Concluding remarks

Authors


Dear Sir,

We read with interest the “Letter to the Editor” IJC-06-2595 by Mannello and Tonti titled “Gelatinase concentrations and zymographic profiles in human breast cancer: matrix metalloproteinases circulating in plasma are better markers for the subclassification and early prediction of cancer: the coagulation/fibrinolysis pathways alter the release, activation and recovery of different gelatinases in serum.” We would like to comment on this letter since it specifically refers to our article “Circulating MMP2 and MMP9 in breast cancer—potential role in classification of patients into low risk, high risk, benign and breast cancer categories.”1 Mannello and Tonti's letter state “We read with concern their article which described that total, active and activity of MMP were significantly increased in serum of cancer patients with respect to benign disease and control group, concluding that serum may classify patients with breast disease.” We believe that Mannello and Tonti's evaluation of our study is out of context and their interpretation of our conclusion is not consistent with the message in our paper. We believe that MMP in serum still holds promise as a biomarker and the real or perceived problems associated with the use of serum can be largely mitigated by standardization of protocols, good sample handling, storage and preparation and good experimental design. The higher levels of MMPs in serum could actually be an advantage in terms of providing levels that are far above background signals, if it will be possible to accurately and more sensitively quantify the fractional increase that is due to disease. We also recognize and relate to the weaknesses of correlative studies, but it is still a challenge and largely impossible to design biomarker-based classification studies dependent on mechanisms because of the limited and/or incomplete or even absence of information available on all the mechanisms that may be important. Coagulation and fibrinolysis apparently account for increased MMP9 levels in serum but claiming that this single phenomenon and mechanism dictates that (i) plasma is better than serum, (ii) serum should not be evaluated further and (iii) journals should not publish papers based on serum MMPs is rather myopic.

Before addressing specific issues in Mannello and Tonti's concerns we would first like to provide a brief background on the study in question. The aim of the study was to determine if serum concentration and/or activity are distinct in female donors categorized as low risk (control) or high risk for developing breast cancer (based on the Gail assessment), or pathologically diagnosed with benign disease and breast cancer.1 This is a follow-up study to our earlier study using plasma from a comparable population.2 Critical to the success of MMP measurement is the effect of preanalytical conditions that may influence the detected MMP levels of serum or plasma.3 In designing our plasma and subsequently serum experiments, we were very much aware of the arguments in favor and against the measurement of MMPs in plasma and serum, and the challenge in interpreting the data.3–5 We therefore designed both plasma and serum studies to ensure that data obtained from both can be compared. The details of our sample collection methods are presented in a publication, which describes the comprehensive sample collection and tissue-banking endeavor of the clinical breast care project (CBCP) of Windber Research Institute, Windber PA and Walter Reed Army Medical Center, Washington, DC6 to which readers were referred in our MMP publications. We used state-of-the-art tools and standardized protocols that were designed to mitigate or in the worst scenario standardize errors attributable to differences due to preanalytical factors. All samples (serum and plasma) were collected from all donors using a standardized CBCP protocol. Some of the study participants donated both serum and plasma thereby allowing us to compare the profile of MMPs in serum and plasma. For the serum study, samples were collected into tubes without anticoagulant, processed within 3 hr of collection and stored in 1-ml aliquots in freezers to avoid repeated freeze-thaw cycles. Another important aspect of our study that requires special note is the total number of subjects enrolled, the variables measured and the statistical analysis employed to determine if the differences we observed were statistically significant, and hence not due to a random occurrence. We used serum from 345 female donors, consisting of 61 controls (with no breast disease), 46 high risk, 150 benign disease and 88 breast cancer patients. In each sample, we quantitatively measured: (i) concentration and (ii) activity of total (active + activatable proenzyme forms) and active MMP2 and MMP9. Thus a total of 8 variables were quantitatively measured and used for classifying the donors into 4 categories (low-risk, high-risk, benign disease, breast cancer). Nonparametric tests (Kruskal-Wallis and Mann- Whitney) were employed to determine statistical significance of the differences in concentration (pg/ml protein) and activity (U/mg protein) of serum MMP2 and MMP9 between the study groups. p-values ≤0.05 were considered statistically significant. The sample size was sufficient to detect differences of ∼67% between groups based on the measured variables with 80% power. Two independent bioinformatics tools confirmed the findings that some of the donors can be accurately classified based on the MMP2 and MMP9 variables measured. These important aspects of the study must be taken into consideration before a valid conclusion can be drawn on the usefulness of the data and results obtained.

Our study demonstrated among others, significantly increased MMP2 concentration and activity in controls compared with benign and breast disease and significantly increased concentration and activity of MMP9 and MMP2 in breast cancer compared with benign disease. Mannello and Tonti's letter focuses on the MMP9 and the biochemical and molecular mechanisms that may partly explain why higher levels of MMP9 may be detectable in serum as compared with plasma. They present a zymogram from a single patient showing higher MMP9 in serum compared with plasma and nipple aspirate fluid, a cartoon depicting the role of coagulation and fibrinolysis factors in increased serum MMP9 levels, and cited a number of articles that state coagulation/fibrinolysis may lead to increased MMP9 levels in serum and subsequently conclude that plasma is better than serum. On the basis of the aims of the study and data obtained under the experimental conditions described, we stand by the results published. Prior to our publication, the potential role of serum MMPs as prognostic markers in breast cancer,7, 8 head and neck cancers9 and ovarian cancer10 has been demonstrated. It is apparent that MMP levels although higher in serum is still different between individual patient groups to allow patient classification to some degree. In dealing with this situation, the general recommendation is that researchers should provide details of preanalytical processes to allow appropriate data interpretation11, 12 and methods be optimized for specific applications.11, 13 This is because there may not be a general rule for all cases based on the associated variables, which include: measurement of MMP2, MMP9 or other MMPs, measurement of the latent or active forms and measurement of concentration or activity of the MMP. The technological platform utilized constitutes another variable and these include the use of zymography, ELISA or immunoassays. We presented the details of our study design to avoid misinterpretation, and called attention to important caveats related to some of the observed relationship between the groups we studied. In these cases, our data did not reflect what you would expect if it is assumed that MMP2 and MMP9 are associated with the events that lead to the transition of a donor from low risk/high risk to benign disease and/or breast cancer. Obviously people diagnosed with breast cancer may not always be at high risk or have been previously diagnosed with benign breast disease prior to the onset of breast cancer. Mannello and Tonti's statements “…serious preanalytical flaw in this other wise well-conceived study” and “…undermines confidence in the author's conclusion…” suggests that they missed the point and are viewing our paper out of context. Research reports are expected to provide all the information observed to allow objective reviews by peers. Mannello and Tonti provide a single zymogram and a cartoon related to coagulation/fibrinolysis as evidence to reject a carefully designed study that measured 8 variables in 345 subjects. The references they have cited to support their letter can not be directly compared with our study, thereby begging the questions; (i) how many breast cancer and normal individuals do they have data on serum MMP9, MMP2 or other MMPs and (ii) How many MMP variables have they measured from a population of normal and breast cancer patients?

Regarding the statement “The conclusion of Somiari et al., about usefulness of MMP in serum strongly contrasts with growing evidence that plasma MMP9 is considered better marker for early diagnosis, progression and prognosis of breast cancer” and “Although these articles unequivocally demonstrate the “artificially” higher levels of serum MMPs with respect to plasma, manuscripts ignore these statements and continue to be published on measurement of serum MMPs” is unfortunate. We would like to point out that our results represent data generated and analyzed under stringent conditions and data analyses were performed with multiple algorithms. Our study demonstrates that in spite of the “artificially higher MMPs in serum,” simultaneous measurement of concentration and activity of total and active forms of serum MMP2 and MMP9, allows accurate classification of some individuals into control, high risk, benign and breast cancer. The overall predictive power (that is the ability to correctly classify the 345 subjects into 4 defined categories) we obtained was 56%. The sensitivity and specificity for predicting the likelihood of a donor being in the breast cancer group compared with the control group (no breast disease) was 78 and 89% respectively. While there may be questions and issues associated with the use of serum for measurement of MMPs because of their generally higher levels,14, 15 and the effects of preanalytical processing,3–5 our results strongly suggest that there are still detectable differences in the MMP levels in serum that are apparently sufficient to allow distinction between different breast disease categories with statistical confidence. Our paper is important because it represents one of the few (if not the only one) to measure 4 MMP2 and 4 MMP9 variables in >300 serum samples collected from a population in which plasma has been analyzed. Since the serum and plasma studies employed similar populations and experimental designs/protocols, it is possible to compare the data obtained and make a more objective evaluation of the potential role of plasma and serum. Both plasma and serum studies revealed common relationships as follows: both serum and plasma data showed that the control groups had significantly higher levels of active MMP2 than the benign and cancer patients and the cancer patients had higher levels of total MMP9 than the benign patients. Additionally, total MMP2 activity was significantly lower in the control groups than high risk, benign and cancer while activity of MMP9 was significantly lower in the control than benign and cancer in both studies. These observations are summarized in Table I. Insights gained by critical evaluation of the plasma and serum data have led us to ask the question “will the classification efficiency improve if MMP2 and MMP9 are simultaneously measured in serum and plasma of patients?” Interestingly, our data, which is currently being prepared for publication, displays classification efficiency that is better than what we obtained independently for plasma or serum. This suggests that the fractions of MMP2 and MMP9 in serum are important variables.

Table I. All Similar Pairwise Relationships Observed for Plasma and Serum Studies
 p value
  1. MMP concentrations and activities were quantitatively determined in controls (CO), high risk (HR), benign disease (BD) or breast cancer (BC) cases. Similar relationships (statistically significant higher (>) or lower (<) levels) were observed in plasma and serum. p-values ≤£ 0.05 were considered significant.

Plasma MMP 
Concentration 
MMP2 active 
CO > BN0.021
CO > CA0.033
MMP9 total 
CA > BN0.030
Activity 
MMP2 total 
CO < HR<0.001
CO < BD<0.001
CO < CA<0.001
MMP9 active 
CO < BN<0.001
CO < CA<0.001
Serum MMP 
Concentration 
MMP2 active 
CO > BN<0.001
CO > CA<0.001
MMP9 total 
CA > BN<0.001
Activity 
MMP2 total 
CO < HR<0.001
CO < BD<0.001
CO < CA<0.001
MMP9 active 
CO < BN0.026
CO < CA<0.001

The statement “…the presence of MMP9 forms in higher amounts in serum, is not only related to the disease but mainly linked to coagulation/fibrinolytic pathways…” means that Mannello and Tonti actually accept that part of the MMP measured in serum may be associated with the disease process. Our study may therefore be demonstrating that the fraction that is attributable to the disease is significantly different between: “normal” and “breast cancer” to allow classifications with 78% sensitivity and 89% specificity. Mannello and Tonti's statements “Somiari et al. reported increased serum MMP9 levels and activity in more than 66% of patients studied to the results previously obtained in plasma…” and “…we provide biological mechanisms explaining their (that is, Somiari et al.) contrasting results, reinforcing the concept that plasma is the best and more reliable source for MMP clinical studies” is surprising. Although MMP9 level was generally higher in serum, this was not the situation in all cases. As shown in Figure 1 (this data was presented as a supplementary data to our paper) we did not observe higher serum concentration in 100% of the groups for all the measured variables (or 100% increased serum MMP9 as would be expected based on the idea of Mannello and Tonti). We have not seen published data from any laboratory including Mannello and Tonti's that unequivocally establishes that “plasma is the best and more reliable source for MMP clinical studies” as strongly stated by Mannello and Tonti.

Figure 1.

(a) Concentration of serum and plasma MMP2 and MMP9 in female donors classified as control (n = 61), high risk (HR; n = 46), benign disease (BD; n = 150) and breast disease (BD; n = 88). Serum MMP concentration was higher in 62.5% (10/16) of the groups. (a1) Active MMP2 in serum was higher in all the groups (100%). (a2) Active MMP9 in plasma was higher in 75% (3/4) of the groups. (a3) Total MMP2 in plasma was higher in 75% (3/4) of the groups. (a4) Total MMP9 in serum was higher in all the groups (100%). (b) Serum and plasma MMP2 and MMP9 activity in female donors classified as control (n = 61), high risk (HR; n = 46), benign disease (BD; n = 150) and breast disease (BD; n = 88). Serum MMP activity was higher in 75% (12/16) of the groups. (b1) Active MMP2 in serum was higher in all the groups (100%). (b2) Active MMP9 in serum was higher in 75% (3/4) of the groups. (b3) Total MMP2 in plasma was higher in 75% (3/4) of the groups. (b4) Total MMP9 in serum was higher in all the groups (100%). [Color figure can be viewed in the online issue, which is available at www.interscience.wiley.com.]

In conclusion, we think Mannello and Tonti are unreasonable in their judgment. Our data indicates that even though MMPs are generally higher in serum than plasma quantitative measurements using sensitive techniques and statistical analyses reveals differences that are apparently sufficient to classify over 50% of donors into 4 categories with confidence. The serum- model promises to be more effective in distinguishing “normal” from “cancer” based on the 78% sensitivity and 89% specificity we have achieved so far. We will like to point out that our plasma paper on MMP2 and MMP92 which is highly acclaimed by Mannello and Tonti did not allow significantly better classification of donors into the same 4 categories (as would have been expected) for us to have abandoned the serum study. We note also that Mannello and coworkers have shown that even the use of plasma is not without controversy because the preanalytical effects of different anticoagulants are sometimes controversial.13, 16–18 Our studies based on plasma and serum suggests that, the simultaneous measurement of the concentration and activity of total and active forms of MMP2 and MMP9, using a commercially available immunoassay kit holds promise for breast cancer diagnosis. The results from our laboratory do not change any of the already existing information that has been generated from other laboratories, but rather provides additional data that will stimulate more research, enable the design of more robust and appropriately powered studies necessary to take a serum- and/or plasma-based test into the clinic.

The importance of MMPs as biomarkers will be significant in many cancers and indeed other disease conditions. The generation of independent and collaborative data from multiple laboratories working together to address these issues will have to continue before we can claim complete resolution of any of the sticky issues. Interestingly, Mannello and Tonti do agree to this when they say “…the identification of the best source between plasma and serum for surrogate markers is a matter of strenuous efforts and debate, the scientific interest in the measurement of MMP in plasma/serum continues to mount, enhancing their promising development as reliable biomarkers…” Our studies indicate that MMP differences between breast disease and normal conditions apparently relate not only to the latent forms but also the active forms. This is not surprising because MMPs occur either as inactive proenzymes, active enzymes and inactive enzymes bound by various inhibitors, indicating that multiple biological mechanisms will account for the differences in the levels of MMP detected. The need to accurately measure these differences in serum and/or plasma cannot be over emphasized, and techniques like zymography and models taking only coagulation and fibrinolysis into consideration may not be sufficient to move MMPs as biomarkers into the clinic. Indeed, to achieve classification efficiencies approaching 100% it may be necessary to develop more sensitive tests than are currently available and to consider the sensitivity of the assay, half-life of MMPs in serum and plasma, and the contribution of idiopathic differences of patients, unknown/incompletely elucidated biochemical mechanisms (i.e. different from coagulation/fibrinolysis) and the real physiological status of donors. Although gelatin zymography has been utilized successfully in the discrimination between inactive and active MMPs, it is the development of chromogenic or fluorogenic synthetic peptide substrates19 and assay protocols that have enabled more quantitative and accurate measurements of MMP activity.20 These have led to the development of commercially available immunocapture assay kits for simultaneous measurement of already active or endogenous MMP and proenzyme (activatable) forms.1, 2 Hopefully additional new detection strategies and studies will lead to the standardization and optimization of methods for more accurate and reproducible measurement of MMPs in plasma and/or serum. Currently available information sets the stage for new research that will enable us to move forward with plasma and/or serum MMPs as disease biomarkers. We encourage researchers to be open minded and willing to review all data as they relate to specific experimental designs and disease models rather than prematurely ruling out certain models.

Yours sincerely,

Stella Somiari*, * Windber Research Institute, Windber, PA.

Ancillary