Sep 14, 2011 - correlate with disease course and inflammatory properties ... metalloproteinase; MMR: Mismatch repair; MSH2: MutS homolog 2; MSI-H: high level microsatellite ... deficient male mice are more susceptible to developing car-.
IJC International Journal of Cancer
Serum MMP-8 levels increase in colorectal cancer and correlate with disease course and inflammatory properties of primary tumors Juha P. Va¨yrynen1, Juha Vornanen1, Taina Tervahartiala2, Timo Sorsa2, Risto Bloigu3, Tuula Salo4, Anne Tuomisto1 and Markus J. Ma¨kinen1 1
Department of Pathology, Institute of Diagnostics, University of Oulu, Oulu, Finland Department of Oral and Maxillofacial Diseases, Institute of Dentistry, University of Helsinki and Helsinki University Central Hospital, Helsinki, Finland 3 Medical Informatics Group, University of Oulu, Oulu, Finland 4 Department of Diagnostics and Oral Medicine, Institute of Dentistry, University of Oulu and Oulu University Hospital, Oulu, Finland
Matrix metalloproteinases (MMPs) form a family of zinc-dependent endoproteases participating in cancer pathogenesis by promoting invasion and regulating growth signaling, apoptosis, angiogenesis and immune responses. MMP-8 is an intriguing MMP with recently discovered antitumor activity and immunoregulatory properties, but its role in colorectal cancer (CRC) has not been studied extensively. Preoperative serum MMP-8 levels (S-MMP-8) of 148 CRC patients and 83 healthy controls were measured using an immunofluorometric assay and related to clinical and pathological parameters. The patients had higher S-MMP-8 than the controls (median 63.0 vs. 17.2 ng/ml, p 5 1.5E 2 9), and a receiver operating characteristics analysis yielded an area under the curve of 0.751 in differentiating the groups. In univariate analyses, S-MMP-8 correlated positively with disease stage (p 5 4.5E 2 4), the degree of primary tumor necrosis (p 5 0.0024) and blood neutrophil count (Pearson r 5 0.523, p 5 2.5E 2 9). Particular interest was also addressed to the inflammatory properties of the tumors, and both variables studied, peritumoral tumor-destructing inflammatory infiltrate and Crohn’s-like lymphoid reaction (CLR), showed a negative correlation with S-MMP-8 (p 5 0.041 and p 5 0.0057, respectively). In a multiple linear regression analysis, high S-MMP-8 associated with elevated blood neutrophil count, distant metastases, low-grade CLR and low body mass index. Overall, our results indicate that MMP-8 is involved in the course and progression of CRC influencing the immune response against the tumor and contributing to the resolution of necrosis. Serum or plasma MMP-8 may prove to be a worthy biomarker for CRC.
Matrix metalloproteinases (MMPs) are a family of structurally related but genetically distinct zinc-dependent endoproteases participating in extracellular matrix (ECM) degradation. MMP activities are strictly regulated at different levels: gene expression, zymogen activation, interaction with ECM and inhibition by endogenous regulators, most notably tissue inhibitors of metalloproteinases (TIMPs).1 MMPs have traditionally been
considered to be involved in cancer pathogenesis by promoting invasion and metastasis.2–4 However, recently, they have been implicated in other hallmarks of cancer, as well, by regulating growth signaling, apoptosis, angiogenesis and immune responses.3,4 In general, increased expression of certain MMPs can be observed in most types of human cancer, and this overexpression is often associated with poor survival.3
Key words: colorectal cancer, matrix metalloproteinase-8, biomarker, cancer immunology Abbreviations: AUC: area under the curve; BMI: body mass index; CEA: carcinoembryonic antigen; CI: confidence interval; CLR: Crohn’s like lymphoid reaction; CRC: colorectal cancer; DAB: diaminobenzidine; ECM: extracellular matrix; ELISA: enzyme-linked immunosorbent assay; IFMA: immunofluorometric assay; IHC: immunohistochemistry; IQR: interquartile range; MLH1: MutL homolog 1; MMP: matrix metalloproteinase; MMR: Mismatch repair; MSH2: MutS homolog 2; MSI-H: high level microsatellite instability; NG: necrosis grade; Pearson r: Pearson correlation coefficient; qRT-PCR: real-time quantitative polymerase chain reaction; ROC: receiver operating characteristics; RT/ CRT: radiotherapy or chemoradiotherapy; S-MMP-8: serum level of matrix metalloproteinase-8; S-TIMP-1: serum level of tissue inhibitor of metalloproteinases-1; SD: standard deviation; TIMP: tissue inhibitor of metalloproteinases. Additional Supporting Information may be found in the online version of this article. Grant sponsors: Academy of Finland, Emil Aaltonen Foundation, Finnish Cancer Foundation, Northern Finland Cancer Foundation, and the Research Foundation of Helsinki University Central Hospital DOI: 10.1002/ijc.26435 History: Received 25 Feb 2011; Accepted 8 Sep 2011; Online 14 Sep 2011 Correspondence to: Markus J. Ma¨kinen, Department of Pathology, Institute of Diagnostics, University of Oulu, Aapistie 5, POB 5000, FIN-90014, Oulu, Finland, Fax: þ358-8-537-5953, E-mail: markus.makinen@oulu.fi
C 2011 UICC Int. J. Cancer: 131, E463–E474 (2012) V
Early Detection and Diagnosis
2
Serum MMP-8 in CRC
Early Detection and Diagnosis
E464
Colorectal cancer (CRC) is one of the most common malignancies and leading causes of cancer deaths in the industrialized world.5 There is abundant evidence linking MMPs and TIMPs to the multistep process of colorectal carcinogenesis. Using real-time quantitative polymerase chain reaction (qRT-PCR) and immunohistochemistry (IHC), it has been shown that CRCs and normal colorectal mucosa display significant differences in the expression of certain MMPs including MMP-1–3, -7 and -9–14.6–8 Peripheral blood MMP levels are thought to reflect local MMP concentrations in the tumor microenvironment. Accordingly, higher levels of MMP-99,10 and TIMP-110–12 have been reported in serum or plasma samples of CRC patients compared to healthy controls. Furthermore, higher serum or plasma MMP-1,13 MMP-914 and TIMP-115–17 have also been associated with worse prognosis. Consequently, these peripheral blood MMPs and TIMPs are considered potential biomarkers in CRC. Despite being regarded as one of the most interesting MMPs, the role of MMP-8 in CRC has not been studied extensively. Also known as collagenase-2 and neutrophil collagenase, MMP-8 belongs to a collagenase subgroup of MMPs. Besides collagen, its substrates include other ECMproteins, proteases, protease inhibitors, cell adhesion proteins, growth factors and cytokines.18,19 MMP-8 is expressed by a wide variety of different cell populations including neutrophils, macrophages, plasma cells, T-cells, fibroblasts and epithelial cells.18 In colorectum, MMP-8 expression has been observed in murine colon mucosal cells by IHC.20 In turn, Asano et al.6 assessed MMP expression in human colon by qRT-PCR, but the expression levels of MMP-8 were too low both in normal and in cancerous tissue to allow a meaningful comparison. There are few other reports on the expression and function of MMP-8 in healthy or diseased colon, and as a consequence, little is known of the role MMP-8 plays in CRC. What separates MMP-8 from most other MMPs is the antitumor activity it has been reported to possess. MMP-8deficient male mice are more susceptible to developing carcinogen-induced skin tumors,21 which has been attributed to the ability of MMP-8 to participate in the processing of inflammatory mediators,21,22 and its ability to modulate tumor cell adhesion and invasion.23 Similarly, MMP-8-deficient female mice develop tongue cancer more often than wild-type mice, and MMP-8 has also been suggested to possess a protective role in human tongue cancer.24 Moreover, reducing MMP-8 levels in breast cancer cells by ribozyme knock-down technique increases their metastatic potential,25 and genetic variation in MMP-8 gene has been proposed to influence breast cancer prognosis by inhibiting metastasis.26 However, opposite reports also exist with local MMP-8 expression levels correlating positively with tumor stage, grade and poor prognosis in, e.g., ovarian cancer,27 indicating that some of the actions of MMP-8 might be tissuedependent.
The objective of this study was to investigate the possible role of MMP-8 in CRC pathogenesis. In more detail, we aimed to find out the differences in serum levels of MMP-8 (S-MMP-8) between CRC patients and healthy controls and to correlate the serum levels with clinicopathological parameters of tumors. Because MMP-8 is regarded as an important regulator of the immune system, particular interest was addressed to peri- and intratumoral inflammatory cell infiltrates. Specifically, serum levels of MMP-8 were of interest also providing an opportunity to identify the potential diagnostic or prognostic usefulness of peripheral blood MMP-8 measurements. Serum TIMP-1 (S-TIMP-1) analysis was included in the study to offer data of peripheral blood antiprotease-levels and blood neutrophil/lymphocyte ratio—a systemic inflammation-based prognosticator of CRC28—was included to provide a variable of comparison for S-MMP-8. Finally, tissue MMP-8 expression in selected CRC cases was studied to elucidate the sources of S-MMP-8 in CRC.
Material and methods Subjects
The study population consists of CRC patients operated consecutively in Oulu University Hospital between April 2006 and January 2010 and their age- and sex-matched controls. All the subjects were asked for a written informed consent to participate in the study, thus yielding 148 CRC patients and 83 healthy controls. Controls younger than 65 years were healthy blood donors (Finnish Red Cross, Oulu, Finland), and those aged 65 years or more were recruited from patients undergoing cataract surgery in Oulu University Hospital. The study design was approved by the Ethical Committee of Oulu University Hospital (58/2005 and 184/2009). Exclusion criteria for both patients and controls included earlier or simultaneous diagnosis of other cancer diseases. Controls younger than 65 years had to meet some other criteria, too, due to the regulations in blood donation including hemoglobin levels (135–195 g/l for men and 125–175 g/l for women) and absence of acute infections and certain chronic illnesses requiring continuous medication like severe allergies and asthma, epilepsy, diabetes and coronary artery disease. The details of age, sex, height, weight, medication and previous illnesses of controls and patients were acquired by a questionnaire. Height and weight were used to calculate body mass index (BMI). Age and sex data were available for all the subjects while data on height and weight of two patients was unavailable. Medical histories and clinical details of the patients were reviewed from patient’s case records, including preoperative differential white blood cell count and, if available by April 2011, data on recurrences by the 24-month follow-up visit. Thirty-two of 72 rectal cancer patients received preoperative radiotherapy or chemoradiotherapy (RT/CRT). Twenty-four of them received a short-term radiotherapy (a total radiation dose of 25 Gy in 1 week), whereas six received a 4-week-long radiotherapy (total radiation doses of 45–54 C 2011 UICC Int. J. Cancer: 131, E463–E474 (2012) V
Va¨yrynen et al.
Gy) along with a treatment with 5-fluorouracil, a cytotoxic drug, and two received a 4-week-long radiotherapy (total radiation doses of 45 and 48 Gy) without a cytotoxic drug. Small sample sizes did not enable us to study the effects of different treatment schemes on S-MMP-8 separately. Therefore, all preoperatively treated were considered as a group (RT/CRT group), and a stage-matched control group for them (RT/CRT control group) was randomly generated from other rectal cancer patients.
E465
NG0 denoted rare areas of necrosis, NG1 denoted frequent small areas of necrosis and NG2 denoted broad areas of necrosis. The evaluations of inflammatory reactions and necrosis were done independently by two researchers after which the cases with divergent evaluations were viewed again and mutual agreement was achieved. Three cases of 148 had inadequate sample material to evaluate the inflammatory cell reactions and the amount of necrosis accurately.
Preoperative serum samples of 148 CRC patients and serum samples of 83 controls were collected in glass tubes without a clot activator. After centrifugation, the supernatants were collected and stored at 70 C until the analysis. Serum MMP-8 concentrations were determined by a time-resolved immunofluorometric assay (IFMA).29 TIMP-1 ELISA (R&D Systems, Minneapolis, MN) was performed according to the manufacturer’s instructions with 1:300 dilutions of the serum.29 Histopathological analysis of primary tumors
The histopathological analysis and the preparation of tissue samples were conducted according to standard guidelines used in Department of Pathology, Oulu University Hospital. The samples from surgical specimens were fixed in 10% buffered formalin solution, embedded in paraffin and 5 lm sections were stained with haematoxylin and eosin. Staging was done according to TNM 6,30 and WHO histological grading31 was used to estimate the differentiation of the tumors. On average, 15 lymph nodes were examined. Related to palliative therapy, two cases of all 148 had inadequate sample material for staging while only one case was excluded from grading. Peritumoral inflammatory cell infiltrate was determined from haematoxylin-eosin-stained sections using the method described earlier.32 In short, a score of 0 was given when there was no increase of inflammatory cells, 1 denoted mild and patchy increase of inflammatory cells at the invasive margin, a score of 2 was given when inflammatory cells formed a band-like infiltrate at the invasive margin with some evidence of destruction of cancer cell islets and a score of 3 denoted a very prominent inflammatory reaction with frequent destruction of cancer cell islets. The scores were then classified as low-grade (scores 0 and 1) or high-grade (scores 2 and 3). Crohn’s-like lymphoid reaction (CLR; Fig. 1a), a lymphoid follicle reaction around the tumor, was evaluated according to the criteria established by Graham and Appelman.33 In short, cases were classified into three classes: CLR0 (no reaction) denoting no or at most one single lymphoid aggregate in all tumor sections, CLR1 (mild reaction) defined as occasional lymphoid aggregates with rare or absent germinal centers and CLR2 (intense reaction) denoting numerous lymphoid aggregates with germinal centers. Finally, haematoxylin-eosin-stained sections were also graded for the amount of necrosis (Fig. 1b) using a three-grade scale: C 2011 UICC Int. J. Cancer: 131, E463–E474 (2012) V
About 3.5 lm sections cut from paraffin-embedded specimens were deparaffinized in xylene and rehydrated through graded alcohols. For antigen retrieval, sections were pretreated with Tris–EDTA buffer (pH 9.0) in a microwave oven at 800 W for 2 min and at 150 W for 15 min. After cooling down to room temperature and neutralizing endogenous peroxidase activity, the sections were incubated at room temperature with mouse anti-human monoclonal antibody for MLH1 (1:200 dilution, BD-Pharmingen, San Diego, CA) for overnight or MLH2 (1:150 dilution, BD-Pharmingen, San Diego, CA) for 1 hr. Bound antibodies were detected using the NovoLinkTM Polymer detection system (Leica Biosystems, Newcastle, UK) or the EnVisionTM system (Dako, Copenhagen, Denmark), respectively. Diaminobenzidine (DAB) was used as the chromogen and haematoxylin as the counterstain. The expression was rated negative if there was no staining in any of the cancer cell nuclei. Normal proliferating tissue, e.g., crypt epithelium or germinal centers of lymphoid follicles was used as an internal positive control. Cancer cells devoid of the expression of either MLH1 or MSH2 were considered mismatch repair (MMR) enzyme deficient, while others were considered MMR-proficient. One subject of 148 had insufficient sample material to evaluate the expression. In our earlier report,34 MLH1 and MSH2 immunohistochemistry detected CRCs with high-level microsatellite instability with 97.8% specificity and 90.1% sensitivity. MMP-8 immunohistochemistry
Antigen retrieval for 3.5 lm sections cut from paraffin-embedded specimens was conducted with Tris–EDTA buffer (pH 9.0) in a microwave oven at 800 W for 2 min and at 150 W for 15 min. A rabbit polyclonal MMP-8 antibody,24 which has been shown not to cross-react with other members of MMP-family, was used in a 1:500 dilution with an incubation time of 1 hr at room temperature. Bound antibodies were detected using the EnVisionTM system. DAB was used as the chromogen and haematoxylin as the counterstain. When rating the expression, neutrophils inside blood vessels were used as an internal positive control. Statistical analyses
Normally distributed continuous variables are presented as mean (standard deviation), whereas other continuous variables are presented as median (interquartile range). The statistical analyses were carried out using statistical analysis
Early Detection and Diagnosis
Mismatch repair enzyme immunohistochemistry Collection of serum samples and measurement of MMP-8 and TIMP-1 levels
Early Detection and Diagnosis
E466
Serum MMP-8 in CRC
Figure 1. (a–b) Example histological images of Crohn’s like reaction (CLR) and a necrotic tumor. (a) Histological image demonstrating Crohn’s like reaction (CLR) defined as lymphoid follicles adjacent to the tumor. (b) Histological image demonstrating a necrotic tumor accompanied by a neutrophil infiltrate. (c–f) Serum MMP-8 in colorectal cancer and associations with clinical and pathological variables. (c) Boxplot showing the differences in S-MMP-8 between CRC patients and age and sex matched healthy controls. The median S-MMP-8 of the patients was more than three times higher. (d) Receiver operating characteristics (ROC) curve for S-MMP 8 in separating colorectal cancer patients from healthy controls. (e) Boxplot showing the association between S-MMP-8 and TMN stage. Higher MMP-8 levels associated with high-stage disease. (f) Boxplot showing the association between S-MMP-8 and CLR. Intense CLR was connected to lower S-MMP-8. (g) Boxplot showing the association between S-MMP-8 and necrosis. Extensive necrosis related to higher S-MMP-8. (h) Correlation scatter plot for logarithmically transformed S-MMP-8 and blood neutrophil count. All the p values are for Mann–Whitney U test. In boxplots, * denotes an outlier and * denotes an extreme value. Abbreviations: AUC: area under the curve; CI: confidence interval; Pearson r: Pearson correlation coefficient.
software PASW Statistics 18 (IBM, Chicago, IL). Statistical significances of differences in S-MMP-8, S-TIMP-1 and MMP-8/TIMP-1 ratio between different study groups and clinicopathological variable categories were analyzed by Mann–Whitney U test (comparing two classes) or Kruskal– Wallis test (comparing three or more classes). Pearson corre-
lation coefficients (r) were used to assess the correlations of MMP-8 and TIMP-1 with other peripheral blood parameters. A multiple linear regression analysis of the correlation of MMP-8 with selected clinicopathological factors was performed using the stepwise method. v2 test, Fisher’s exact test and Mann–Whitney test were used in the recurrence analysis. C 2011 UICC Int. J. Cancer: 131, E463–E474 (2012) V
Va¨yrynen et al.
E467
Table 1. Patient characteristics
Age, mean (SD)
All CRC patients (n ¼ 148)
RT/CRT group (n ¼ 32)
RT/CRT control group (n ¼ 32)
CRC patients without RT/CRT (n ¼ 116)
Healthy controls (n ¼ 83)
66.7 (11.1)
63.4 (10.3)
67.9 (10.2)
67.6 (11.2)
67.5 (10.7)
Sex Male
80 (54.1%)
22 (68.8%)
20 (62.5%)
58 (50%)
43 (51.8%)
Female
68 (45.9%)
10 (31.3%)
12 (37.5%)
58 (50%)
40 (48.2%)
Yes
32 (21.6%)
32 (100%)
0 (0%)
0 (0%)
No
116 (78.4%)
0 (0%)
32 (100%)
116 (100%)
48 (32.4%)
0 (0%)
0 (0%)
48 (41.4%)
Preoperative RT/CRT
Tumor location
Distal colon
28 (18.9%)
0 (0%)
0 (0%)
28 (24.1%)
Rectum
72 (48.6%)
32 (100%)
32 (100%)
40 (34.5%)
WHO grade
(n ¼ 147)
(n ¼ 115)1
1
Grade 1
21 (14.3%)
5 (15.6%)
5 (15.6%)
Grade 2
108 (73.5%)
22 (68.8%)
25 (78.1%)
86 (74.8%)
Grade 3
18 (12.2%)
5 (15.6%)
2 (6.3%)
13 (11.3%)
TNM stage
(n ¼ 146)2
(n ¼ 114)2
Stage I
27 (18.5%)
Stage II
54 (37.0%)
Stage III
46 (31.5%) 19 (13.0%)
1 (3.1%)
1 (3.1%)
Stage IV MMR enzyme status
16 (13.9%)
8 (25.0%)
8 (25.0%)
19 (16.7%)
9 (28.1%)
8 (25.0%)
45 (39.5%)
14 (43.8%)
15 (46.9%)
32 (28.1%)
(n ¼ 147)1
18 (15.8%) (n ¼ 115)1
MMR-deficient
11 (7.5%)
0 (0%)
0 (0%)
11 (9.6%)
MMR-proficient
136 (92.5%)
32 (100%)
32 (100%)
104 (90.4%)
1
No tumor tissue sample available for one patient. 2No tumor tissue sample or lymph node sample available for two patients. Abbreviations: CRC: colorectal cancer; MMR: mismatch repair; RT/CRT: radiotherapy or chemoradiotherapy; SD: standard deviation.
In all the tests, a two-tailed, exact p value less than 0.05 was considered statistically significant.
Results Patient characteristics
The study included two different study groups, CRC patients and healthy controls. The patient group was further divided into three subgroups (Table 1). The average age of the patients was 66.7 years. About 51.4% had colon cancer and 48.6% had rectal cancer. Eleven CRC cases of 147 studied showed negative expression of MLH1, whereas all the cases expressed MSH2. MMP-8 and TIMP-1 levels of different study groups
S-MMP-8, S-TIMP-1 and MMP-8/TIMP-1 ratios of the study groups are presented in Table 2. To first assess the potential confounding effect of preoperative RT/CRT, the values were compared between patients, who received the therapy and their stage and tumor location matched control group. Those who received the therapy had lower MMP-8 levels (median C 2011 UICC Int. J. Cancer: 131, E463–E474 (2012) V
42.6 vs. 33.4 ng/ml, p ¼ 0.227). Although this difference was not statistically significant, we decided to first exclude the RT/CRT group from the univariate analyses of clinical and pathological parameters to avoid any confounding effects, taking into consideration the fact that preoperative RT/CRT affects the local characteristics of the tumor.35 For comparison, the analyses were then repeated for the all CRC patients group and for the RT/CRT group alone, and the results of these analyses are displayed in the Supporting Information. The values of the healthy controls were also compared to the patients who did not receive RT/CRT to better model the situation before the diagnosis of the disease. The median SMMP-8 of the patients was more than three times higher than that of the healthy control group (63.0 vs. 17.2 ng/ml, p ¼ 1.5E 9; Fig. 1c). A receiver operating characteristic (ROC) analysis was conducted to evaluate the potential of SMMP-8 in separating the CRC patients from the controls (Fig. 1d). The analysis yielded an area under the curve (AUC) of 0.751 (95 % confidence interval 0.685–0.817). Using a cut-off value of 63.4 ng/ml, specificity was 90.4 %,
Early Detection and Diagnosis
Proximal colon
Serum MMP-8 in CRC
E468
Table 2. S-MMP-8 and S-TIMP-1 levels of the patient and control groups included in the study S-MMP-8 (ng/mL)
S-TIMP-1 (ng/mL)
MMP-8/TIMP-1
Variable studied and study group
Median (IQR)
All subjects (n ¼ 231)
31.9 (14.7–75.4)
183.2 (158.6–214.3)
0.17 (0.08–0.42)
All CRC patients (n ¼ 148)
52.7 (22.6–110.3)
180.2 (148.8–214.5)
0.30 (0.12–0.57)
p value
p value
Median (IQR)
Median (IQR)
p value
Preoperative RT/CRT RT/CRT group (n ¼ 32)
33.4 (23.5–57.4)
RT/CRT control group (n ¼ 32)
42.6 (18.5–137.4)
0.227
173.2 (145.9–190.7)
0.638
180.9 (146.4–203.3)
0.21 (0.12–0.36)
0.212
0.33 (0.10–0.72)
Patients and controls Patients without RT/CRT (n ¼ 116)
63.0 (22.1–128.7)
Healthy controls (n ¼ 83)
17.2 (9.6–40.0)
1.5E 9
185.0 (145.5–217.4) 188.8 (168.5–212.5)
0.382
0.34 (0.12–0.66)
7.6E 10
0.09 (0.05–0.21)
p values are for Mann–Whitney U test. Abbreviations: CRC: colorectal cancer; IQR: interquartile range; RT/CRT: radiotherapy or chemoradiotherapy.
and sensitivity was 50.0 %. S-TIMP-1 values did not differ significantly between any of the groups studied leaving also serum MMP-8/TIMP-1 ratio elevated in CRC.
Early Detection and Diagnosis
Associations between serum MMP-8 levels and clinical parameters of patients and histopathological properties of tumor
Table 3 presents the associations between S-MMP-8 and clinicopathological parameters. First, the effects of age, sex, BMI, tumor location and mismatch repair enzyme status on SMMP-8 were determined, but no statistically significant differences were found. Next, S-MMP-8 was correlated with TNM stage and WHO histological grade. Higher S-MMP-8 was observed in more advanced stage disease (p ¼ 4.5E 4). Stage IV disease, signifying the presence of distant metastases, yielded the highest S-MMP-8 values (Fig. 1e). The median S-MMP-8 in stage III was—although insignificantly— lower than in stages I and II (26.2 vs. 42.6 ng/ml, p ¼ 0.686). Also, TIMP-1 levels correlated positively with advancing stage (p ¼ 0.0066), but MMP-8/TIMP-1 ratio remained elevated in stage IV relative to other stages (p ¼ 0.0031). Similarly, increasing depth of primary tumor invasion (pT) associated with higher S-MMP-8 (p ¼ 0.0035). Well-differentiated tumors had lower S-MMP-8, but the difference was not statistically significant. Of the inflammatory parameters studied, higher S-MMP-8 associated with both lower-grade peritumoral tumor-destructing inflammatory cell infiltrate (p ¼ 0.041) and CLR (p ¼ 0.0057; Fig. 1f). A positive correlation was observed between S-MMP-8 and necrosis (p ¼ 0.0024; Fig. 1g). The differences were most prominent when CLR2 and NG2 were compared to other classes (p ¼ 0.0014 and 5.6E 4, respectively). The analyses, which included RT/CRT group, showed a visible confounding effect especially in the inflammatory variables and WHO grade (Supporting Information Tables 1 and 2). For comparison, the relations between blood neutrophil/lymphocyte ratio and the clinicopathological variables were also determined (data not shown), and in our material, the only variable to show an
association with blood neutrophil/lymphocyte ratio was higher-grade necrosis (p ¼ 0.019). To find out any possible interconnections between SMMP-8, S-TIMP-1 and differential white blood cell count parameters in CRC patients, Pearson correlation coefficients were calculated (Table 4). Logarithmic transformations were applied for all the parameters studied to reduce positive skewness. In the analysis, high S-MMP-8, most notably, associated with high blood neutrophil count (Pearson r ¼ 0.523; Fig. 1h) and high blood leukocyte count (Pearson r ¼ 0.411). Positive correlation of S-MMP-8 with S-TIMP-1 was also observed. S-TIMP-1 correlated positively with blood neutrophil and monocyte counts. Comparable results were acquired for the all CRC patients group (Supporting Information Table 3) and for the RT/CRT group (Supporting Information Table 4). A multiple linear regression analysis was carried out to further analyze the association of S-MMP-8 with selected clinicopathological factors. Logarithmic transformations were applied for S-MMP-8, blood neutrophil count and BMI. Several categorical variables were also reclassified for the analysis. TNM parameters were grouped into distant metastases (M0 vs. M1), nodal metastases (N0 vs. N1–N2) and primary tumor invasion through muscularis propria (T1–2 vs. T3–4). Based on the univariate analyses, we combined CLR grades CLR0 and CLR1 and necrosis grades NG0 and NG1. Those who received preoperative RT/CRT were also included in the analysis, and preoperative RT/CRT was added as a potential confounding factor in all the models studied. We decided to restrict the number of independents in the final model to five due to our relatively small study population. Considered variables included age, sex, BMI, mismatch repair enzyme status, blood neutrophil count, tumor location, distant metastases, nodal metastases, invasion through muscularis propria, CLR, tumor-destructing peritumoral inflammatory infiltrate and necrosis. Besides preoperative RT/CRT, the model presented consists of four best-correlating independents added using the stepwise method (Table 5). Cases were excluded likewise, C 2011 UICC Int. J. Cancer: 131, E463–E474 (2012) V
Va¨yrynen et al.
E469
Table 3. Associations of serum MMP-8 and TIMP-1 levels and MMP-8/TIMP-1 ratio with clinical parameters of patients and histopathological properties of tumors S-MMP-8 (ng/mL) Clinical or pathological variable
Median (IQR)
S-TIMP-1 (ng/mL)
MMP-8/TIMP-1
p value
Median (IQR)
p value
Median (IQR)
p value
0.788
173.7 (134.9–220.5)
0.294
0.34 (0.13–0.68)
0.562
Age (years)