The opinion and assertions contained herein are the private views of the authors and are not to be construed as official or as representing the views of the Department of the Army or the Department of Defense.
Matrix metalloproteinases (MMPs) are a family of highly homologous, zinc- and calcium-dependent extracellular enzymes classified into 5 groups (collagenases, gelatinases, stromelysin, matrilysin and the membrane-type MMP) based on substrate specificity, protein domain structure, sequence homology and ability/inability to be secreted.1, 2, 3 They are implicated in tumor invasion and metastasis.4, 5 Signals from tumor cells activate the production of MMPs, which degrade/digest the extracellular membrane, thus aiding angiogenesis/cell migration. This process sets the stage for hematological spread/metastasis.6, 7 Because tissue MMP may leach into the blood stream and increase the circulating levels, it is believed that MMP profile in the blood could serve as biological markers for disease onset, progression or monitoring.
MMP2 and MMP9 (gelatinases A and B or 72- and 92-kDa type IV collagenases) are of particular interest because of their role in early cancer development and progression. Specifically, MMP2 and MMP9 degrade gelatin and type IV collagen, the main components of basement membrane (BM), which is the main barrier separating in situ and invasive carcinoma.8, 9 Degradation of the BM is associated with tumor spread. Immunohistochemical studies of breast tumors from lymph node-negative and node-positive patients have indicated possible prognostic value of MMP2 and MMP9,10, 11 suggesting that MMP2/MMP9 may be useful as markers for early breast cancer (BC) detection, disease classification/prognostication or treatment monitoring.
In biological systems/fluids, MMPs occur either as inactive proenzymes, active enzymes and inactive enzymes bound by various inhibitors.12 Activation in vivo usually occurs by proteolytic removal of the prosequence by other proteases through a proteolytic cascade.13 The functional status of MMP's can be obtained by measuring their activity. While gelatin zymography has been utilized in the discrimination between inactive and active MMPs (aMMPs),14, 15 it is the development of chromogenic or fluorogenic synthetic peptide substrates16 and assay protocols that have enabled more quantitative and accurate measurement of MMP activity.12, 17, 18, 19 These have led to the development of commercially available immunocapture assay kits that allow the simultaneous and precise measurement of already aMMP or endogenous MMP and the proenzyme (activatable) form.12, 19
aMMP, active MMP; BC, breast cancer; BD, benign disease; BM, basement membrane; CBCP, Clinical Breast Care Project; CO, control; DCIS, ductal carcinoma in situ; HR, high risk; MMP, matrix metalloproteinase; NAF, nipple aspirate fluid; pMMP, pro-MMP; TIMP, tissue inhibitor of metalloproteinase; tMMP, total MMP; WRAMC, Walter Reed Army Medical Center.
Although available reports suggest that MMP2 and MMP9 may be involved in the early process of tumor development1, 20 and that elevated levels of MMP2 and MMP9 occur in serum and plasma of cancer patients,21, 22 there are no studies that have quantitatively measured and compared the concentration and activity of the active-forms and proforms in serum of patients at low/HR of developing BC (based on the Gail model) to those with breast disease and BC. In an earlier study,19 we utilized an immunocapture assay to quantitatively determine the level and activity of MMP2 and MMP9 in plasma of 124 donors and correlated our results with breast disease status or risk of developing BC (determined by the Gail Model). We observed that the plasma concentrations and activity of total MMP2 (tMMP2) and aMMP9 were significantly lower in the control group than in the HR, benign and cancer groups and the concentration of aMMP2 was significantly higher in the control than HR, benign and cancer. Activity of MMP2 in plasma was significantly lower in HR individuals compared to benign, cancer and the control groups. While our study19 further supported the potential role of MMP2/9 in breast disease detection and classification, answers are still needed to the questions regarding the possible contribution of other factors,23, 24 for example, the best medium to use (plasma or serum), and how the use of serum or plasma impacts the sensitivity and usefulness of MMP2 and MMP9 as breast disease markers. We believe that the full potential of exploiting MMP2 or MMP9 as circulating markers for disease monitoring or classification can only be determined after establishing whether the levels and activity in serum and plasma are comparable and whether one medium (serum or plasma) is better suited for the measurement of MMP2/9.
The aims of the present study were therefore to determine the following: (i) whether MMP2 and MMP9 concentration or activity are distinct in the serum of donors classified as low risk or HR (based on the Gail Model) or donors diagnosed with benign breast disease or BC (based on pathological classification) and (ii) whether the concentration or activity trend in serum correlates with the trend observed in plasma of other female donors19 enrolled in the same research protocol. The present study therefore mirrors our earlier plasma study, but specifically provides the best direct comparison of the levels and activity of circulating MMP2 and MMP9 in serum of females with breast disease, BC and at risk of developing BC. Additionally, it elucidates which medium (plasma or serum) is likely the best for quantitative measurement of specific forms of circulating MMP2 and MMP9.
Material and methods
The Institutional Review Boards at Walter Reed Army Medical Center (WRAMC), Washington DC and Windber Medical Center, Windber PA, approved the protocol for our study. Study subjects were fully informed and consented donors attending clinics of the Clinical Breast Care Project (CBCP) in Windber, PA, Johnstown PA and WRAMC between 2001 and 2004. Patients were pathologically classified based on a standardized CBCP format25 into BC or benign disease (BD) and by the Gail Model26 into HR (5-year risk ratio ≥ 1.67%) or low risk, that is the control (5-year risk ratio < 1.67%). Serum collection and processing was as previously described.25 Samples were appropriately coded (to protect donor confidentiality) and stored in a tissue bank. Accompanying questionnaires and data collection sheets were similarly coded and entered in a relational database. Our experimental design called for the inclusion of only female individuals (in our database) who donated serum samples prior to any clinical evaluation and disease diagnosis. We queried the CBCP database for such patients/volunteers who were subsequently diagnosed with BC or BD, or estimated to be at HR or low risk (control) of developing BC. A total of 345 patients met the inclusion criteria. These consisted of 88 BC cases (age range 22–81 years, mean age 56.4), 150 BD cases (age range 19–85 years, mean age 48.4 years), 46 HR cases (age range 45–77 years, mean age 57.4 years) and 61 low risk or control cases (CO, age range 24–68 years, mean age 42.9 years). Donor details are as shown in Table I.
Table I. Characteristics of Donors Within the Study Group
Total protein concentration in each sample was determined by the Bradford (Bio-Rad Protein Assay, Bio-Rad Laboratories, CA) microassay procedure.19 Briefly, 160 μl of appropriately diluted standards and serum samples were placed in wells of a microtiter plate and 40 μl of dye reagent concentrate added to each well. After mixing and incubation for at least 5 min at room temperature, the absorbance was measured at 595 nm. Bovine serum albumin was used as standard.19
Measurement of concentration and activity of MMP2 and MMP9
We utilized the BiotrakTM system (GE Healthcare, Piscataway, NJ) for determination of serum concentration of aMMP and total MMP (tMMP). Briefly, endogenous levels of aMMP or tMMP (i.e. aMMP + pMMP [pro-MMP]) were captured from 100 μl of appropriately diluted serum incubated overnight at 4°C with 100 μl assay buffer (0.01 M sodium phosphate buffer pH 7.0 containing 0.05% TweenTM 20). After the overnight incubation, wells were washed 4 times with wash buffer. Then 50 μl of aminophenylmercuric acetate was added to wells containing the standards and into the wells where tMMP was to be measured, whereas 50 μl of assay buffer was added into wells where aMMP was to be measured. For the MMP9 assay, a further incubation period of 1.5 hr at 37°C was carried out before detection. The detection process involved the addition of 50 μl of detection reagent (modified urokinase and chromogenic peptide substrate S-2444) to all wells and incubation at 37°C for 3–4 hr before reading the absorbance at 405 nm on a spectrophotometer (SPECTRAmax 190, Molecular Devices Corporation, Sunnyvale, CA). The concentrations of active and total MMP2 and MMP9 were extrapolated from the standard curve and these values were normalized by individual patient total serum protein to obtain concentrations in pg/mg protein (Eq. (1))
Activity of MMP2 and MMP9, which is equivalent to the rate of change of absorbance at 405 nm, was determined as previously reported.19 Briefly, the activity is directly proportional to the intensity of the color generated following the cleavage of the chromogenic substrate (Eq. (2)).
where h is the incubation time (hr).
Data was expressed as arbitrary units defined as δAb405/h2 × 1,000 and normalized by individual patient total protein (units/mg protein) as previously reported.19 The total activity attributable to the already active forms of MMPs (aMMPa) and to the activatable pro-MMP forms (pMMPa) is obtained from Eq. (3).
Test for statistical significance
We employed the Kruskal–Wallis and Mann–Whitney nonparametric tests to determine the statistical significance of the differences in the concentrations and activity/mg protein of serum MMP2 and MMP9 between the study categories. p-values ≤ 0.05 were considered statistically significant in each case. The sample size available for our study was sufficient to detect differences of ∼67% between patient groups based on the measured variables with 80% power, using a Mann–Whitney test with two-sided significance level of 0.05.
Linear discriminant analysis
Linear discriminant analysis (also referred to as canonical discriminant analysis) was applied to the concentration and activity datasets to extract a set of linear combination of the quantitative variables that best reveal the differences among patient groups.19 The linear discriminant analysis tool was also used to build a predictive/descriptive model of group discrimination based on observed predictor variables and to classify each observation into one of the groups.19 Independent validation of classification accuracy was estimated based on the leave-one-out cross-validation approach. In this method, each case in the analysis was classified by the functions derived from all cases other than that case. Data were analyzed on a natural log scale and all statistical analyses were performed using SPSS® for Windows version 12.0 (SPSS, Chicago, IL).
Test for the effect of age, tumor stage/grade and tumor metastatic status
We employed a two-way ANOVA with interaction and log-transformed data27 to determine whether concentration and activity of serum MMP2 and MMP9 was (i) influenced by age of donor or (ii) tumor stage, grade and metastatic status. The donors were grouped into “below” and “above” 50 years of age at the time of disease diagnosis and patients with invasive BC were stratified into those with well, moderate or poorly differentiated tumors, and tumor stages I and II (A and B).
Concentration and activity of MMP2 in serum
Concentration of aMMP2 ranged from a mean of ∼339 pg/mg protein in BC patients to a mean of ∼854 pg/mg protein in the HR donors while the concentration of tMMP2 ranged from a mean of ∼8,903 pg/mg protein in the controls to a mean of ∼10,813 pg/mg protein in the HR individuals. Statistical tests showed that the concentration of aMMP2 was higher in the HR (median = 699.5 pg/mg protein) than the cancer and benign (p < 0.001 in both cases). No significant difference was observed between the HR and control groups (p = 0.162) (Fig. 1a). The control group had significantly higher concentration of aMMP2 (median = 495.5 pg/mg protein) than benign and cancer (p < 0.001, respectively; Fig. 1a) but the concentration of aMMP2 in the cancer group was not different from the benign (p = 0.474) group. tMMP2 concentration was significantly higher in the HR group (median = 9669.7 pg/mg protein) than control (p = 0.012), benign (p = 0.001) and cancer (p = 0.007) groups (Fig. 1b).
The activity of MMP2 ranged from a mean of ∼4 U/mg protein in the benign group to a mean of ∼10 U/mg protein in HR groups, whereas the total activity (active-form + proform; Eq. (3)) ranged from a mean of ∼80 U/mg protein in controls to a mean of ∼210 U/mg protein in BC patients. Statistical tests for significance showed that MMP2 activity was higher in the control (median = 5.2 U/mg protein) than benign (p < 0.001) and cancer (p = 0.026) (Fig. 1c). MMP2 activity was also higher in the HR group (median = 7.5 U/mg protein) than the control (p = 0.009), benign and cancer (p < 0.001, respectively). The activity of tMMP2 (active + pro form) was higher in the cancer (median = 197.3 U/mg protein) compared to HR (p = 0.008), benign (p = 0.044) and control (p < 0.001; Fig. 1d). Significantly lower activity of tMMP2 was observed in the control (median = 77.9 U/mg protein) group compared to the HR and benign (p < 0.001, respectively; Fig. 1d).
Concentration and activity of MMP9 in serum
The concentration of aMMP9 ranged from a mean of ∼12 pg/mg protein in the HR donors to a mean of ∼15 pg/mg protein in the control donors and the level of tMMP9 ranged from a mean of ∼1,885 pg/mg protein in patients with BD to a mean of ∼2,534 pg/mg protein in BC patients. Concentration of aMMP9 was significantly higher in the control group (median = 13.9 pg/mg protein) compared to the HR (p = 0.005) and benign (p < 0.001) groups (Fig. 2a) and cancer patients displayed significantly higher concentrations of aMMP9 than the benign group (p = 0.002). No significant differences in the concentration of aMMP9 were observed between HR (median = 12.2 pg/mg protein) and cancer or benign, and between cancer (median = 13.6 pg/mg protein) and control (Fig. 2a). tMMP9 concentration was not significantly different among groups except between cancer and BD, where the levels were significantly higher in the cancer group compared to the benign group (p < 0.001; Fig. 2b).
The activity of MMP9 ranged from a mean of ∼0.11 U/mg protein in controls to a mean of 0.25 U/mg protein in BC patients, whereas the tMMP9 activity (active-form + proform) ranged from a mean of ∼20 U/mg protein in controls to a mean of ∼41 U/mg protein in BC patients. Activity of MMP9 was lower in the control (median = 0.1 U/mg protein) than benign (p = 0.026) and cancer (p < 0.001; Fig. 2c). tMMP9 activity was higher in cancer (median = 37.5 U/mg protein) than HR (p = 0.007) and benign (p < 0.001) but lower in the control (median = 17.1 U/pg protein) compared to benign (p = 0.002) and cancer (p < 0.001) groups; Fig. 2d).
The overall profile and pair-wise relationship between the concentration and activity of serum MMP2 and MMP9 in controls and donors classified as HR or with breast disease and BC are summarized in Table II.
Table II. Pair Wise Comparison Showing the Relationship Between Serum Concentration and Activity of Active and Total MMP2 and MMP9 in Female Donors with Low Risk for Breast Cancer (Controls, CO), High Risk for Breast Cancer (HR), with Benign Disease (BD) and Breast Cancer (BC)
Serum MMP concentration
Serum MMP activity
MMP concentrations and activities were quantitatively determined as described in the methods section. Controls (CO), high risk (HR), benign disease (BD) or breast cancer (BC) cases exhibited statistically significant higher (>) or lower (<) levels in some cases. In other cases, the concentration and activity were not statistically different (=) between two comparisons. p values ≤ 0.05 were considered significant.
CO > BD
CO > BC
CO > BC
CO > BD
HR > BC
CO < HR
HR > BD
HR > BC
HR = CO
HR > BD
BD = BC
BD < BC
CO < HR
CO < HR
CO = BC
CO < BD
CO = BD
CO < BC
HR > BC
HR < BC
HR > BD
BD < BC
BD = BC
HR = BD
CO > HR
CO < BD
CO > BD
CO < BC
CO = BC
CO = HR
HR = BC
HR = BC
HR = BD
HR = BD
BD < BC
BD = BC
CO = BC
CO < BD
CO = BD
CO < BC
CO = HR
CO = HR
HR = BD
HR > BC
HR = BC
HR = BD
BC > BD
BD > BC
Effect of the inclusion or exclusion of ductal carcinoma in situ and atypical hyperplasia
The results based on concentration remained unchanged whether 26 patients with noninvasive BC (ductal carcinoma in situ [DCIS]) or 13 patients with atypical hyperplasia were included or excluded in the cancer or benign groups, respectively. Only a slight statistically significant change was observed (activity relationship) when 26 DCIS and 13 atypical hyperplasia were excluded from the cancer and benign groups. The observed change was a higher activity of aMMP9 in the cancer group than benign group (p = 0.044, p was 0.137 before exclusion).
Effect of donor age, tumor stage/grade on serum MMP2 and MMP9 concentration and activity
To test the potential effect of age and disease stage/grade on MMP2 and MMP9 concentration/activity we further (i) stratified the donors into below and above 50 years of age and (ii) grouped patients based on tumor metastasis, grade and stage. A two-way ANOVA with interaction was performed and all data was log-transformed.27 Results indicated that there were no significant interactions of age (p value, 0.21–0.99) and no major effect attributable to age across groups (p = 0.067–0.88, data not shown). In relation to tumor grade, only individuals with well-differentiated tumors had significantly higher aMMP9 (median 0.37 U/mg protein) than those with poorly (p = 0.04) or moderately (p = 0.01) differentiated tumors.
Linear discriminant analysis
Linear discriminant analysis was performed with all the measured variables (concentrations and activity of MMP2 active/total and MMP9 active/total proteins) to determine which combination (of variables) best discriminates the study groups/categories or better reveal group specific features. When all 4 groups (control, HR, benign and cancer) were considered, the first discriminant function differentiated HR/control individuals from cancer/benign patients on the basis of lower activity of tMMP2 and higher activity of aMMP2. The discriminant function coefficients are shown in Table III (Group 1). These function scores are displayed on a scatter plot (Fig. 3a). Overall, 56% (192/343) of original grouped cases for Group 1 (control, HR, benign and cancer) were correctly classified with the benign group having the highest correct classification (78%, 117/150), (Table IV, Group 1). When the controls were excluded, the first discriminant function also identified HR individuals as those more likely to have high activity of aMMP2 (Table IV, Group 2; Fig. 3b). Overall correct classification for Group 2 (HR, benign and cancer) increased to 64.5% (182/282) (Table IV, Group 2) with 86% correct classification for benign (Table IV, Group 2). Cross validation of these results showed 53% (183/343) for overall correct classification of original grouped cases (Group 1: control, HR, benign and cancer); again, the benign group had the highest correct classification (74%, 111/150; Table V). Overall classification for the cross validation without the controls (Group 2: HR, benign and cancer) increased to 63% (178/282; Table V). Details of the cross-validation classification and original results are as shown (Supplemental Table 1a and b).
Table III. Standardized Canonical Discriminant Function Coefficients (CDFC) using (Group 1) all Four Groups (Control, High Risk,Benign and Cancer) and (Group 2) 3 Noncontrol Groups (High Risk, Benign and Cancer)
log MMP2 active (U/mg protein)
log MMP2 total (U/mg protein)
log MMP9 active (U/mg protein)
log MMP9 total (U/mg protein)
log MMP2 active (pg/mg protein)
log MMP2 total (pg/mg protein)
log MMP9 active (pg/mg protein)
log MMP9 total (pg/mg protein)
Functions at group centroids
Table IV. Classification Results for High Risk, Benign, Cancer and Control Groups (Group 1) and High Risk, Benign and Cancer Groups (Group 2) Based on Linear Discriminant Analysis of Serum MMP2 and MMP9 Measurements
Table V. Cross-Validated Results for High Risk, Benign, Cancer and Control Groups (Group 1) and High Risk, Benign and Cancer Groups (Group 2) Based on Linear Discriminant Analysis of Serum MMP2 and MMP9 Measurements
In the United States, BC is still the most common cause of death in women age 40–49.28 Because current screening methods (i.e. mammography, self breast examination and clinical breast examination) do not provide sufficient early diagnosis in all patients, new biomarkers that complement existing screening tools need to be developed. An ideal molecular marker should correlate well with disease and be easily monitored, for example by a simple blood test. So far no single biomarker with these attributes has been found.29
In our study, we examined the serum levels of MMP2 and MMP9, two of the most widely studied MMPs. Previous studies by our group and others have provided evidence of their possible prognostic significance in BC development and progression.10, 15, 19 Only few reports are available on the activity of circulating MMP2 and MMP9 in cancers in general and BC in particular. Most studies have focused on measuring MMP by ELISA (enzyme linked immunosorbent assay), which employs specific polyclonal or monoclonal antibodies. However, since MMPs are activated through an activation network and occur in biological fluids as inactive precursors, active enzymes or enzyme inhibitor complexes, it will be important to understand how the concentration and activity of these different forms vary and correlate with disease status. Although zymography has been utilized for discriminating between active and pMMP2/9, this technique is known to be very time consuming, only semiquantitative at best and unlikely to find large-scale adoption in the clinic. The use of a more quantitative method that is amenable to automation, such as the one employed in our study, will be more appropriate for transition into clinical practice. The commercial assay employed for our study does not measure activity if the MMP is inhibited by TIMP (tissue inhibitors of metalloproteinase); therefore, the activation process and activity values obtained are not influenced by increased TIMP production.19
To utilize circulating MMPs as blood-based breast disease markers, we must first accurately determine if and how circulating levels correlate with different forms of breast malignancies and also establish the best medium for its measurement. In our present study, we quantitatively measured concentration and activity of MMP2 and MMP9 in serum and observed detectable intergroup differences in the concentration and activity of total and aMMP2/9 in serum of donors grouped as control, HR, benign and cancer. The HR group had significantly higher levels of aMMP2 than benign and cancer groups (p < 0.001, respectively). Notably, the aMMP2 levels of the HR group were not different from that of the control group (Fig. 1b, Table II), an indication that the levels of aMMP2 are apparently similar between the female individuals classified with the Gail Model as “high risk” and “low risk” for development of BC and as such may not be useful for distinguishing between these two groups.
Concentration of aMMP2 was significantly higher in the controls compared to the benign and cancer groups (p > 0.001, respectively). This observation is noteworthy because the assumption is generally that the concentration of MMP2 would be higher in the cancer group where cancer associated tissue remodeling is expected to be higher. Our data suggests that aMMP2 is associated with normal tissue remodeling, an indication that it is the balance between the active and total forms of MMP that maybe important in BC classification. aMMP2 concentration was similar between the cancer and benign group (Fig. 1b, Table II), suggesting that tissue remodeling mediated by aMMP2 are similar in cancer and benign conditions. Interestingly, aMMP2 levels were also similar in control and HR, but significantly different between control and BC, the conditions at the opposite ends of the cancer spectrum. The observed similarity of aMMP2 concentration between the cancer and benign group suggests that MMP2 may not be actively involved in the complex events that initiate or promote progression from the benign to the cancer condition. The cancer group displayed significantly higher concentrations of both active and tMMP9 compared to the benign group (p = 0.002 and p < 0.001 respectively; Table II, Figs. 2a and 2b), suggesting that unlike MMP2, MMP9 plays a more active role in the events that characterize the benign and cancer conditions.
Activity of tMMP2, on the other hand, was significantly lower in the controls than HR, benign and cancer (p < 0.001, respectively) and significantly higher in the cancer than HR (p = 0.008) and benign group (p = 0.044). The lower tMMP2 activity that apparently occurs in controls compared to the other groups may have important biological significance and possibly form the basis for distinguishing between the control and the other breast disease categories (HR, benign and cancer). Such distinction could have significant impact on early detection, monitoring and management of BC and could lead to the establishment of proper intervention measures.30
Our results indicate that the concentration and activity of MMP2/9 in serum is progressively different in the different groups, suggesting that changes in MMP level can be used to monitor or predict the onset or progression of BC. But we observe that the inclusion or exclusion of individuals with noninvasive cancer (DCIS) from the cancer group or individuals diagnosed with atypical hyperplasia from the benign group did not significantly impact the classification accuracy and this may suggest that although the serum concentration and activity of MMP2/9 may be different between categories, this may not be sufficient to permit intracategory differences. The observation that patients with BC who had well-differentiated tumors had higher MMP9 activity compared to those with moderately or poorly differentiated tumors needs to be validated to determine its real significance in the classification of patients based on metastatic potential. However, the observation that increased activity of MMP9 is associated with BC has been previously reported by other investigators.15
The caveat is that some of our observed differences (between and within groups) do not reflect or correlate with an expected transition from low risk/HR to BD and/or BC in all the cases (Supplemental Fig. 1). Only the significant differences observed for the concentration of aMMP2 and the activities of (i) tMMP2 and (ii) tMMP9 (Table I) reflect the trend that would be expected if it is supposed that normal individuals will progress from HR to BD or BC. The likely reason for this scenario is unclear, but it may suggest that (i) a larger and more evenly distributed sample size may be needed to uncover the true significance of these observations and (ii) because of the likely influence of idiopathic differences, the activity and concentration of MMP2 and MMP9 alone may not be sufficient to accurately classify all individuals into the 4 main groups studied.
An important aspect of our study is the use of the Gail model26 to identify and classify donors into the HR and low risk groups. The Gail model is a mathematical model that utilizes epidemiological risk factors (age, cancer family history, etc.) to categorize women as low, intermediate or HR for BC.26 There are limitations to the use of the Gail model and other similar models31, 32, 33 because they are mainly based on subjective measures and may be influenced by idiopathic differences between individuals. The accuracy of any model will depend on the identification of strongly linked risk factors, accuracy of estimation of risk factor effects in specific populations and accurate knowledge of an individual's medical, familial and demographic history. While these issues may impact the Gail model, and indeed our results, it is still the most widely used, well-understood and validated model for breast-related malignancies.34 Considering that an elevated Gail risk is an accepted and validated epidemiological indicator of increased BC risk35 it provided us with a clinically accepted basis for stratification of donors. By including individuals at HR of developing BC based on the Gail Model, we were able to determine whether these individuals can be distinguished from others with low risk or with BD/BC on the basis of the concentration or activity of MMP2/9.
Although there have been previous attempts at correlating epidemiological risk with biology,36, 37 our present study provides the first attempt at correlating epidemiological risk with the concentration and activity of MMP2 and MMP9 in serum. Studies by Tice et al.36 demonstrate that adding nipple aspirate fluid (NAF) cytology results to the predictor variables used to calculate the Gail risk for women modestly improved the discriminatory accuracy of the model. They concluded that NAF cytology may improve prediction models of BC incidence, particularly for HR women.36 Molecular differences have been observed in mathematically determined high, low and intermediate risk for BC based on the observation that promoter methylation of cyclin D2 occurred almost exclusively in malignant tumor samples, whereas promoter methylation of APC, RAR/β2 and RASSF1A occurred more frequently in benign samples from HR women than from low or intermediate risk women as determined by the Gail model.37 Our data as well as those from other laboratories36, 37 now provide information on the biological markers that correlate with epidemiological risk. The combination of molecular signatures with risk models, such as the Gail model, will improve prediction of the near-term risk of BC, and this may become clinically useful in the areas of risk management and early breast disease detection.
Linear discriminant analysis is a powerful and widely accepted method for determining which combination of variables best discriminates between groups/categories or that better reveals group specific/distinguishing features.19 Given a nominal variable and several quantitative attributes, the analysis extracts a linear combination of the quantitative variables (canonical variables) that capture between class variations. We employed LDA to determine which combination of the measured MMP variables discriminates between 2 or more of the 4 patient/study categories: BC, BD, HR and control. Linear discriminant analysis has been used in classifying tumors based on a small number of variables.38 In our study, 86% of the benign individuals were correctly classified, but overall classification efficiency was 64.5%, probably because of the small sample size. If the overall classification does not improve after increasing the sample size or analyzing the data with novel algorithms, then the achieved classification efficiency may not be robust enough and suitable for adoption in the clinic. This will mean that only about 65% of female individuals may be classified accurately in the 4 categories on the basis of total and active levels of MMP2 and MMP9.
The best medium (plasma or serum) for use for quantitative measurement of MMP2/9 concentration and activity is still debatable. Availability of this information will help the development of standardized and reproducible protocols for the measurement of MMP2/9. We observed that the average concentration and activity of MMP2 or MMP9 were higher in serum compared to plasma in more than 66% of the cases studied (Supplemental data, Figs. 2a and b). However, similarities in the relationship between individual categories did exist between our earlier study19 and the present one.
Our results are encouraging because they appear to support earlier studies that indicate that circulating MMP2 or MMP9 could be useful in staging/prognosis10, 15, 39 of breast and other cancers. It also demonstrates that a correlation may exist between mathematical risk model assessment (i.e. Gail model) and molecular changes (i.e. MMP2/9), consistent with previous findings involving NAF cytology36 and tumor DNA methylation analysis.37 A combination of risk models with molecular data may improve prediction of near-risk for BC.36, 37 Taken together, our studies indicate that measurement of MMP2 and MMP9 in plasma and serum is informative because both MMPs effectively partition in similar trends/proportions in plasma and serum.
While issues related to preanalytical conditions and methodologies23, 40 still remain unresolved, before fully establishing whether MMP2/9 can be exploited as circulating BC biomarkers, this article has addressed some of the important issues including (i) the suitability of serum, (ii) the usefulness of simultaneous measurements of concentration and activity and (iii) the ability to classify sub categories. The full potentials of these markers may need to be determined using a larger sample size study prior to a detailed prospective clinical assessment. Additionally, since MMP2 and MMP9 seem to partition relatively well in serum and plasma, it will be interesting to simultaneously measure MMP2 and MMP9 in plasma and serum of each donor to determine if the combined measurement will improve classification accuracy.
This work was performed under the auspices of the Clinical Breast Care Project (CBCP).