|Home | About | Journals | Submit | Contact Us | Français|
The cell surface protease membrane-type serine protease 1 [MT-SP1]/matriptase is often upregulated in epithelial cancers. A dysregulation in MT-SP1/matriptase levels with respect to its cognate inhibitor hepatocyte growth factor activator inhibitor-1 [HAI-1] suggests that it is an increase in proteolytic activity that significantly differentiates malignant from normal tissue. Here we use antibodies to demonstrate that MT-SP1 is active on cancer cells and that this activity may be targeted for tumor detection in vivo. A proteolytic activity assay with the MT-SP1-positive human cancer cell lines MCF-7, HT29, LNCaP, and MDA-MB-468 showed that the antibodies, which inhibit recombinant catalytic MT-SP1, are able to bind and inhibit the full-length enzyme. The same experiment with the MT-SP1-negative breast cancer cell lines MDA-MB-231, COLO 320DM and HT1080 showed no inhibition of proteolysis. Fluorescent microscopy then confirmed localization of labeled antibodies to the surface of MT-SP1-positive cells. To evaluate these antibodies as probes for targeting MT-SP1 activity in vivo, 0.7-2 nanomoles of fluorescently labeled antibodies were administered to xenograft mouse cancer models. The antibodies localized to the MT-SP1-positive MCF-7 and MCF-7/Luc+ tumors (n=3), permitting visualization of MT-SP1 activity. Fluorescence was not observed in MT-SP1-negative MDA-MD-231/Luc+ tumors (n=2), suggesting that MT-SP1 activity is a novel biomarker for epithelial cancer and these antibodies provide a non-invasive method for detecting this activity in vivo.
Molecular imaging of cancer has the potential to facilitate early detection and to provide a more detailed assessment of disease. Currently there are fewer than ten imaging probes approved for use in the clinic, and they are often limited in use (1). Unfortunately, a lack of specificity and/or sensitivity limits most of these probes for use in evaluating patients with previously diagnosed cancers, and the majority are useful only in a small subset of cancers. Therefore the search for new biomarker molecules, and imaging probes with which to detect them, continues.
Proteases are a class of enzymes that shows promise for cancer detection and characterization. Proteolytic processing is necessary in nearly every stage of cancer growth and progression, from angiogenesis to extracellular matrix remodeling, cell-cell signaling and metastasis (2-5). Finely regulated through activation and inhibition, changes in relative levels of proteases or their cognate inhibitors are often associated with cancer, suggesting that it is a dysregulation of proteolysis which contributes to malignant growth (4, 6).
Membrane-type serine protease 1 [MT-SP1], also referred to as matriptase, TADG-15, PRSS14, SNC19, prostamin and gene ST14, is a cell surface protease which, along with its cognate inhibitor hepatocyte growth factor activator inhibitor-1 [HAI-1], is dysregulated in a number of epithelial cancers. This dysregulation has been correlated to cancer stage, and the down regulation of inhibitors indicates that MT-SP1 is likely more active in the disease state (7-20). MT-SP1 has been shown to cleave a number of cancer-promoting substrates and has recently been associated with a pro-metastasic signaling pathway in breast cancer, but its exact role in cancer has not been precisely defined (21-23). The ability to non-invasively monitor active MT-SP1 throughout disease progression would be useful for a better understanding of the role of MT-SP1 in tumor development and as a method for tumor detection.
The use of fluorescently-labeled substrate-based probes in mouse models has confirmed that proteolytic activity is a viable marker for cancer imaging in vivo (24-27). Antibodies provide an alternate approach for targeting biomarkers and have the advantage of being very potent and specific for the target protease. Pharmacokinetics may also be altered by building the epitope recognition region into antibody-derived structures such as single chain variable fragments (scFvs), Fabs or minibodies. Human antibodies are also non-immunogenic and can be functionalized for multiple imaging modalities. Accordingly, many antibody-derived tools for cancer detection and treatment are already FDA-approved for in vivo use (1, 28).
MT-SP1, anchored to the extracellular surface of cancer cells, is localized to malignant tissue and accessible to antibodies, making it an ideal candidate for targeting in vivo. Phage display was used to isolate two antibodies that are selective for the active form of MT-SP1 and inhibit the protease. Kinetic assays with a number of closely related proteases showed no measurable inhibition, indicating that these antibodies are also very selective for this enzyme (unpublished results, 29, 30). Here, we demonstrate that MT-SP1 is active and can be inhibited on the surface of human cancer cells. Antibodies are then fluorescently labeled and used as probes for MT-SP1 activity in xenograft mice, ultimately showing that MT-SP1 is active at the tumor site and that this activity may be used for non-invasive tumor detection.
E2 scFv and Fab and A11 Fab were expressed in E. coli and purified as previously described (29, 31). Diabody was synthesized by complete deletion of the poly-glycine linker region of the scFv using primers (Eurofins MWG Operon, Hunstville, AL): Forward primer is 5′-GGCCAAGGCACCCTGGTGACGGTTAGCTCAGCGGATATCGTGATGACCCAGAGC CCACTGAGCCTGCC-3′ and Reverse is: 5′-GGCAGGCTCAGTGGGCTCTGGGTCATCAC GATATCCGCTGAGCTAACCGTCACCAGGGTGCCTTGGCC-3′. The scFv expression vector was purified from E. coli using Miniprep kit (Qiagen, Valencia, CA) and PCRed with primers using reagents from Stratagene QuikChange kit (Agilent Technologies, La Jolla, CA) – 1 cycle at 97° C for 60 seconds, 20 cycles of 97° C for 60 s, 55° C for 80 s, 68° C for 5 min, and 1 cycle of 68° C for 12 minutes. The PCR product was sequenced to confirm deletion and retransformed into DH12S cells and expressed in the same way as the scFv. MT-SP1 and the inactive zymogen mutant R15A were expressed in E. coli and purified as described (29, 31). A11 IgG was expressed in CHO cells using the V genes from the selected A11 Fab clone and an IgG expression vector as previously described (32).
The KI for A11 IgG and E2 diabody were determined as previously described (30). Antibody was incubated with recombinant MT-SP1 for 4 hours and proteolysis measured through activation of the chromogenic substrate Spectrozyme tPA (America Diagnostica Inc, Stamford, CT). A11 concentration was varied from 0-250 nM, E2 diabody was varied from 3 to 250 pM, and Spectrozyme concentration from 0 - 1mM. KI* values for each concentration of substrate was determined by fitting the data to the equation
where υi is the reaction velocity in the presence of inhibitor and υs is the reaction velocity without inhibitor for a given concentration of substrate. The KI* values were then plotted against substrate concentration to obtain the KI:
The KI of the diabody was determined to be 8 ± 2.3 picomolar, similar to that of the scFv.
scFv, diabody, Fab and IgG were labeled with Alexa Fluor 594 (for microscopy) or Alexa Fluor 680 (for in vivo imaging) (Invitrogen, Carlsbad, CA) according to manufacturer's protocol. Protein was purified from unreacted dye on a Superdex 75 FPLC column (GE Healthcare, Little Chalfont, UK). Degree of labeling was determined using UV/Vis spectrometry as directed in manufacturer's protocol. In fluorescent experiments, concentrations given refer to dye molecules rather than the labeled protein.
Binding curves were obtained as previously described, with the following modifications: A11 Fab immobilization level was ~70 RU, and MT-SP1 and the inactive zymogen MT-SP1 R15A were injected in concentrations from 50 nM to 1 μM at 20 μl/min to minimize mass transfer effects (31). Surface regenerations were performed with 100 mM Glycine pH 2.2, allowing a complete return to baseline. The sensorgram of the reference surface was subtracted from the ligand conjugated surface for each injection.
Human cancer cell lines HT-29, PC-3, MDA-MB-231, MCF-7, MDA-MB-468, HT1080 and COLO 320DM and LNCaP were obtained from the American Type Culture Collection (Manassus, VA) and maintained in the recommended media for less than six months after receipt or resuscitation. MCF-7/Luc+ and MDA-MB-231/Luc+ cells were modified by lentiviral transduction to express Firefly luciferase as previously described (33, 34).
T-75 flasks of 70-90% confluent adherent cells were rinsed in PBS and lifted using Enzyme-Free Cell Dissociation Buffer (Invitrogen). Cells were washed twice in serum-free media and counted, then resuspended in serum-free media and aliquoted into round-bottomed 96-well plates, ranging from 30,000-60,000 cells per well, depending on the cell line. E2 Fab and serum-free media were added for a final volume of 95 μl and final inhibitor concentration of 200 nM. For total inhibition, 5 μl of 25× Complete Inhibitor Cocktail (Roche, Basel, Switzerland), a broad-spectrum protease inhibitor cocktail, in water was added along with 90 μl of serum-free media. After 1-1.5 hour incubation at 37° C, 5% CO2, Spectrofluor-tPA (American Diagnostica Inc.) was added to a final concentration of 500 μM. Fluorescence was measured on a SpectraMax Gemini EM plate reader (MDS Inc., Mississauga, Ontario, Canada) with an excitation/emission wavelengths of 380/460 nm. Fluorescence was measured for one hour or until proteolysis ceased to be linear. Fluorescence was also measured in wells containing only 100 μl of serum-free media to correct for non-proteolytically-mediated substrate hydrolysis. Prior to inhibition assays, these experiments were carried out with 10,000-100,000 cells per well to ensure that the number of cells used was in the range where fluorescence increased linearly with cell number. Activity assays were conducted in sextuplicate.
Glass microscope cover slips were flame-sterilized and placed in 12 well plates. Cells were passaged into these wells and grown to a confluency of 40-90%. 12-16 hours prior to imaging, cells were switched to serum-free media. One hour prior to imaging, fresh serum–free media was added with enough AlexaFluor 594-labeled E2 scFv to obtain a final fluorphore concentration of 300 nM. Cells were returned to the incubator for 1-1.5 hours after which slides were removed, rinsed in PBS, and immediately imaged on a Nikon Eclipse E800 fluorescent microscope outfitted with a G-2E/C filter combination (Nikon, Tokyo, Japan). All cells were imaged within 10 minutes of removal from incubator. For the HAI-1 blocking experiment, recombinant human HAI-1 (R&D Systems, Minneapolis, MN) was diluted to 1 μM in PBS and added along with fresh serum-free media to a final concentration of 200 nM and cells incubated under normal conditions for three hours. At this time, fluorescently labeled E2 scFv was added to the blocked cells, as well as to unblocked cells, for a final dye concentration of 100 nM. Cells were incubated for another hour and then rinsed in PBS and imaged. Images were collected in the .zvi format with a Zeiss Axiocam using Axiovision Software (Carl Zeiss, Oberkochen, Germany). Images were converted to jpegs and corrected for camera condensation artifacts, as determined by blank (no cell) images and using Adobe Photoshop software (Supplementary Data Figure 1).
Six to seven week-old female nude [NCR nu/nu athymic female mice, 6-7 weeks old; (Taconic Farms, Hudson, NY)] were subcutaneously [s.c.] implanted at the base of tail with 60-day sustained release 0.72-mg 17h-estradiol pellets (Innovative Research of America, Inc., Sarasota, FL). Two days later, 1×107 MCF-7 or MCF-7/Luc+ breast cancer cells were implanted s.c. in the upper back area as a 0.2-mL suspension. Tumor growth was measured by caliper along the largest (length) and smallest (width) axes twice a week. Tumor volumes were calculated using the following formula (20): tumor volume = [(length) × (width) × (width)]/2. At approx 30 days post-tumor implantation (mean tumor volume, 600 mm3), animals were imaged as described below.
Six to seven week-old female nude [NCR nu/nu athymic female mice, 6-7 weeks old; (Taconic Farms)] were anesthetized with avertin. A midline incision along the abdomen followed by angled bilateral incision between the number 4 and 5 nipples was then made to expose the number 4 mammary gland. MDA-MB-231/Luc+ cells (2×105 cell suspended in 40 μl serum-free media) were injected directly into the surgically exposed mammary fat pads with a 500 μl tuberculin syringe. The abdominal skin flaps were closed with wounds clips. Wound clips were removed 14 days post surgery. At approximately 45 days post-tumor implantation (mean tumor volume, 400 mm3), animals were imaged as described below.
Mice were fed an alfalfa-free diet of Harlan Teklad Global 2018 rodent feed (Indianapolis, IN) to minimize background fluorescence. Mice were anesthetized with 1.5-2% isoflurane. Alexa Fluor 680-labeled A11 IgG, E2 Fab, and E2 diabody were injected to the tail veins with total amount of injected dye ranging from 0.5-2 nanomoles. For fluorescent imaging of mice with E2 diabody and Fab, two MCF-7 mice were each injected with ~ 2 nanomoles of dye (Fab) and ~ 3 nanomoles of dye (diabody). Images were collected in fluorescent mode on an IVIS 50 using Living Image 2.50.2 software (Caliper Life Sciences, Hopkinton, MA) at set intervals depending on the antibody construct injected. For the IgG studies, two MCF-7 and one MCF-7/Luc+ mice were injected with approximately 2 nanomoles of dye and anesthetized and imaged at regular intervals for 50 hours. Two MDA-MB-231/Luc+ tumor-bearing mice were injected with approximately 0.7-1 nanomoles of dye and imaged in the same manner. In the images presented, region of interest analysis of the entire mouse using Living Image software indicated the relative signal coming from each mouse four hours after injection. Intensity minima and maxima of each set of the presented images were adjusted to compensate for the difference in total signal from the two sets of mice. For bioluminescent images, Luc+ mice were anesthetized as described and intravenously injected with ~ 1 mg of luciferase and imaged in bioluminescent mode on the IVIS 50 with Living Image software at 12-15 minutes post injection. All in vivo studies were performed as directed under institutional approval.
We have developed two antibodies, E2 and A11, which exclusively bind to the active form of MT-SP1 with picomolar affinity (29, 30). As was shown for E2, here we demonstrate that the A11 Fab is selective for active MT-SP1 (Figure 1) (31). Mutating the arginine residue necessary for MT-SP1 autoactivation generates the inactive zymogen MT-SP1 R15A. While binding of active MT-SP1 can be seen at 200 nM, no binding was observed for MT-SP1 R15A at concentrations up to 1 μM, demonstrating selectivity for the active enzyme.
Because they are both selective for active enzyme and share similar mechanisms of inhibition and potencies against MT-SP1, E2 and A11 are used interchangeably in these experiments, as summarized in Table 1. Although multiple antibody-based constructs are used – scFv, Fab, IgG – they will be referred to throughout as “antibodies,” with the specific construct noted in materials and methods.
Though the antibodies inhibit the catalytic domain of MT-SP1, it has not been shown that the full-length enzyme is active and can be inhibited by the antibodies on human cells. An assay was developed to address this by monitoring whole cell-associated proteolysis. In this assay the P1-Arg specific substrate Spectrofluor-tPA was added to cells in serum-free media and proteolysis measured over time. When the MT-SP1-specific (α-MT-SP1) antibodies were added to these cells, any decrease in the rate of proteolysis could be attributed to this particular enzyme, confirming that MT-SP1 is active and that the antibodies inhibit the full-length protease. Figure 2 shows the results of this assay as performed with seven different human cancer cells, chosen based on previously published mRNA expression data (35). The MT-SP1-positive cells showed a 42-89% decrease in proteolysis of the P1-Arg substrate upon the addition of MT-SP1-specific antibody-based inhibitors, while the MT-SP1-negative cells showed no significant decrease in activity. Additionally, the proteolysis was completely inhibited in the presence of a broad spectrum inhibitor cocktail in all cell lines. This assay demonstrates two key points: first, that uninhibited, active MT-SP1 is present on the surface of these cancer cells and second, our antibodies are able to bind and inhibit the full-length protease.
For an antibody to be useful as a probe for molecular imaging of MT-SP1 activity, it must be efficiently labeled without destroying its ability to bind to the enzyme. For fluorescent detection, commercially available dye-succinimidyl ester conjugates were used to nonspecifically label the antibodies via accessible lysines. Based on structural data of both A11 and E2 Fabs bound to recombinant MT-SP1, the labeling of free lysines should not interfere with enzyme binding (Supplementary Data Figure 2). Depending on the construct – scFv, diabody, Fab or IgG – we were able to conjugate an average of 1-6 dye molecules per protein. Inhibition assays with recombinant protein showed minor (0-5 fold) increases in IC50 values, and given the high potency of these inhibitors, such an increase was not considered a significant barrier to antibody binding. To test the functionality of the scFv against full length protein, human cancer cells were incubated with the labeled MT-SP1 antibodies and fluorescently imaged to look for association with the membranes of these cells (Figure 3). Three MT-SP1-positive cell lines - HT-29, MCF-7 and LNCaP - show labeling with the fluorescent antibodies, while the negative control lines MDA-MB-231 and COLO 320 DM do not. Interesting to note is the difference in the nature of this labeling; the HT-29 signal appears to come evenly from the membrane of the cell while the signal on LNCaP and MCF-7 cells is more punctate. Whether this is a result of MT-SP1 congregating on the surface, or an internalization of the antibody/enzyme complex has not been determined. Nevertheless, these results show that these labeled antibodies serve as imaging probes to selectively target MT-SP1-positive cells ex vivo.
MT-SP1 is putatively present on the surface of epithelial cells in at least three different isoforms – the inactive zymogen, active protease, and HAI-1-inhibited protease. Our experimental results indicate that the antibodies are binding to and inhibiting active protease on the cell surface, but the question remains whether the antibody is displacing cognate inhibitor upon binding. In this case the signal would not be representative of free, active protease on the cell surface. A study of the rat homologues of MT-SP1 and HAI-1 observed picomolar inhibition, and MT-SP1 which has been shed from the cell surface is found only in complex with its cognate inhibitor, suggesting extremely tight binding in vivo (36). To test for HAI-1 displacement by these antibodies, the immunofluorescent cell labeling was carried out with HT-29 cells which were first incubated with recombinant HAI-1 (Figure 3c & d). Cells which were treated with HAI-1 before the addition of fluorescently labeled scFv showed much lower membrane-associated fluorescence than those which were not treated. This data shows that our antibodies only minimally displace HAI-1 under the conditions used in the fluorescent and the cell activity experiments, indicating that the majority of the signal comes from free active MT-SP1.
Having successfully labeled active MT-SP1 in cell culture, the antibodies were evaluated as probes for MT-SP1 activity in vivo. Based on cell culture data, xenograft mouse models were generated using MCF-7, MCF-7/Luc+ and MDA-MB-231/Luc+ breast cancer cell lines. The “Luc+” designation indicates that the cells have been transfected with an imaging vector which includes the gene for Firefly luciferase so that the tumor can be imaged via bioluminescent detection independent of MT-SP1-based fluorescence detection (34). These mouse models were injected with fluorescently labeled diabody, Fab or IgG, and imaged for up to 50 hours to assess biodistribution of antibodies and any tumor localization. The anti-MT-SP1 diabody and Fab localized to the tumor in MCF-7 xenograft mice, but failed to achieve high tumor/background contrast due to high levels of signal retained in the excretory system (Supplementary Data Figure 3). The IgG, however, localized to the tumor in both MCF-7 and MCF-7/Luc+ mice and remained so until free protein was cleared, achieving excellent tumor to background contrast by 50 hours (Figure 4a and Supplementary Data Figure 4). Similar injections into MDA-MB-231/Luc+ tumor-bearing mice showed no tumor associated signal over the same time period (Figure 4b). Luciferin administered to these mice generated a tumor-specific signal, validating both the presence of the MCF-7/Luc+ and MDA-MB-231/Luc+ cells at this location and sufficient vasculature to deliver the labeled antibodies to the tumor. These experiments indicate that MT-SP1 is active in the tumors which are positive for MT-SP1 expression, and this activity can be targeted in vivo for non-invasive cancer imaging using the antibodies as probes.
Proteases have long been recognized as cancer biomarkers, but based on indications that it is activity which is important for cancer progression, perhaps a more direct strategy for disease assessment is to measure the active form of the enzyme. The approach described here, from phage display to in vivo imaging, could be readily applied to any number of cancer-associated proteases, particularly those anchored to the cell membrane, to identify additional sites for targeting tumors in vivo.
The tools described here may have many uses in studying MT-SP1 both in the laboratory and the clinic. Though immunohistochemistry and expression data have strongly connected MT-SP1 to cancer, these experiments have focused on overall protein levels rather than the amount of proteolytic activity. The antibody-mediated targeting of active MT-SP1 described here may provide a new way to measure protease activity and connect it with specific downstream effects both in cell culture and in mouse models, such as the recently proposed pro-metastatic signaling pathway involving MT-SP1, the growth factor MSP and its receptor RON (22, 23).
These antibodies may also prove useful for studying the biology of MT-SP1 at the cellular level. Recent results have shown that HAI-1 recycles between the membranes of polarized epithelial cells, and it is proposed that MT-SP1 may be transported between membranes in a similar fashion, perhaps in complex with HAI-1 (37). Though further studies are necessary, it is possible that the punctuate fluorescence observed when the labeled antibodies are added to unfixed MCF7 and LNCaP cells is a result of internalization of antibody-bound enzyme, indicating that MT-SP1 is indeed capable of reentering the cell after it is exocytosed to the extracellular membrane.
MT-SP1 detection and inhibition may also find use in the clinic. Small molecule inhibitors of MT-SP1 have shown that in vivo inhibition has potential in anti-cancer therapy (38). Antibodies could achieve similar results while allowing for more specific targeting and inhibition, minimizing toxicity. As an imaging biomarker, MT-SP1 expression has been correlated with cancer stage and/or subtype, and a non-invasive method for assessing the activity in the malignant tissue may aid in tumor classification and potential therapeutic design. Independent of the utility of MT-SP1 inhibition in cancer therapy, it can be used as an imaging biomarker to assess tumor margin or monitor tumor response to alternative therapies in MT-SP1-positive tumors.
Previous studies connecting MT-SP1 and related type-two transmembrane serine proteases to cancer have suggested their potential as biomarkers (39). This work demonstrates that MT-SP1 activity is present in tumors and can be targeted and inhibited on the surface of cancer cells using specific antibodies, validating an approach that can be used to target cell-surface proteolysis for cancer assessment and intervention.
The authors thank Lily Darragh for technical consultation and Dr. Mark Moasser for useful discussions and critical reading of the manuscript.
This research was funded by the National Science Foundation Graduate Research Fellowship (MRD), and NIH CA72006 (CJF, CSC) and NIH grants CA108462-04 (ELS).
The authors declare no competing commercial interest with the work presented here.