Differential Associations of MMP-2 and MMP-14 with Stromal Amounts and T-lymphocyte Presence in Ovarian Cancer

Differential Associations of MMP-2 and MMP-14 with Stromal Amounts and T-lymphocyte Presence in Ovarian Cancer M Caroline Vos1*, Elise Bekers2, Brenda L den Oudsten3, Ella M. de Boed2, Anneke A M van der Wurff2, Toin H van Kuppevelt4, Katarina Wolf5 and Leon F A G Massuger6 1Department of Obstetrics and Gynaecology, ElisabethTweesteden Hospital, The Netherlands 2Department of Pathology, ElisabethTweesteden Hospital, The Netherlands 3Department of Medical and Clinical Psychology, Tilburg University, The Netherlands 4Department of Biochemistry, Nijmegen Centre for Molecular Life Sciences, Radboud University Medical Center, the Netherlands 5Department of Cell Biology, Radboud Institute for Molecular Life Sciences, The Netherlands 6Department of Obstetrics and Gynaecology, Radboud University Medical Center, The Netherlands


Introduction
Ovarian cancer is known for its bad prognosis due to the lack of screening possibilities and the late onset of symptoms leading to diagnosis mostly in advanced disease. Despite intensive treatment with debulking surgery and chemotherapy, 5-year survival rates remain low [1].
In the past few years, the investigation of the stromal compartments of tumors has increasingly gained importance [2][3][4][5]. Stromal alterations are commonly found around the invasive front of a malignant tumor. One such alteration can be described as a desmoplastic stroma reaction, characterized by a dense fibrotic network with deposition of various types of fibrillar collagen. The presence of desmoplastic stroma correlates with adverse prognosis in for example endometrial cancer as well as in a subset of ovarian cancers [5][6][7].
The maintenance of stroma often correlates with the upregulation of matrix metalloproteinases (MMPs), especially MMP-14, as a feedback loop to keep stroma at constant density [8]. In cancer however, this process is often deregulated and connective tissues develop into dense tumor stroma despite MMP upregulation and the possibility to break up tissue again [9][10][11].
MMP's are ubiquitous zinc-dependent proteases that are important in numerous physiological and pathophysiological processes, including Page 2 of 7 peripheral blood vessels and must travel through connective tissue compartments in order to reach and kill the tumor cells ( Figure 1) [19].
The efficiency of T-cell trafficking is influenced by extracellular matrix (ECM) deposition, its geometry, which includes a guidance as well as a barrier function, and proteolytic remodeling which may open tracks of least resistance for moving cells. In malignant stroma the network densifies, develops thickened matrix fibers and thus may hinder T-lymphocyte migration [3].
In ovarian cancer, both the amount of stroma deposition relative to the tumor mass ("stroma percentage") and T-lymphocyte presence influence prognosis [19][20][21]. Both survival-promoting presence of tumour infiltrating T-lymphocytes and survival-impairing high stroma percentage is reported for advanced ovarian cancer [19][20][21]. However, whether and how these survival-determining factors correlate with modulated MMP-expression typical in ovarian cancers remains unknown.
We here explored in a large regional cohort with long-term followup whether MMP-2 and MMP-14 expression was associated with these stromal and stroma-derived processes in ovarian cancer for the reduction of stroma mass by ECM turnover and attraction of T-cell influx. Our data indicate, opposite to our expectations, that MMP-2 expression positively correlates with relative stromal amounts and MMP-14 negatively associates with T-cell presence in tumor tissue.

Clinical data
Included in this retrospective cohort were all 116 patients diagnosed with ovarian cancer between January 1997 and December 2003 at St. Elisabeth Hospital and Tweesteden Hospital, both in Tilburg, and at Amphia Hospital, Breda all in, the Netherlands. Patients who had undergone neo-adjuvant chemotherapy with secondary cytoreductive surgery were excluded (n = 11), because previous chemotherapy may have influenced the immunohistochemical results of MMP's and T-lymphocytes of histopathological material obtained at the time of surgery [22]. All patients were followed until March 2013. Four patients emigrated and were thus lost to follow-up. For 15 patients, no residual tumor material was available for complete immunohistochemistry, leaving 86 patients for which MMP-14 and MMP-2 expression and T-lymphocyte presence were determined. For a flowchart of the patients see Figure 2.
The patients underwent a laparotomy with staging or debulking when indicated by clinical stage and frozen-section results. Debulking surgery was found to be optimal if the maximum diameter (length, width or depth) of the individual residual tumor deposits were less than 1 cm. In patients with FIGO stage Ia or Ib ovarian cancer with differentiation grade I, no adjuvant therapy was given. All other patients received 6 to 9 courses of adjuvant platinum-based chemotherapy. The only treatment change during the study period was in adjuvant polychemotherapy. In 1999, the combination of cisplatin and cyclophosphamide was changed to carboplatin or cisplatin combined with paclitaxel. FIGO stage (I to IV) was categorized according to World Health Organization (WHO) criteria. Survival data were collected from the medical records, hospital-registered death certificates or registration by the general practitioner of the patient.

Histopathological data
All histopathological results and original slides were reviewed by an independent pathologist, specialized in gynecological pathology. Histology and differentiation grade were categorized according to WHO criteria, grading was based on the observer's impression of both architectural and cytological features [23]. In the original hematoxylin eosin slides, both the areas of tumor epithelium and stroma percentage were quantified from three random images taken adjacent to the epithelial tumor fields ( Figure 3) and averaged according to Labiche et al. [21]. Necrotic or adenofibromatotic components of the tumor were excluded from the percentage calculation.

Semiquantitative immunohistochemistry
Paraffin-embedded blocks were selected from the archives of the histopathology laboratory and immunohistochemistry was performed as described before [13]. In brief, sections of 3 μm thickness were deparaffinized in xylene and rehydrated with graded alcohol. Each slide included positive controls for MMP's consisting of placenta. Each run included negative controls without primary antibody. Endogenous peroxidase was blocked with 3% H 2 O 2 and 5% normal goat serum. The primary antibodies used for staining are described in Table 1. After each incubation step, slides were washed two times with phosphate buffered saline (PBS).  Zhang et al. on clinical characteristics, we followed their methodology in the number and selection of fields and markers despite newer insights into the markers for regulatory T cells [19,25].

Statistical analysis
Statistical analysis was performed using the Statistical Package for Social Sciences 20.0 (SPSS Inc., Chicago, Il). Clinicopathological and survival data and immunohistochemistry scores of early-stage and advanced-stage patients were compared using independent samples t-tests for continuous variables if normally distributed, Mann-Whitney tests if continuous variables were not distributed normally and chisquare tests for categorical variables.
To determine the correlation coefficients between MMP-14 and MMP-2, T-lymphocytes and stromal volume, Spearman's rho correlation coefficients were calculated; a value of 0.1 being considered a small correlation, 0.3 medium and 0.5 a high correlation.

Results
Of the 101 patients in this consecutive series from a regional laboratory, 33 were early-stage patients (FIGO Ia-IIa). The median follow-up time was 59 months (minimum 0-maximum 195). Patient characteristics including histopathological subtypes reflect the distribution pattern of diagnosis in the Netherlands 1989-2009 [1]. The patient characteristics are summarized in Table 2. Because the pathophysiology of early-stage ovarian cancer and advanced-stage ovarian cancer are different, we present the results for the two groups in the tables seperately. To relate MMP staining to epithelial tumor or stromal areas and determine the number of T-cell infiltrates, firstly, MMP immunohistochemistry was performed and quantified (Figures 3 and 4) ( Table 3 upper part). The MMP-14 and MMP-2 expression in the tumor epithelium was positive (Overall Score 1) in 53/94 and 84/94 of all patients (early and late stage together), respectively, and in the tumor stroma of 56/94 and 30/94 patients (with staining intensity scored weak or strong, respectively). Second, Table 3 (lower part) shows the median T-cell count for the different markers. For the T-lymphocyte parameters, a large range of counts was observed for both the tumor epithelium and the stromal compartment. T-cells counts were recoded into < 5/High Power Field (HPF), 6-20/HPF and > 20/HPF according to Zhang [19]. Approximately half of the patients had five or less positive T-lymphocytes per HPF in the tumor epithelium (58.1% for CD45Ro, 58.1%for CD8 and 50.0% for CD3). In the stromal compartment, more CD45Ro, CD8 and CD3 positive T-lymphocytes were present. Only 27.9%, 32.6% respectively 19.8% of patients had five or less positive T-lymphocytes in the stromal compartment of the tumor. More T-cells were found -both epithelial and stromal-in advanced stage patients compared to early-stage patients, but only for cytotoxic T-cells (CD8) in stroma this difference was significant.
Third, in early-stage patients, the median percentage of stroma was 40% and in advanced-stage patients 30% (Table 3). Figure 3 illustrates different examples of a tumor with a low and a tumor with a high percentage of stroma. In the lower panels, the slide with CD3 immunostaining is presented.
In Figure 4 different examples of stroma are illustrated with A) necrotic and hemorrhagic stroma, B) more fibrillary and dense stroma and C) normal stroma with inflammatory cells.
Median overall survival for advanced-stage patients was 25 Slides were incubated with secondary antibody, poly-HRP-GAM/ R/R IgG Powervision (Immunologic, Duiven, the Netherlands) for 60 minutes at room temperature. Staining was done with diaminobenzidine (Immunologic, Duiven, the Netherlands) for 5 minutes. Slides were counterstained with hematoxylin.
T-lymphocyte count was recorded according to Zhang et al. [19]. In short, T-lymphocytes were counted in three fields of the tumor compartment and three for the stromal compartment and averaged. CD3 was used as a pan-T-lymphocyte marker. CD8 was used as a cytotoxic T-lymphocyte marker as the presence of CD8 positive cytotoxic lymphocytes is most important for tumor cell killing [19]. The total amount of intratumoral T-cells consist of the CD3 positive cells. CD3 positive cells minus CD8 positive cells will make up the helper T-lymphocyte compartment. Of the helper T-lymphocyte compartment, the regulatory T cells with CD45Ro as a regulatory T-lymphocyte marker, seem to be the most interesting for prognosis [24]. For these markers, the same fields were counted as the fields in the slides used for CD3 determination. In order to make our result comparable to the results of  months, respectively. These differences were not statistically significant, but in line with the previous reported results on survival [19].
If the stroma percentage is dichotomized at the median of 30% for advanced-stage patients, median overall survival is 19 (minimum 0-maximum 174) months for patients with <30% stroma and 44 (minimum 6-maximum 191) months for patients with ≥30%. Also, this difference is not statistically different.

Correlation between MMP-14 and MMP-2 expression, T-lymphocytes and stromal volume
Contrary to our original hypothesis that MMP-14 expression in the tumor would positively correlate with T-cell influx, we found that epithelial MMP-14 expression negatively correlated with the presence of T-lymphocytes and yielded correlation coefficients rho -0.36, p < 0.001 for CD3; rho -0.30, p < 0.005 for CD8; and rho -0.24, p < 0.03 CD45Ro (Table 4). No significant correlations between stromal MMP-14 and MMP-2 and the presence of T-lymphocytes in the stromal compartment were found. Neither significant correlations were found for MMP-2 expression in the tumor epithelium nor in the stroma on T-cell influx. Interestingly, high MMP-2 in stroma significantly correlated with high stroma percentage (rho 0.29, p < 0.008), which is also contrary to our original hypothesis.
In conclusion, MMP-14 expression in the tumor epithelium negatively correlates with T-cell influx and stromal MMP-2 expression correlates with the amount of stroma.

Discussion
In this study, we investigated the relationships between MMP-    Table 3: Results of immunostaining with p= values between early-stage and advanced stage using Mann-Whitney U tests.

Page 5 of 7
14 and MMP-2 expression in the tumor epithelium and the stromal compartment, together with the presence of T-lymphocytes and the amount of stroma in a regional cohort of ovarian cancer patients. We found a significant negative association between MMP-14 expression within the tumor epithelium and the presence of intratumoral T-lymphocytes. In addition, high stromal MMP-2 expression was positively correlated with stroma percentage. No other significant relationships were found.
It is challenging to interpret our results from a pathophysiological viewpoint. The ovaries are a unique environment to study relationships between MMP parameters and stromal factors. Most ovarian cancers probably do not arise in the ovaries themselves, [26,27] and in terms of disease characteristics, ovarian tumors are strikingly different to tumors in other organs. The main distinguishing features of ovarian tumors are their often cystic nature, their large size at detection and the prolonged confinement of metastases to the peritoneal cavity.
Regarding stroma percentage, our results are not fully comparable with those in the study by Labiche et al. [21]. For advanced stage, the median stroma percentage was 30% in our group and 50% in their group. In early-stage tumors, we found a median percentage of 40%, while early stage tumors were not included in their study. This may be attributable to the different populations that the patients come from, ours being a large regional cohort and theirs from a university referral center, resulting in a possible selection bias. In line with the larger stroma percentage we found in early-stage patients, the relationship between stroma percentage and survival was also different than found in Labiche's study. In our group of advanced stage patients, those with a stroma percentage larger than the median percentage of 30 had a longer survival, although not statistically significant. Perhaps this longer survival goes along with overall less formed tumor mass in relation to the stroma mass. Also, perhaps is more stroma mass a result of less MMP-2 production which regulates part of the, though uncontrolled, stroma turnover. In line with this, in a study on MMP-2 blockade and stroma formation in skin cancer, even though the amount of stroma relative to tumor epithelium was not described in that study, an inhibitory effect on vessel formation and invasion was found [28], reasoning for a better overall survival.
We found a negative association between epithelial MMP-14 expression and T-lymphocyte presence, indicating a role for tumor-expressed MMP-14 in inhibiting T-lymphocyte trafficking towards the tumor epithelium. Our result of the inverse effect of MMP-14 expression on T lymphocyte presence is controversial. For T-lymphocyte infiltration, structural characteristics of malignant stroma are important, in particular, how the arrangement of collagen fibers facilitates or hinders T-lymphocyte trafficking, as shown in an example of lung cancer. When treated with collagenases, T-lymphocyte movement towards the tumor increases. This is in favor of the concept of partial degradation and thus widening of the collagen network for increased T-cell migration [29]. More T-lymphocyte infiltration, especially CD3+CD8+ lymphocytes will result in a better prognosis for the patients [25]. In our advanced-stage patients, we found significantly more CD8+ lymphocytes in stroma. In another example in lung cancer, lymphoid structures containing a reticular network are sometimes located far from the tumor at the boundary between tumor stroma and healthy tissue [3]. In ovarian cancer, this boundary is often less clear, because of the lack of an invasive front by microscopic detection. However, also other adverse roles of MMP-14 are possible. Studies on genetic profiling in ovarian cancer suggest that some tumor types with a mesenchymal, and thus MMP-14-positive profile have less lymphocyte infiltration than other tumor types such as high graded serous and endometroid tumors [7]. From this perspective, high MMP-14 expression in an ovarian tumor may reflect a mesenchymal profile and thus contribute to less T-lymphocyte infiltration, for yet unknown reasons. Unfortunately, analyzing different histopathological groups in our sample will lead to a too small sample size to draw any conclusions.
Comparison of our results with those of earlier studies on T-lymphocytes yields the following conclusions. Zhang et al, established the influence of the presence of tumor-infiltrating lymphocytes on prolonged survival in ovarian cancer [19]. This is in line with our results. However, the underlying mechanisms for this finding are poorly understood.
The roles that MMP-14 and MMP-2 play in malignant tumor development are largely unknown. From the studies on normal ovaries [8,30] and on auto-immune disease [31,32] MMP-14 expression strength should positively correlate with T-cell infiltration, but negatively correlate with stroma percentage. The findings in our presented study show opposite results with the presence of MMP-14 in the tumor epithelium correlating negatively with T-lymphocytes in the tumor epithelium. No relation between MMP-14 and stroma percentage is found, only for stromal MMP-2 a positive correlation is found. The results from the present study thus suggest that in the ovarian environment, the roles of MMP-14 and MMP-2 in malignant stroma are quite different from the roles that MMP-14 and MMP-2 may play in healthy ovarian stroma with an expected positive correlation between MMP-14 and T-cell infiltration and a negative correlation between MMP-2 and stroma percentage [3].
In auto-immune disease, MMP-14 and MMP-2 are important for T-lymphocyte influx via CD44, the integrin pathways and TGF-β [31,[33][34][35]. In ovarian cancer, we found in contrast a negative correlation between the presence of the different types of T-lymphocytes and MMP-14 expression. It seems logical to assume that the degrading action of MMP's on the ECM would facilitate T-lymphocyte infiltration of the tumor stroma followed by infiltration of the tumor mass, but clearly other actions of MMP-14 on signaling pathways are more important. It is possible that chemotactic signals in the micro-environment of the tumor could be more important than the arrangement of the collagen fibers [29,33]. The negative association of MMP-14 on T-lymphocyte influx is in accordance with the negative effect of MMP-14 expression and survival [16]. This negative correlation that we have found would favor the use of MMP-14 inhibitors, which have been, however, already been reported to be ineffective in advanced ovarian cancer [36]. Since correlation coefficients of this MMP-14/T-cell presence relationship are around 0.30, indicating a moderate effect, other factors must also play a role, like maybe CD44 and TGF-β [33,37].