Clinical significance of stromal apoptosis in colorectal cancer

Background: Epithelial and stromal cells play an important role in the development of colorectal cancer (CRC). We aimed to determine the prognostic significance of both epithelial and stromal cell apoptosis in CRC. Methods: Total apoptosis was determined by caspase-3 activity measurements in protein homogenates of CRC specimens and adjacent normal mucosa of 211 CRC patients. Epithelial apoptosis was determined by an ELISA specific for a caspase-3-degraded cytokeratin 18 product, the M30 antigen. Stromal apoptosis was determined from the ratio between total and epithelial apoptosis. Results: Epithelial and stromal apoptosis, as well as total apoptosis, were significantly higher in CRC compared with corresponding adjacent normal mucosa. Low total tumour apoptosis (⩽median caspase-3 activity) was associated with a significantly worse disease recurrence (hazard ratio (HR), 95% confidence interval (95% CI): 1.77 (1.05–3.01)), independent of clinocopathological parameters. Epithelial apoptosis was not associated with clinical outcome. In contrast, low stromal apoptosis (⩽median caspase-3/M30) was found to be an independent prognostic factor for overall survival, disease-free survival and disease recurrence, with HRs (95% CI) of 1.66 (1.17–2.35), 1.62 (1.15–2.29) and 1.69 (1.01–2.85), respectively. Interpretation: Stromal apoptosis, in contrast to epithelial apoptosis, is an important factor with respect to survival and disease-recurrence in CRC.

The classical normal mucosa -adenoma -carcinoma sequence of colorectal cancer (CRC) is associated with a resistance to apoptosis (Koornstra et al, 2003). The use of apoptosis indices or measurements as a prognostic or predictive factor for CRC has been limited by the reliability of the detection techniques and time consuming evaluations, and have shown conflicting results (de Bruin and Medema, 2008). Caspases, or death proteases, are cysteine proteases that are responsible for the morphological changes within cells during apoptosis (Kumar, 2007). Caspase-3 is at the point of convergence of the two main apoptotic pathways and cleaves most of the cellular substrates in the apoptotic process (Porter and Janicke, 1999). Measurement of caspase-3 activity is therefore a major and the most reliable determinant of apoptosis (Kaufmann et al, 2008). Caspase-3-degraded proteins are also used to detect apoptosis, for instance by the M30 antibody that detects an epithelial-specific caspase-3-degraded product of cytokeratin 18 (CK18), that is, CK18-Asp396, and therefore specifically detects apoptotic epithelial cells (Leers et al, 1999).
In recent years it has become clear that stromal cells within the tumour, and their interactions with the tumour cells, are important in the initiation and progression of cancer (Mueller and Fusenig, 2004), illustrated by the prognostic relevance of tumour stromalepithelial ratios for CRC patients (Mesker et al, 2007). Apoptosis of stromal cells might therefore also be of prognostic importance for the patients. To investigate the prognostic impact of stromal apoptosis, we determined total cellular apoptosis by caspase-3 activity measurement, and epithelial apoptosis by an ELISA for caspase-3-degraded CK18 (M30 antigen), in a large series of CRC specimens. Correction of the total apoptosis for the epithelial apoptosis (caspase-3/M30) revealed the apoptosis of non-epithelial cells, that is, stromal apoptosis, which was found to be of major clinical relevance in the CRC patients.

Patients
The study population consisted of 211 CRC patients that were admitted to the Leiden University Medical Center for tumour resection between December 1983 and September 1991, as previously described (Langers et al, 2008). Fresh tissue was collected from the non-irradiated surgical specimen immediately after resection and attention was paid to collect material from the non-necrotic mid-central part of the tumour. Normal mucosa samples were obtained whenever possible at a distance of approximately 10 cm from the tumour. Tissue samples were frozen and stored at À701C until use. Clinical data and follow-up information were available for a period of at least 5 years after surgical intervention. Macroscopic (diameter and localisation of the tumour) as well as microscopic data were assessed, including Dukes' stage and differentiation (and ulceration) grade of the tumour according to the WHO classification. Colonic cancers were classified as being proximal or distal to, and including, the splenic flexure. The study was performed according to the guidelines of the Medical Ethics Committee of the Leiden University Medical Center and in compliance with the Helsinki Declaration.

Tissue homogenisation and protein determination
Frozen tissue specimens were weighed and homogenised on ice for 2 min in 1 ml Tris-HCl, 0.1% Tween 80, pH 7.5 per 60 mg tissue using a Potter device (B Braun, Germany), and centrifuged twice at 8000 g for 2.5 min at 41C, as described before (Langers et al, 2008). Protein content was measured according to Lowry et al. (1951) and standardised by bovine serum albumin (BSA).

Caspase-3 activity measurement
For measurement of caspase-3 enzymatic activity, a colorimetric assay was used as described before . Samples were incubated with a saturating concentration of 25 mM specific enzyme substrate Ac-Aps-Glu-Val-Asp-AMC (Ac-DEVD-AMC; Bachem, Heidelberg, Germany) in a total of 100 ml 100 mM HEPES buffer, pH 7.25, containing 10% (w/v) sucrose, 0.1% (v/v) Nonidet-P40 and 10 mM dithiothreitol. During incubation at 371C, fluorescent AMC release by active caspase-3 was monitored at an excitation of 360 nm and an emission of 460 nm using a Fluostar Optima plate reader (BMG Labtech GMBH, Offenburg, Germany). Calibration curves were constructed using free AMC. Caspase-3 activity was indicated in pmol AMC min À1 mg À1 protein. The intra-and interassay variabilities of the caspase-3 activity assay were between 5 and 15%.

M30 and CK18 antigen level detection
M30 antigen and total CK18 antigen levels were detected by M30 and M65 ELISAs, according to the manufacturer's protocol (Peviva BV, Bromma, Sweden), and expressed as antigen units per mg protein (U mg À1 ). Both ELISAs had intra-and interassay variabilities o10%, in line with the manufacturer's specifications.

Immunohistochemistry for active caspase-3, M30 antigen and cytokeratin
Paraffin tissue sections (5 mm) were deparaffinised, blocked in 0.3% H 2 O 2 in methanol for 30 min at RT and rehydrated through graded ethanol to phosphate-buffered saline (PBS). Antigen retrieval was performed by boiling the slides in 10 mM citrate buffer (pH 6.0) for 10 min in a microwave oven. After cooling down to RT and rinsing with PBS for three times the slides were blocked with 5% goat serum (Dako) in 1% BSA in PBS for 20 min at RT. Afterwards the sections were incubated with the primary antibody in 1% BSA/PBS overnight at 41C: mouse-anti-pancytokeratin (1 : 5000, clone C11; Santa Cruz Biotechnologies, Santa Cruz, CA, USA), mouse M30 Cytodeath antibody (1 : 400; Roche Applied Science, Penzberg, Germany) and rabbit-anti-activecaspase-3 (1 : 200; Cell Signaling Technology). After washing with PBS, the slides were incubated with biotinylated goat-anti-mouse or goat-anti-rabbit secondary antibodies (both 1 : 200; Dako) in 1% BSA/PBS for 45 min. After thorough washing with PBS, the slides were incubated with streptavidin-avidin -biotin complex/HRP (Dako) for 45 min at RT. The signal was developed with 0.015% H 2 O 2 (Merck, Darmstadt, Germany) in 0.05% diaminobenzidinetetrahydrochloride (Sigma, Zwijndrecht, The Netherlands) in 0.01 M Tris-HCL pH 7.6 for 10 min resulting in a brown staining product. Sections were counterstained with Mayers' haematoxylin (Merck), dehydrated and mounted in entallan (Merck). Slides without primary antibody incubations were included as negative controls. Photomicrographs of representative sections were taken with a Nikon Elipse E800 microscope equipped with a Nikon DXM 200 camera.

Statistical analysis
Statistical analysis was performed with Statistical Package for Social Sciences (SPSS) statistical software (version 12.0 for Windows; SPSS Inc., Chicago, IL, USA). Wilcoxon signed rank and Spearman's rho tests were used to compare paired observations that did not follow a normal distribution. Kruskall -Wallis and Mann -Whitney U-tests were used to compare two or more groups that do not follow a normal distribution. The entry data for survival analysis were the time of surgery for the primary tumour. Biomarker expressions were dichotomised and overall survival, disease recurrence and disease-free survival were evaluated using Kaplan -Meier methodology with death, local and distant recurrence, or both, as events, including log-rank tests. Univariate Cox proportional hazard models were used to explore the association of markers with clinical outcome. Variables with a Pp0.1 in the univariate analyses were subjected to multivariate analyses. Statistical tests were two-sided and Po0.05 was considered statistically significant. Graphs were made with Graphpad Prism (version 4.0; Graphpad Prism Inc., La Jolla, CA, USA) software.

Total, epithelial and stromal apoptosis in colorectal cancer
The caspase-3 activity, M30 antigen and CK18 levels were determined in colorectal carcinoma tissue of all 211 patients, whereas due to tissue/homogenate availability these parameters could only be determined in adjacent normal tissue from 177 patients. M30 antigen, CK18 and caspase-3 activity levels in the colorectal tumour tissue samples were significantly higher than those in the corresponding normal tissue samples in these 177 patients (Table 1). Immunoblot analysis confirmed that tumours with a high caspase-3 activity had high protein levels of active caspase-3 compared with low caspase-3 activity tumours ( Figure 1A). Overall there was a weak but significant correlation between tumour caspase-3 activity and tumour M30 antigen levels, as expected, because the M30 antigen is generated by active caspase-3 ( Figure 1B). Some tumours had really high caspase-3 activity levels, without high M30 antigen levels, indicating high stromal apoptosis.

Low High
Pro-caspase-3 Caspase-3 activity Immunohistochemical analysis confirmed that in these tumours active caspase-3 was indeed mainly expressed by diverse nonepithelial cells and were almost M30 antigen-negative ( Figure 1C, represented by Tumour 1), compared with other tumours in which epithelial cells were expressing most of the active caspase-3 and M30 antigen-positive (represented by tumour 2). Some high power magnifications of tumour 1 to illustrate the diverse nature of stromal apoptosis are shown in Supplementary Figure 1. Our measurements enabled us to further differentiate in cell type origin of apoptosis by calculating the percentage of apoptotic epithelial cells through the formula M30/CK18 Â 100 and for stromal apoptosis through the formula caspase-3/M30. This epithelial as well as stromal apoptosis was also significantly higher in tumour tissue compared with corresponding normal tissue (Table 1).

Relation between total, epithelial and stromal apoptosis and clinical variables
The association between the clinicopathological characteristics of the 211 colorectal cancer patients and the tissue caspase-3 activity, as a measure of total apoptosis, is summarised in Table 2. Caspase-3 activity level in the tumours correlated with gender, tumour location and Dukes' stage, that is, significantly higher in females (P ¼ 0.006) and in tumours located in the proximal colon (P ¼ 0.02), and decreased with increasing Dukes' stage (P ¼ 0.01). Caspase-3 activity in normal mucosa was also found to be higher in females (P ¼ 0.04) and in the proximal colon (P ¼ 0.04), but did not correlate with Dukes' stage. Tumour M30 antigen levels correlated with grade of differentiation, with well-differentiated tumours having significantly higher levels (Table 3; P ¼ 0.02). The total amount of epithelial cells (CK18 per mg protein) in the tumour samples correlated with the tumour diameter, larger tumours having significantly lower amounts of epithelial cells compared with smaller tumours (P ¼ 0.007). Epithelial cell apoptosis (M30/CK18 Â 100) was significantly higher in tumours with ulceration (P ¼ 0.02). Stromal apoptosis (caspase-3/M30 ratio) was related to gender and tumour location, that is, significantly higher in females (P ¼ 0.02) and proximal colon tumours (Po0.001).
Stromal apoptosis predicts overall survival, disease-free survival and disease recurrence High caspase-3 activity (4median) in the tumour was associated with better overall patient survival; median overall survival: 58 vs 34 months (Figure 2A). These patients with high tumour caspase-3 activity (4median) also had a significantly better disease-free survival with a median disease-free survival of 47.5 vs 27 months for patients with low tumour caspase-3 activity (pmedian; Figure 2B). Low tumour caspase-3 activity (pmedian) was also accompanied by a shorter time to recurrence compared to patients with a high tumour caspase-3 activity ( Figure 2C), with 5-year recurrence rates of 46.1% and 30.3%, respectively. Caspase-3 activity in the normal adjacent mucosa was not associated with overall survival, disease-free survival or disease recurrence of the patients (not shown). Tumour M30 antigen, CK18 or M30/CK18 Â 100 levels did not show any relation with disease-free survival ( Figure 2D, E and F), overall survival or disease recurrence (not shown). In contrast, the caspase-3/M30 ratio of the tumours correlated very strongly with the disease-free survival of the patients ( Figure 2G), with a high ratio (4median) having a significantly better median disease-free survival of 45 vs 26.5 months for the patients with tumours with a low ratio.
To underline the importance of stromal apoptosis to the clinical outcome of the CRC patients, we divided our patient population in four groups based on median tumour M30 antigen and median tumour caspase-3 activity levels: (1) low caspase-3/M30 low (n ¼ 67), (2) low caspase-3/M30 high (n ¼ 39), (3) caspase-3 high/M30 high (n ¼ 68) and (4) caspase-3 high/M30 low (n ¼ 37). The M30 antigen levels did not have an additive effect in the two groups with low caspase-3 activity, and therefore groups 1 and 2 were pooled. In the groups with high caspase-3 activity, the disease-free survival was found to be the best for patients with a high tumour caspase-3 activity and low M30 antigen level ( Figure 2H; group 4 vs groups 1 and 2, log-rank value: 5.68, P ¼ 0.02), which strengthened our notion that stromal apoptosis in the tumour is an important prognostic factor for the patients' disease-free survival.
The estimates of relative risk of the evaluated parameters to disease outcome, as calculated by hazard ratios in the univariate Cox regression analysis, are shown in Table 4. The clinicopathological factors that were significantly associated with worse outcome were found to be old age and advanced Dukes' stage, as expected. With regard to the apoptosis parameters, a low total (caspase-3 activity) or low stromal (caspase-3/M30) apoptosis were consistently found to be associated with a worse overall and disease-free survival as well as with a higher risk of recurrence. Adjustment of the apoptosis parameters to the most relevant clinicopathological parameters, that is, gender, age, Dukes' stage, tumour localisation and tumour diameter in the multivariate Cox proportional hazards analysis, identified stromal apoptosis as an independent prognostic factor for overall survival, disease-free survival and disease recurrence. In contrast, total apoptosis was found to be an independent prognostic factor for disease recurrence only (Table 4).  (Ishiguro et al, 2006). Important signalling pathways that are frequently disrupted in CRC, such as the transforming growth factor (TGF)-b (Naber et al, 2008), bone morphogenic protein (BMP; Hardwick et al, 2008) and Wnt pathway (Macheda and Stacker, 2008), play a role in both the cancer-associated stromal and epithelial compartment in the development of CRC. The presence of fibroblasts , myofibroblasts  and endothelial cells (Baeten et al, 2006) in the cancer-associated stroma contributes to progression of CRC and a worse patient prognosis. Also, the expression of other proteins in the cancer-associated stroma, that is, hypoxia regulated proteins (Cleven et al, 2007), matrix degrading proteases (Hilska et al, 2007) and cyclooxygenase-2 (Sonoshita et al, 2002;Adegboyega et al, 2004), is correlated to CRC progression and disease prognosis. The importance of the cancer-associated stroma in CRC is most simply illustrated by the reported prognostic significance of the tumour-stroma ratio in CRC (Mesker et al, 2007). The cancer-associated stromal cells, and their interaction with the epithelial tumour cells, are therefore speculated to be important targets for future therapy (Mueller and Fusenig, 2004). Programmed cell death, that is, apoptosis, is an important mechanism to maintain tissue homoeostasis. Defects in the two main apoptotic pathways, the extrinsic and intrinsic pathways, lead to a resistance to apoptosis and increased cell survival. Epithelial apoptosis in CRC has been widely studied but still the significance for the prognosis of CRC patients, as well as the significance for the response to treatment such as chemotherapy, is unclear (Koornstra et al, 2003;de Bruin and Medema, 2008). As apoptosis in the cancer-associated stroma could also play an important role, and its induction could be relevant for treatment response, we investigated the contribution of stromal apoptosis, as well as epithelial apoptosis in the clinical outcome of CRC. We showed that both epithelial and stromal apoptosis, as well as total apoptosis, were significantly higher in non-irradiated CRC than in normal adjacent mucosa, as found in other studies (Leonardos et al, 1999;Jonges et al, 2001;Koornstra et al, 2003;de Heer et al, 2007;de Oca et al, 2008). As cancer cells usually have apoptotic defects and are more difficult to drive into apoptosis (Igney and Krammer, 2002;Brown and Attardi, 2005), higher levels of apoptosis in cancer, compared with normal tissue, might seem to be somewhat unexpected. However, the level of apoptosis and sensitivity to apoptotic signals are two, although related, different things. Importantly, in this study, total CRC apoptosis decreased with increasing Dukes' stage, whereas no such correlation was found in adjacent normal mucosa, which indeed indicates that cells in the tumour become resistant to apoptotic signals during tumour progression resulting in a decreased caspase-3 activity in more advanced tumour stages. Moreover, total apoptosis in the tumour showed a strong association with risk of recurrence and diseasefree survival. Cox multivariate hazards analysis identified low caspase-3 activity in the tumour as an independent risk factor of recurrence, with patients with tumour caspase-3 activity below the median value having a 1.77 times higher relative risk of recurrence, similar as reported by others for both colon and rectal cancer de Oca et al, 2008). Caspase-3 activity corrected for the percentage of epithelial cells, as determined on slides, showed similar results compared with the uncorrected caspase-3 activity in rectal cancer in the study by de Heer et al (2007), and therefore the authors concluded that caspase-3 activity is a prognostic factor for local recurrence in rectal cancer without the previous knowledge of epithelial-stromal ratios in the tumour. Correction of caspase-3 activity for the amount of epithelial cells (measured by CK18 antigen levels), caspase-3/CK18, in this study indeed showed similar results with respect to clinicopathological parameters and clinical outcome, as caspase-3 activity alone (not shown). However, this does not indicate the source of the caspase-3 activity; it could be that most of the caspase-3 activity is produced by apoptotic epithelial cells, and the apoptotic stromal cells could also be an important source. To investigate the source we determined M30 antigen levels by ELISA in the same CRC homogenates, as a measurement of epithelial caspase-3 activity/ apoptosis (Leers et al, 1999;Hagg et al, 2002). If apoptotic epithelial cells were mainly responsible for the caspase-3 activity in this study, M30 antigen levels should have shown a similar correlation with patient and tumour characteristics, which was clearly not the case. Our measurement of epithelial apoptosis, determined by M30/CK18 Â 100 (%), with median values of 3.72% for tumour tissue and 1.83% for normal tissue, is very much in line with epithelial apoptosis found in other studies (Koornstra et al, 2003). M30 antigen levels, either corrected for CK18 or uncorrected, did not correlate with overall survival, disease-free survival or disease recurrence, in contrast to caspase-3 activity, indicating the importance of non-epithelial caspase-3 activity, that is, stromal apoptosis. This was confirmed by determining caspase-3/M30 ratios, as stromal apoptosis levels, which even showed a stronger association with overall survival, disease-free survival or disease recurrence. Low stromal apoptosis (pmedian caspase-3/ M30) was an independent prognostic factor for overall survival, disease-free survival and disease recurrence with an HR (95% CI) of 1.66 (1.17 -2.35), 1.62 (1.15 -2.29) and 1.69 (1.01 -2.85), respectively.
Epithelial tumour cell apoptosis has been extensively studied in CRC, but still the relevance of the apoptotic index as a prognostic factor is a matter of debate (Koornstra et al, 2003;de Bruin and Medema, 2008). The detection technique is likely to influence the apoptotic levels, as some detection techniques are really specific for epithelial apoptotic cells, that is, M30 immunohistochemistry and M30 ELISA (Leers et al, 1999;Hagg et al, 2002), and others are (more or less) specific for apoptotic cells in general, that is, TdTmediated dUTP-biotin nick end-labeling (TUNEL) (Stadelmann and Lassmann, 2000;Barrett et al, 2001), active caspase-3 immunohistochemistry or caspase-3 activity measurements (Kaufmann et al, 2008), and also detect apoptotic stromal cells. The results presented here, that is, stromal apoptosis as an important prognostic factor for CRC, suggests that discrepancies found in the literature might be explained by the different detection techniques that are used, that is, 'specific for epithelial apoptosis' vs 'specific for apoptosis in general'.
In conclusion, this study confirms that caspase-3 activity within tumour tissue is an important denominator of disease recurrence and patient survival in CRC, with high levels of caspase-3 associated with good outcome. Remarkably, this caspase-3 activity was found to predominantly reside within the non-epithelial compartment of the tumours. These observations underline the importance of stromal cell apoptosis in CRC progression and identify the cancer-associated stroma as a potential therapeutic target.