for DSS after cystectomy in two independent cohorts. This novel marker was the strongest predictive factor in multivariate analysis in patients receiving cystectomy.
• MRE11 was shown to be a predictive marker for DSS after radiotherapy.
• We have shown that TIP60 and MRE11 hold the potential to guide patients with invasive bladder cancer to either cystectomy or radiotherapy. This study was based on retrospective material and consequently we suggest that these markers should be validated in a prospective study.
Standard treatment for patients with localised muscle-invasive bladder cancer (MIBC) is cystectomy. However, radiotherapy alone or combined with chemotherapy have achieved comparable results to cystectomy in some retrospective cohorts [1–4], and in the latest European Association of Urology guideline pT2N0M0 selected patients can be offered multimodality bladder-sparing therapy as an alternative . Furthermore, in a recent published randomised trial James et al. showed that neoadjuvant chemotherapy (5-fluorouracil and mitomycin C) combined with radiotherapy results in better local control of the cancer without increasing toxic side-effects. Treatment decisions by clinicians are guided by performance status, TNM classification, and request from the patient. However, there is no histopathological or molecular parameter for selecting the optimal treatment regimen (cystectomy or radiotherapy) that would benefit the individual patient.
The DNA-damage response (DDR) mediates either DNA repair or apoptosis upon the DNA damage created by radiation or chemotherapy. Double-strand breaks (DSBs) are the most extreme damage that ionizing radiation and chemotherapy can inflict in the cell. To initiate the DDR upon DSBs activation of the MRN (MRE11, Rad50 and Nijmegen breakage syndrome 1 [NBS1]) complex, TIP60 acetyltransferase, and ataxia telangiectasia mutated (ATM) kinase is crucial . Briefly, MRN complex binds around the DSB and TIP60 and ATM are recruited, then ATM is activated by phosphorylation of serine 1981 (pATM) leading to activation of p53, which then stops cell proliferation by activation of p16 and retinoblastoma protein (Rb) among others until the DNA is repaired. Failure in DNA repair may result in cell death by mitotic catastrophe, senescence or apoptosis. Defects in the DDR (e.g. in ATM and p53) have been shown to render cancer cells more sensitive to DSB induced by radiation and chemotherapy . In contrast, loss of cell-cycle regulators (e.g. Rb, p16, p53, p21) have been shown to be related to poor survival in patients receiving cystectomy [9,10]. However, the potential clinical impact of such findings is often hampered by insufficient validation of significant results in suitable cystectomy and radiotherapy cohorts.
In the present study we evaluated the predictive value of a series of cell cycle and DDR proteins (p16, ATM, TIP60, Rb, p53, Ki67 and MRE11) by immunohistochemistry (IHC) in patients receiving cystectomy and patients treated with radiotherapy combined with chemotherapy, respectively.
PATIENTS AND METHODS
This study included three different patient cohorts (Table 1). We studied two cohorts (A and B) of patients undergoing radical cystectomy for urothelial carcinoma, and one cohort (C) of patients undergoing radiotherapy. Only T1–4a N0M0 patients at time of treatment were included, and patients without disease-specific death had at least 36 months follow-up. Patients were censored after 96 months of disease-specific survival (DSS). Cohorts A (162 patients) and B (273) were operated at the Department of Urology, Aarhus University Hospital, Denmark between 1980 and 2003 (A), and between 1992 and 2008 (B). Cohort B was defined for a previous research project  and consisted of all consecutive patients with accessible cystectomy tissue from this period. No patients from cohort A were included in cohort B. No patients in cohort A or B received neoadjuvant or adjuvant chemotherapy. Cohort C (148 patients) was treated at Erlangen University Hospital, Germany, with >54 Gy radiation to the bladder (achieved in 85% of the patients, and all patients received >45 Gy) and 45 Gy to the lymph nodes. Concurrent chemotherapy was administered to 83% of the patients, and all regimens contained cisplatin (72%) or carboplatin (28%) as monotherapy (59%), combined cisplatin and carboplatin (5%), or in combination with 5-fluorouracil (35%) or gemcitabine (2%) [4,12]. In all, 116 patients (78%) had complete response locally at 3 months after radiotherapy. Due to lack of local control (20 patients) or bladder dysfunction (two), 15% (22) received cystectomy after radiation. Tumour stage was determined using the American Joint Committee on Cancer recommendations from 1997 for cohort A and from 2002 for Cohort B and C. In brief, tumours invading the deep muscle (outer half) were classified as T3a tumours in the 1997 edition and as T2b tumours in the 2002 edition. Lymph node stage was based on pathology reports after surgery in cohort A and B, while it was based on clinical evaluation and imaging in cohort C. The material was: formalin-fixed paraffin-embedded (FFPE) full-slide sections from preoperative transurethral resection of bladder (TURB) biopsies (cohort A); tissue microarray (TMA) with FFPE biopsies from cystectomy specimens (cohort B); TMA with FFPE biopsies from pretreatment TURB (cohort C).
Table 1. Demographics of cohorts A, B, and C and statistical analysis of parameters in relation to outcome (DSS)
*HRs from univariate and multivariate Cox regression analysis in relation to DSS for each cohort. Backwards eliminating regression models was used for multivariate analysis. Here we used all clinical parameters and markers, and the parameter with the highest P value was eliminated and the multivariate analysis was rerun. Elimination continued until all parameters had a P < 0.2. †Preop., T-stage after TURB or Final, T-stage after cystectomy. Significant findings are highlighted in bold.
For IHC, the staining procedure was based on the EnVision+TM System HRP (Dako) as previously described . Antibodies used: p16 (1:50 Pharmigen), Rb (1:200 Rb-1 DAKO), p53 (1:600 DO-7 DAKO M7001), TIP60 (1:8500 Novus Biologicals), ATM (1:300 Rockland pS1981), Ki67 (1:75 DAKO M7240), and MRE11 (1:150 Abcam plc). All staining procedures were conducted at the Department of Molecular Medicine, Aarhus, Denmark.
For scoring of IHC staining, TMA slides were scanned using Hamamatsu Nanozoomer scanner (Hamamatsu Corporation, Hamamatsu City, Japan) and scoring was performed manually using VIS visualization software (Visiopharm A/S, Hørsholm, Denmark). Full-slide sections were scored under the microscope. Thresholds used: p16 >10% was considered positive ; Rb was negative when no staining of carcinoma cell nuclei was observed (positive control staining of stromal cell nuclei) [15,16]; Ki67 >20% positive carcinoma cell nuclei was considered high [16,17]; p53 >20% was considered positive [9,16,17]; TIP60 and MRE11 were scored on a continuous scale and the lowest quartile considered ‘low’. pATM-positive carcinoma cell nuclei were scored on a continuous scale and the optimal threshold value defined by receiver operating characteristic curve. Scoring was performed by two observers independently ‘blinded’ to outcome. Scoring differences >10% on the continuous scale, and differences in the dichotomized scorings, were reviewed and consensus was reached. Inter-observer agreement was calculated using Cohen's κ (range 0.70–0.91).
Specificity of antibodies against MRE11 and TIP60 was tested by Western blotting (Fig. S1). The remaining antibodies have previously been documented by Western blotting [19–23].
Categorical data were compared in univariate analysis using the chi-squared test or Fisher's exact test and censored data was compared using the log-rank test. Hazard ratios (HRs) were estimated from Cox proportional hazard models. Backwards eliminating regression models were used for multivariate analysis. Here we used all clinical parameters and markers in multivariate analysis, and the parameter with the highest P value was eliminated and the multivariate analysis was re-run. Elimination continued until all parameters had a P value of <0.2. Harrell's concordance index was used to evaluate the predictive accuracy of the parameters.
We used IHC to study the expression of the DDR markers and cell cycle markers p16, ATM, TIP60, Rb, p53, Ki67 and MRE11 in cancers from three different patient cohorts (Table 1). Cohorts A and B differed in their relative distribution of T-stages, with low T-stage tumours being more frequent in B (Table 1). Before treatment, cohort C had significantly more T1 tumours compared with B (P= 0.001), and significantly less T2 tumours (P < 0.001). There was no significant difference in DSS between cohorts B and C (P= 0.18), while cohort A had significantly shorter DSS than cohorts B and C (P < 0.001 and P= 0.042; Fig. S2). Detailed description of the cohorts is listed in Table 1.
MARKER EXPRESSION AND CORRELATION TO CLINICAL PARAMETERS AND OUTCOME
Initially we compared marker expression with clinical parameters for the three patient cohorts (Table S1). We found that high TIP60 staining was significantly associated with high disease stage and lymphovascular invasion (LVI) in cohort C. High MRE11 staining was significantly associated with high disease stage in cohort B and finally, high pATM staining was significantly associated with disease stage in cohorts B and C, and with carcinoma in situ (CIS) in cohort B.
We then analysed expression of p16, TIP60, pATM, Rb, p53 and Ki67 in cohort A according to outcome (Fig. 1). Patients with cancers staining positive for p16 (P= 0.001; Fig. 2A) and with high TIP60 expression (P < 0.001; Fig. 2D) had significantly longer DSS. Expression of Rb, pATM, p53 and Ki67 was not significantly correlated with DSS (Table 1). To support our initial findings, we then studied the expression of p16, TIP60, Rb and pATM in cohort B. Again, we found that patients with cancers staining positive for p16 (P= 0.032; Fig. 2B) and with high TIP60 expression (P= 0.002; Fig. 2E) had significantly longer DSS. Furthermore, patients with cancers staining positive for Rb (P= 0.017; Fig. 2G) had significantly shorter DSS. We then investigated if the markers had predictive value in bladder cancer patients receiving radiotherapy (cohort C). Here, only patients with cancers staining positive for p16 had significantly longer DSS (P= 0.032; Fig. 2C). A summary of the results is presented in Table 2. Using Backwards eliminating regression models for multivariate Cox regression analysis, we found TIP60 and p16 to be significantly correlated to DSS in cohorts A (P= 0.004 and P= 0.009) and B (P= 0.009 and P= 0.001). In contrast, none of these markers were significant in multivariate analysis in cohort C (Table 2). Using Harrell's concordance index we found that TIP60 increased the predictive accuracy (PA) of the available clinical parameters (stage, grade, CIS, LVI, and age) from 52.2% to 61.8% in cohort A and from 62.2% to 65.5% in B (Table 3).
Table 2. Univariate and multivariate Cox regression analysis of disease specific survival as function of molecular markers
HRs from univariate and multivariate Cox regression analysis in relation to DSS for each cohort. Backwards eliminating regression models was used for multivariate analysis. Here we used all clinical parameters and markers, and the parameter with the highest P value was eliminated and the multivariate analysis was rerun. Elimination continued until all parameters had a P < 0.2. Significant findings are highlighted in bold.
TIP60 and p16
TIP60 and MRE11
p16 and MRE11
Table 3. PA of clinical parameters alone and combined with molecular markers based on Harrell's concordance index
High expression of MRE11 in cancer cells has previously been reported to be associated with good prognosis after radiotherapy . In patients undergoing radiotherapy, high expression of MRE11 was significantly associated with long DSS (P= 0.005; Fig. 2I), but we observed no predictive value of this marker in the cystectomy cohort (Fig. 2H). Furthermore, in multivariate analysis, MRE11 was significantly associated with long DSS in cohort C (P < 0.001; Table 2) and increased PA from 66.7% to 73.2%. MRE11 expression was not evaluated in cohort A due to a lack of tissue sections.
We then compared patients from cohorts B and C and found that patients receiving cystectomy had significantly better DSS (P= 0.008, HR 0.54, 95% CI 0.35–0.85; Fig. 3A) relative to patients receiving radiotherapy, when the analysis was restricted to patients with high expression of TIP60. In a similar way, cystectomy resulted in a significantly better DSS (P= 0.009, HR 0.42, 95% CI 0.23–0.79; Fig. 3B) compared with treatment by radiotherapy in patients with low expression of MRE11. There were no differences between the two treatment groups in patients having tumours with either high MRE11 or low TIP60.
COMBINATION OF MARKERS
Stratification based on combinations of TIP60 and MRE11 staining resulted in significant differences in DSS between cohorts B and C (Fig. 3 C+D). Cystectomy resulted in better DSS compared with radiotherapy (P= 0.01, HR 0.30, 95% CI 0.12–0.76; Fig. 3E) in patients with cancers showing low expression of MRE11 and high expression of TIP60. In contrast, cystectomy resulted in worse DSS compared with radiotherapy (P= 0.012, HR 3.58, 95% CI 1.25–10.27; Fig. 3F) in patients with cancers showing high expression of MRE11 and low expression of TIP60. There was no difference between cystectomy and radiotherapy in patients with cancers showing simultaneously high or low expression of both MRE11 and TIP60. Stratification based on p16 and TIP60 expression resulted in significant differences in DSS and increased PA in both cystectomy cohorts A and B (P < 0.001, PA 64.8% and P= 0.003, PA 69.0%; Fig. S1 B+C and Table 3) but had no effect in cohort C. The HRs between both markers being high, and both being low, were 0.25 (95% CI 0.13–0.49) and 0.21 (95% CI 0.09–0.47), respectively. Stratification based on p16 and MRE11 expression resulted in significant differences in DSS and PA in cohorts B and C (P= 0.039, PA 63.7% and P= 0.003, PA 76.4%; Fig. S1 E+F).
Finally, we simulated the potential clinical value of the results. If treatment of all the patients was guided by TIP60 and MRE11 expression levels, then all patients with cancers having low expression of MRE11 and high expression of TIP60 received cystectomy, and all patients with cancers having high expression of MRE11 and low expression of TIP60 received radiotherapy. In this scenario, 12% of the patients originally receiving cystectomy would benefit from receiving radiotherapy instead, and 18% of the patients originally receiving radiotherapy would benefit from cystectomy. The remaining majority of patients receiving their original, and biomarker correct treatment, would have no change in 5-year DSS.
To guide optimal treatment decisions involving radiotherapy and surgery, assuming that the patients can tolerate both options, additional individualised information is needed. In the present study, we asked whether DDR protein markers and cell cycle regulatory protein markers could serve as predictive biomarkers for outcome in these two treatment settings. We found that p16 was predictive of outcome after both treatments and increased PA in combination with the clinical parameters. Furthermore, we successfully showed and validated TIP60 to increase PA and to be predictive of DSS in two independent cohorts of cystectomised patients but not in patients receiving radiotherapy. Finally, we confirmed that high MRE11 expression was associated with better outcome and increased PA in patients with bladder cancer receiving radiotherapy in accordance with a previous report by Choudhury et al.. This result is reinforced by the fact that we used an identical methodology (antibody, staining protocol, and scoring system) to the one used by Choudhury et al. We therefore think that the present findings strongly support the hypothesis that MRE11 is associated with response to radiotherapy. We found no correlation between pATM and survival in either of the cohorts, suggesting that even though it is positioned in the same complex as TIP60 and MRE11, it has no predictive property, as previously shown for the non-phosphorylated form .
Each DDR marker has many different targets and the entire network is only partially understood [7,24]. Here, we showed that MRE11 and TIP60 belonging to the same response pathway have opposite predictive significance in patients treated with cystectomy or radiotherapy. MRE11 predicted only radiotherapy response and TIP60 predicted only outcome after cystectomy. We then combined the patients in cohorts B and C by expression of either TIP60 or MRE11, and identified subgroups favouring cystectomy. However, when we combined the two markers, we could significantly identify two of the four possible subgroups, where the patients would benefit significantly from receiving either cystectomy or radiotherapy. A bias may be associated with the selection of patients for cohorts B and C, because there were no patients without cancer in the bladder at cystectomy represented in cohort B. However, when we compared the demographics of the two groups they appeared similar and comparable.
In the present study, we found that p16 was a useful predictive marker, and regardless of treatment, low p16-expression correlated with a poor prognosis. We are not the first to show this correlation , but importantly, we validated our findings in two independent cohorts, using previously published IHC scoring threshold values . When we used p16 in combination with either TIP60 or MRE11 we increased the HRs and the PAs between the top and bottom subgroups but decreased the number of patients (Fig. S1).
In the present study, expression of Rb had a significant effect on patient survival after cystectomy in one of the cystectomy cohorts but not in the radiotherapy cohort. This is inconsistent with previous findings [9,15,16], where Rb was found to be a predictive marker for outcome after radiotherapy. A reason for this could be that we used another antibody and staining protocol, or that the tissues were handled using different fixation protocols. In the present study, expression of p53 and Ki67 had no significant effect on patient survival after cystectomy. This finding may not be surprising considering the substantial controversy that exists with respect to the predictive value of p53 and Ki67 [9,17,18,25,26]. Recently, Stadler et al. investigated the use of p53 expression in bladder cancers in a randomised prospective trial and found no significant association to outcome.
The cohorts have been retrospectively collected and due to our inclusion criteria, a selection bias may have been introduced. Cohort A had significantly worse DSS than B and C, probably due to a higher T-stage and more unidentified lymph node metastases in cohort A, as the operational technique changed to include extended lymph node dissection in January 2004. Furthermore, we obtained low PA values for clinical risk factors alone (52% and 62%) probably as tumours from patients with positive lymph nodes were excluded and other risk factors (CIS, LVI) were missing for some patients cohorts. It is very likely that inclusion of LVI information for cohort B would probably increase the PA to a similar level as for cohort C.
We used the same approach for the novel markers as with MRE11. The exact threshold differs between the cohorts, and this variation in measures of marker expression may be due to differences in pre-analytic factors (e.g. length and type of fixation or embedding temperatures) . We found different threshold values for TIP60 among patients operated at the same hospital in the same period. However, the tissue came from TURB specimens in cohort A and from cystectomy specimens in cohort B. Differences in tissue handling between the different procedures, differences in time from section-cutting to staining, and differences between staining dates may explain this.
Currently, no biomarkers are used in daily clinical practice, as most published markers cannot predict outcome, improve treatment decisions, or monitor therapy response in MIBC [5,27]. The markers reported here for therapy selection, TIP60 and MRE11, offer an ideal combination for this purpose, and should be tested in a prospective setting with standardised procedures to verify their power in therapy outcome prediction. First, in a cystectomy and a chemo-radiation cohort, and then in an intervention study, if the first confirms the promising data reported here.
In conclusion, we have shown that TIP60 and MRE11 hold the potential to guide patients with MIBC to either cystectomy or radiotherapy. The present study was based on retrospective material and consequently we suggest that these markers should be validated in a prospective study.
The study was supported by The John and Birthe Meyer Foundation, the Danish Cancer Society, the Ministry of Technology and Science, and the Lundbeck Foundation. Furthermore, the research leading to these results has received funding from the European Community's Seventh Framework programme FP7/2007-2011 under grant agreement n° 201663.rh