Relaxin Enhances the Collagenolytic Activity and In Vitro Invasiveness by Upregulating Matrix Metalloproteinases in Human Thyroid Carcinoma Cells

The multifunctional heterodimeric peptide hormone relaxin (RLN), a member of the insulin-like superfamily, is an established endocrine factor in the reproductive tract, cardiovascular/neural system, and in cancer biology (1–6). Upregulated in various human cancer tissues, RLN2 contributes to tumor cell proliferation, tissue invasion, and tumor angiogenesis (7–9). Understanding the molecular mechanisms by which RLN2 enhances the tissue invasiveness of tumor cells in patients is of prognostic and therapeutic importance. High serum levels of RLN2 were shown to be associated with increased levels of metastasis in breast cancer patients (10).

The penetration of the basal lamina and remodeling of the extracellular matrix (ECM) are crucial early steps in the process of tumor cell invasion. RLN is known to contribute to ECM remodeling processes in various biological systems by: (i) inhibiting fibroblast proliferation and differentiation (11, 12); (ii) inhibiting collagen synthesis/secretion and tissue deposition in vivo (13–15); (iii) altering the composition and structural integrity of the ECM (16–19); and by (iv) increasing the production/secretion and activity of enzymes that are powerful remodeling factors of the basal lamina and ECM (20, 21). In the human thyroid gland, RLN2 is exclusively detected in thyroid cancer tissues and enhances the production and secretion of the lysosomal acid hydrolases of the cathepsin family which promote penetration of elastin matrices by human thyroid cancer cells (7). Earlier reports supported a role for RLN in ECM remodeling by upregulating cathepsin B and dipeptidyl peptidase I in the mouse pubic symphysis and urokinase and tissue type plasminogen activator in the uterus and cervix of prepubertal gilts (22–24). Members of the family of matrix metalloproteinases (MMP) and their tissue inhibitors of MMP (TIMP1–4) are another group of proteinases responsive to RLN.

The MMP/TIMP system is known to modulate normal (e.g., development, growth, and repair) and pathologic tissue compositions (e.g., fibrosis, tumor spread). Functional imbalance between MMPs and TIMPs can facilitate tumor cell invasion of basal lamina and ECM and enhances tumor angiogenesis (25). The members of the human MMP family of zinc-dependent endopeptidases segregate into 5 major subclasses (collagenases, stromelysins, gelatinases, membrane-type MMPs, and other MMPs; ref. 26). All of these MMPs combined have the ability to degrade essentially any ECM component (27). Apart from the basic structural design of secreted MMPs, which includes an N-terminal signal peptide, a propeptide, and a catalytic domain with a conserved zinc-binding site linked to a hemopexin-like domain, the membrane-type (MT)-MMPs 1–4 possess an additional anchoring transmembrane and cytoplasmic domain C-terminal to the hemopexin-like domain. MMP family members MMP2 and MT1-MMP, for example, not only degrade ECM components but can also cleave and activate growth factors, thereby promoting tumor cell migration/invasion (28, 29) and modulating angiogenesis (30).

MMP2, MT1-MMP, and TIMP2 play a significant role in thyroid cancer. These MMP and TIMP members are frequently upregulated in differentiated thyroid carcinoma and increased MMP2 expression is common in papillary and undifferentiated/anaplastic thyroid cancer, but not in follicular adenoma (31–33). Enhanced production and activation of proMMP2, MT1-MMP, and TIMP2 is positively associated with large thyroid tumor size, increased invasiveness, including lymph node metastasis, and advanced disease stage (31, 34, 35). Human normal thyrocytes and thyroid carcinoma cell lines were shown to express MMP1, 2, and 9, MT1-MMP, and TIMP1–3 (36, 37).

RLN affects the expression and/or secretion of collagenase 1 (MMP1) and 3 (MMP13), gelatinase A (MMP2) and B (MMP9), stromelysin1 (MMP3), and MT1-MMP in the periodontal apparatus and joints (38–40), lung (13), reproductive tissues (24, 41–43), human renal fibroblasts (44), and human breast cancer cells (20). We previously showed that human thyroid cancer tissue is a source of RLN2 and that RLN2 enhances elastin matrix degradation by thyroid cancer cells (7).

In this study, we have determined the tissue distribution of MMP2/gelatinase A and MT-MMP1/MMP14 in human thyroid carcinoma tissues and identified MMP2 to be predominantly present in thyroid tumor tissues. We show that both MMPs are RLN2 target molecules in thyroid carcinoma cells. Of all members of the TIMP family, RLN2 caused no changes in TIMP2 but a significant downregulation of TIMP3 in thyroid cancer cells. The RLN2-mediated upregulation of in vitro invasiveness through collagen matrices coincided with the increased presence of MT1-MMP and MMP2 at the leading migrating edge of invadopodia in human thyroid carcinoma cells. These findings implicate the MMP2–MT1-MMP–TIMP2 system and TIMP3 as potent mediators of RLN2 in promoting tissue invasion and angiogenesis in human thyroid cancer.

One microgram of total RNA was used as a template for first strand cDNA synthesis as described previously (7). PCR amplification conditions had been optimized for each primer pair listed in Table 1.

A total of 106 human thyroid tissues [25 goiter, 30 papillary (PTC), 27 follicular (FTC), and 24 undifferentiated thyroid cancer (UTC)] were obtained at the Department of General, Visceral and Vascular Surgery, Martin Luther University Halle-Wittenberg, Halle (Saale), Germany (Table 2). Tumor size, local tissue invasion, colonization of local lymph nodes, and presence/absence of metastases were determined according to the pTNM tumor staging classification (T = tumor, N = lymph node, and M = metastasis). Tissue specimens were cryopreserved in liquid nitrogen. Human placental tissue sections were used as positive control for the immunodetection of MT1-MMP. The collection of all human tissues studied was approved by the ethical committee at the Martin Luther University, Faculty of Medicine, and all patients gave written consent.

The human follicular thyroid carcinoma cell line FTC-133 and previously described FTC-133 stable transfectants expressing human RLN2 (FTC-133–RLN2) and enhanced green fluorescent protein (EGFP)-expressing FTC-133 negative controls (FTC-133–EGFP; refs. 7, 45) were cultured in Dulbecco's modified Eagle's medium/HAM-F12 medium plus 10% fetal calf serum (FCS; Gibco). Culture medium was changed every 2 to 3 days and transfectants were passaged every 5 to 6 days. Prior to the experiments, cells were either seeded in 25 cm² flasks or in 6-, 24-, or 96-well plates (Techno Plastic Products AG) at 80% confluency. Untransfected FTC-133 at the same concentration were incubated with 100 ng/mL RLN2 (RLN2 was generously provided by Corthera Inc.). All experiments were conducted under serum-free conditions.

For immunohistochemistry, sections of goiter and thyroid carcinoma tissues were investigated for the presence of immunoreactive MMP2, MT1-MMP, and RLN2. Immunohistochemical analysis was done by using the DAKO LSAB Kit (Dako). Tissue sections were fixed with 3% H₂O₂ in cold methanol for 25 minutes, washed in PBS plus 0.1% Tween-20 (PBST), and nonspecific-binding sites were blocked with 10% goat serum in PBST (blocking buffer) for 1 hour at room temperature (RT). Sections were incubated with the primary monoclonal antibodies for MMP2 and MT1-MMP (both Calbiochem) diluted 1:100, or a polyclonal antiserum against RLN2 (Immunodiagnostik) at 1:300 with antibody diluent (Dako) at 4°C overnight (7). Sections were incubated with horseradish peroxidase (HRP)–conjugated secondary antibodies (Dako) for colorimetric detection of target MMPs and RLN2 within the thyroid tissues investigated. Immunostained sections were counterstained with hematoxylin/eosin.

The day before the immunofluorescence experiments, cells were washed twice with 1× PBS and grown in serum-free medium overnight. For exclusive F-actin staining, cells were incubated with phalloidin–AlexaFluor 546 conjugate (1:200; Invitrogen). Immunofluorescent confocal microscopy was done in 12-well plates with insert filters containing polyester membranes with 3.0-μm pore size (Costar). Cells were seeded on the filters and incubated in a 5% CO₂ humidified incubator at 37°C for 3 hours to allow pseudopodia to reach the underside of the filter. Cells on both sides of the filters were rinsed with PBS, fixed with 3.7% paraformaldehyde in PBS for 25 minutes, permeabilized, and incubated with a rabbit polyclonal anti–MT1-MMP (1:50; ab51074; Abcam) and mouse anti-human beta 1 integrin antibody (clone JB1A; 1:1,000; ref. 46) or phalloidin-AlexaFluor 488 conjugate (1:200; Invitrogen). Goat anti-rabbit IgG AlexaFluor 633 and goat anti-mouse AlexaFluor 546 (both Invitrogen) were used for secondary antibodies. Laser-scanning confocal images were acquired by using the Zeiss LSM 710 multispectral confocal microscope and Zen 2009 acquisition software (Carl Zeiss MicroImaging; LLC). Axial plane images were generated from Z-axis fluorescence image stacks acquired at intervals and with pinhole diameters optimal for a plan-apochromatic 63×/1.4 objective (equivalent to 1 AU). Laser lines and spectral windows specific for the selected fluorophores were used as recommended by the manufacturer.

FTC-133–RLN2/FTC-133–EGFP transfectants and untransfected FTC-133 were seeded in 6-well plates at 8 × 10⁴ cells/well and cultured for 24 hours in serum-free medium. Supernatants were collected 72 hours later and centrifuged at 4,000 × g for 30 minutes at 4°C to pellet remaining cells. Cellular proteins were isolated with 2× extraction buffer (125 mmol/L Tris-HCl pH 6, 8; 4% SDS; 20% glycerol; 10% (v/v) mercaptoethanol (ME); 2% (v/v) bromophenol blue) with the addition of a protease inhibitor cocktail (all Sigma-Aldrich). Proteins (30 μg) were run on 12% SDS-PAGE gels and blotted onto Hybond-ECL nitrocellulose membranes (Amersham). Membranes were blocked for 1 hour in PBS blocking buffer containing 0.1% Tween 20, 3% nonfat dry milk, and 2% bovine serum albumin (BSA) for the detection of MMP2 and TBS containing 0.1% Tween 20 and 5% nonfat dry milk for MMP14 and TIMP2 immunodetection. Membranes were incubated for 1 hour in blocking buffer at RT and overnight at 4°C with specific primary antibodies to human TIMP2 (Santa Cruz Biotechnology), MMP2, and MT1-MMP (both Calbiochem). Membranes were washed 3 times in buffered saline followed by a 1 hour incubation at RT with the secondary antibodies, either goat anti-mouse IgG conjugated HRP (for MMP2; Santa Cruz Biotechnology) or goat anti-rabbit conjugated HRP (for TIMP2 and MT1-MMP; Dianova). For lamin A/C detection, MMP14 blots were stripped in 0.2 NaOH for 15 minutes and washed 3 times with TBST prior to blocking with 5% milk in TBST, pH 7.6, for 1 hour at room temperature. Membranes were probed overnight at 4°C with an anti-lamin A/C antibody (1:200; SC-6215; Santa Cruz, CA). Membranes were washed 3 times in blocking buffer followed by 1 hour incubation with a rabbit anti-goat conjugated HRP (1:10,000; SC-2922; Santa Cruz). Blots were visualized by using an ECL kit (Thermo Fisher). Beta-actin detection was used as a marker for equal protein loading per lane. Densitometry data were obtained by using Kodak Digital Science 1D software with the Kodak Digital Science Electrophoresis Documentation and Analysis System 120. The ratio in the levels of each of the probed protein versus β-actin in FTC-133–EGFP cells served as 100% reference for the determination of the relative contents of these proteins in the transfectants.

Pseudopodia assays were conducted with slight modifications as described previously (47). The day before the assay, cells were washed twice with 1× PBS and grown in serum-free medium overnight. Purification of pseudopodia was done in 6-well plates, with insert filters containing 1.0 μm porous track-etched polyethylene terephthalate membranes (BD Biosciences). Each filter was seeded with 7.5 × 10⁵ cells and incubated in a 5% CO₂ humidified incubator at 37°C for 1 hour to allow the tips of pseudopodia to reach the underside of the filter. Pseudopodia were isolated by dipping the underside of each filter in 100 μL of lysis buffer (50 mmol/L Tris-HCl, pH 7.4, 3 mmol/L NaCl, 2 mmol/L MgCl₂, 0.1% SDS, 0.1% NP-40 (Igepal), protease inhibitor cocktail; all Sigma) for 10 seconds. The remaining cell body was collected by adding 100 μL of lysis buffer onto the top of each filter. Pseudopodia and cell body lysates from 3 filters were pooled separately and subjected to 3 freeze/thaw cycles (dry ice for 3 minutes and thawed at 37°C for 1 minute). Lysates were centrifuged at 13,000 × g for 20 minutes at 4°C and supernatants of pseudopodia lysates were dried in a SpeedVac, resuspended in double distilled water to make a 10× concentrated pseudopodia supernatant and protein concentration was measured by using the Micro BCA Protein Assay Kit (Pierce). Proteins from the cell body and the pseudopodia were separated on a 12% gradient gel (Invitrogen) and transferred to nitrocellulose membranes (Thermo Fisher) for the detection of MMP2 and MT1-MMP.

The enzymatic activity of secreted MMP2 derived from stable FTC-133 transfectants and untransfected FTC-133 was determined by using the MMP2/MMP9 activity test (Calbiochem) according to the manufacturers' instructions. Cells (5 × 10³) were seeded per 96-well and grown overnight before cells were cultured for 24 hours in serum-free medium. Supernatants collected from all transfectants and from untransfected FTC-133 treated with recombinant RLN2 (100 ng/mL) were incubated at 37°C with 20 μmol/L of the substrate. The enzymatic activity was measured after 30 minutes at 280 nm.

Migration assays were conducted in 24-well Transwell chambers (Costar). The upper and lower culture chambers were separated by a polycarbonate membrane with 8-μm pore size. Membranes were coated with 1 mg/mL gelatine, collagen A (collagen type I), or fibronectin (all Biochrom). FTC-133, FTC-133–RLN2, and FTC-133–EGFP (each at 1 × 10⁴) in serum-free medium were seeded onto the filter and incubated for 24 hours at 37°C in a CO₂ incubator, in the presence and absence of RLN2 at 100 ng/mL. We used the general MMP inhibitor GM-6001 (Ilomastat; Chemicon/Millipore) at a concentration that inhibits MMP2 (0.5 nmol/L). FTC-133 cells did not express MMP9 (data not shown). Cells that had migrated to the underside of the filter were washed with PBS, fixed for 10 minutes in ice-cold methanol: PBS [1:1 (v/v); Merck], followed by a 20 minutes wash in ice-cold methanol and staining in 0.1% toluidine blue solution (Merck) in sodium carbonate (Roth). Stained cells were counted by light microscopy (Zeiss) in 5 separate fields (magnification ×100) per filter.

The 24-well plates were coated with 1 mg/mL gelatine (Biochrom) and dried overnight in sterile conditions. FTC-133–RLN2 and FTC-133–EGFP transfectants (2 × 10⁴) were seeded onto the coated plates and cultured for 48 hours at 37°C in a CO₂ incubator. Proteins were stained with Scarlet Bibrich solution (Wako). Stained plates were visualized by light microscopy (Zeiss). For collagen zymography, collagen-I from bovine Achilles tendon (1 g; Sigma) was dissolved in 0.2 mol/L acetic acid and stirring at 4°C overnight. The collagen solution was neutralized and adjusted to pH 7.2 with concentrated NaOH and adjusted with 50 mmol/L Tris-HCl, pH 7.4. Ten microgram of total protein extracts of FTC-133–RLN2 and FTC-133–EGFP transfectants were mixed with 4× extraction buffer (for composition see Protein extraction and Western blot analysis; without mercaptoethanol) and subjected to electrophoresis on a 10% SDS-PAGE copolymerized with 0.035% v/v of solubilized collagen. The gel was washed for 1 hour at RT with 2.5% Triton X-100 in 50 mmol/L Tris pH 7.4, 5 mmol/L CaCl₂, 1 μmol/L ZnCl₂, rinsed in distilled water, and incubated at 37°C for 16 hours in 0.01 mol/L sodium phosphate buffer, pH 7.4, containing 5 mmol/L CaCl₂ followed by staining with 0.5% Coomassie brilliant blue R250 in 30% ethanol, 10% acetic acid for 30 minutes and then destained with 30% ethanol and 10% acetic acid. Collagenolytic activities were detected as unstained bands against the background of Coomassie brilliant blue–stained collagen.

Statistical analysis was carried out with SPSS 12.0 and Excel software and all experimental parameters were calculated for statistical significance by using Student's t test. The values of P < 0.05 were considered statistically significant. Densitometric analysis was carried out by using the Kodak Digital Science 1D software.

We analyzed human goiter tissues and PTC, FTC, and UTC thyroid cancer tissues for the presence of immunoreactive MMP2, MT1-MMP, and RLN2. Goiter tissues showed very weak to absent immunostaining for MMP2 (Fig. 1A), MT1-MMP (Fig. 1B), and RLN2 (Fig. 1C). Only 1 of 15 (7%) and 3 of 10 (30%) of goiter cases showed weak staining for MMP2 and MT1-MMP, respectively. None were positive for RLN2 as described previously (7). All thyroid carcinoma tissues investigated (PTC, FTC, and UTC) displayed increased presence of immunoreactive MMP2 (Fig. 1D, G, and J), MT1-MMP (Fig. 1E, H, and K), and RLN2 (Fig. 1F, I, and L). When the specific antiserum was replaced with a nonimmune serum of the same species, no specific immunostaining was obtained in any of the tissues (Fig. 1a–l). RLN2 colocalized with MMP2- and MT1-MMP–producing cells. We were unable to find a correlation for immunoreactive MMP2 or MT1-MMP thyroid cancer tissues with either gender, age, or tumor size, lymphatic/systemic metastasis (pTNM) status of the patients (Table 2). No correlation with these clinicopathologic criteria was found for the presence of both MMP2 and MT1-MMP in the few thyroid cancer tissue cases (3 PTC, 6 FTC, and 5 UTC) which we were able to test for both MMPs (Table 2). The immunohistochemical staining results suggested that MMP2 may be a suitable marker for thyroid cancer because only a minority of goiter tissue (<10%) but a high percentage (61%–77%) of thyroid cancer tissues expressed immunoreactive MMP2. MT1-MMP was present in both goiter (30%, weak staining) and thyroid cancer tissues (67%–82%; Table 3). As shown previously, MT1-MMP and MMP2 were present in fetal trophoblast cells of human placenta used as positive control (data not shown; ref. 48).

We show that RLN2 increased the transcriptional activity, protein production, and secretion of gelatinase A/MMP2 in previously described RLN2-expressing stable transfectants of the human follicular thyroid carcinoma FTC-133. MT1-MMP protein in either the 64 kDa proform (Fig. 2A) and 54 kDa active form (Fig. 2B) were upregulated in 2 out of 3 or all 3 RLN2 transfectants, respectively, when compared with the EGFP control. All RLN2 clones, but not the EGFP mock, displayed strong immunoreactive bands at 72 kDa corresponding to MMP2 (Fig. 2C). TIMP2 is part of an MMP2 activation complex, which also includes proMMP2 and MT1-MMP (49). The expression of TIMP2 detected as a single immunoreactive protein at 21 kDa did not change significantly in all EGFP and RLN2 clones tested (Fig. 3D). The expression levels determined for MT1-MMP, MMP2, and TIMP2 correlated well with the densitometric measurements derived from the Western blot analysis for MT1-MMP proform (64 kDa) and active form (54 kDa), MMP2, and TIMP2 (data not shown). The upregulation in MT1-MMP and MMP2 correlated with RLN2 production as determined by RLN2 ELISA in the FTC-133–RLN2 stable transfectants (70–95 pg/mL) and negligible levels of RLN2 in the EGFP mock cells (data not shown).

Of the 4 TIMP protein members identified by Western blot in the FTC-133 transfectants, RLN2 caused an exclusive and marked reduction in TIMP3 isoforms at 48 kDa (Fig. 2E) and 27 kDa (Fig. 2F), without affecting the levels of TIMP1 or TIMP4 protein (Table 4). Expression of other members of the MMP family known to be associated with cancer, including MMP1 (collagenase 1), MMP3 (stromelysin 1), MMP8 (collagenase 2), MMP9 (gelatinase B), MMP10 (stromelysin 2), and MMP13 (collagenase 3) was tested in the FTC-133–EGFP and FTC-133–RLN2 transfectants by using reverse transcriptase PCR (RT-PCR). Transcripts were exclusively observed for MMP13 and remained unaltered in the presence of RLN2 (Table 4). FTC-133 cells were devoid of MMP1, 3, 8, 9, and 10 transcripts (Table 4).

Supernatants of RLN2 transfectants contained significantly more bioactive MMP2 when compared with the mock control as determined by MMP2/MMP9 bioactivity assay (Fig. 3A). The enzymatic processing of the substrate was likely caused by the increased MMP2 activity because RLN2 transfectants did not express MMP9 (Table 4). The addition of the general MMP inhibitor GM6001 abolished the enhanced proteolytic cleavage of the fluorescently labeled substrate by the RLN2 transfectants, but not mock transfectants (Fig. 3C). Untransfected FTC-133 thyroid cancer cells exposed to recombinant RLN2 also showed enhanced secretion of active MMP2 (Fig. 3B). Furthermore, protein extracts of RLN2 transfectants (clone 10) subjected to in-gel collagen-I zymography revealed a specific band of digested collagen-I at 72 kDa likely resembling MMP2 (Fig. 3D). In situ zymography assays also showed that the transfectants cultured in gelatine matrices differed in their ability to degrade this matrix as evidenced by the larger bright pericellular halos around RLN2 transfectants, which could be inhibited with the general MMP inhibitor GM6001 at a concentration known to inhibit MMP2 (Fig. 3E).

We conducted migration assays to determine whether the observed increase in MMP2 and MT1-MMP production and MMP2 activity resulted in enhanced penetration of the RLN2 transfectants through collagen type I matrices. All 3 RLN2 transfectants tested (Fig. 4A) and FTC-133 cells treated with human recombinant RLN2 (Fig. 4B) were able to penetrate collagen-I matrix and migrate to the underside of the filter in significantly higher numbers than mock and untreated controls, respectively. In addition to collagen I, MMP2 and MT1-MMP are capable of digesting other extracellular matrix compounds, including gelatine and fibronectin (50). We observed a positive trend, but no significant increase, in migration through gelatine matrices and no change in the ability of clones to invade fibronectin matrix (data not shown).

To further elucidate the potential mechanism by which RLN2 may facilitate the enhanced migration of thyroid cancer cells through collagen matrices, we isolated pseudopodial membrane protrusions extending into 1-μm wide filter pores after a 1 hour migration time. The narrow filter pores only permitted the collection of invadopodia rather than the whole cell body which stayed on top of the filter as shown by the exclusive presence of the nuclear protein lamin A/C in protein preparations from the top, but not the bottom, of the filter (Fig. 4C). This pseudopodia assay allowed us to determine proteins present at the leading edge of migration (bottom of filter) and the rest of the cell body (top of filter). Western blot analysis of protein extracts from the invadopodia (bottom of filter) and rest of the cell body (top of filter) confirmed the increased overall production of MT1-MMP in RLN2 clones as compared with EGFP clones (data not shown; Fig. 2A and B). Protein levels of MT1-MMP were increased in the invadopodia fraction (just below significance) and significantly in the cell body on top of the filter as shown by quantitative Western blot analysis (Fig. 4C and D). Next, we used multicolor fluorescent confocal microscopy on RLN2 and EGFP transfectants to visualize the presence of MT1-MMP in invadopodia invading the 3-μm wide filter pores. A monoclonal antibody against beta 1 integrin was used as a membrane marker. We observed pronounced immunofluorescent staining for both MT1-MMP and beta 1 integrin in the invadopodia of RLN2 clones when compared with EGFP clones at the same exposure time (Fig. 5A). Thus, one way of RLN2 to ensure increased collagen invasion was to increase the concentration of MT1-MMP within the cells and the invadopodia as compared with EGFP clones. In the invadopodia, though, MT1-MMP was distributed over a much smaller surface area (Fig. 5A) and, thus, was more concentrated as shown in the confocal fluorescent images of the membrane pore-penetrating invadopodia (Fig. 5A). RLN2 may alter the actin cytoskeleton and this may contribute to the increased invasion of ECM. Staining of F-actin by the fungal toxin phalloidin revealed distinct differences in the actin cytoskeletal organization in EGFP and RLN2 transfectants, with the latter displaying stress fibers (Fig. 5B). Increased stress fiber formation was also detected in the 3-dimensional (3D) confocal images of invadopodia derived from the RLN2 transfectants (Fig. 5C), suggesting changes to the cytoskeletal actin architecture and increased presence of MT1-MMP in pseudopodia to contribute to invadopodia formation and ECM-penetrating activity.

In this study, we have shown the presence of MMP2 (gelatinase A) and MT1-MMP (MMP14) in human benign and malignant thyroid tissues. Of all the MMP members (MMP1, 2, 8, 9, 10, 13, and 14) and TIMPs (TIMP1–4) investigated here, RLN2 specifically induced the expression of MMP2 and MT1-MMP in human follicular thyroid carcinoma cells. Production of proMMP2 and MT1-MMP has previously been shown in human PTC and MT1-MMP expression levels correlated with the proMMP2 activation ratio, thyroid cancer cell invasion, and PTC lymph node metastasis (31, 35). Similarly, a positive correlation was reported for the presence of immunoreactive MMP2, but not TIMP2, and bad prognosis of patients with medullary thyroid cancer, a tumor derived from C-cells in the thyroid gland (51). Expression of gelatinase A (MMP2), gelatinase B (MMP9), and MT-MMP1 has also been reported in breast, colon, lung, skin, ovary, and prostate cancer and is associated with enhanced invasion and decreased overall survival (52–54). MT1-MMP is important for cell-surface activation of proMMP2 (49) and the MT1-MMP C-terminal domain is essential for promoting cell migration (55, 56). Both MMP2 and MT1-MMP are major players in tumor angiogenesis and linked to lymphangiogenesis, lymphatic invasion, and lymph node metastasis (53, 57). The targeted deletion of MMP2 impairs lymphatic vasculature and the inhibition of MMP2 and MT1-MMP attenuates the formation of both new blood and lymphatic vessels and reduces lymph node metastasis (58). The angiogenic phenotype of MT1-MMP–producing cells is associated with the upregulation of VEGF (59). Mice deficient in MT1-MMP develop skeletal deformations as a result of delayed vascular invasion of the cartilage (60) and overexpression of MT1-MMP induces tumorigenic mammary gland abnormalities in transgenic mice (61). Our finding of an upregulation of MT1-MMP/MMP2 coincided with the RLN2-induced downregulation of TIMP3. TIMP3 is known to inhibit VEGF-mediated angiogenic signalling and lack of TIMP3 has been shown to cause an increase in abnormal vessel formation (62). Our findings may in part explain the increased angiogenesis and growth we had reported previously in xenografts derived from the FTC-133 transfectants expressing RLN2 used in this study (7).

Currently, few factors have been identified which can affect the expression of MMP2 and MT1-MMP. Epidermal growth factor (EGF) was shown to mediate the invasion of differentiated thyroid cancer cells by an EGF receptor–mediated increase in the expression and activation of MMP2 (35). The membrane-bound glycoprotein and MMP1 receptor EMMPRIN (extracellular matrix metalloproteinase inducer, CD147) was reported to induce MMP1, 2, 3, and MT1-MMP, but not MMP9 expression (63). Here, we have identified RLN2 as a novel potent inducer of MMP2 and MT1-MMP in human follicular thyroid cancer cells. RLN2 is strongly and specifically expressed in thyroid cancer tissues and both RLN receptors, RXFP1 and RXFP2, are expressed in the thyroid gland (7, 64, 65). Exposure to RLN2 caused a significant RXFP1-mediated increase in the ability of human thyroid cancer cells to invade elastin matrices as a result of the enhanced production and secretion of cathepsin-L, a potent elastinolytic lysosomal hydrolase (7). Our discovery of MMP2 and MT1-MMP as novel RLN2 targets in human thyroid cancer cells identifies RLN2 as an important multifunctional regulator of a range of potent proteases with overlapping ECM-degrading qualities known to facilitate various stages of (thyroid) cancer cell invasiveness (35, 66–68).

RLN exerts autocrine/paracrine actions on tumor cell migration/invasion and angiogenesis in different human cancer models (1, 5). Follicular thyroid cancer cells and PC-3 prostate carcinoma cells expressing RLN2 displayed enhanced xenograft growth and angiogenesis (7, 69, 70). The importance of the RLN/RXFP1 ligand receptor system for tumor growth is also highlighted by 2 sets of in vivo experiments. We and others have recently shown that the intratumoral injection of a RLN antagonist peptide (71) or RXFP1 antisense coated nanoparticles (72) results in the specific in vivo silencing of RXFP1 signaling pathways in xenografts of human prostate cancer cells and causes a marked reduction in tumor size. High RLN2 expression has been associated with advanced endometrial cancer (EC) tumor stages, myometrial EC invasion, and a significantly shorter overall survival of women with EC (8). The stimulation of the human EC cell lines HEC-1B and KLE with RLN2 resulted in the RXFP1-mediated release of increased levels of MMP2 in KLE cells and MMP9 in HEC-1B cells (8). RLN2 affects breast cancer growth and invasiveness in a cell type-specific manner (73, 74). RLN2 is upregulated in neoplastic breast lesions (75) and systemic RLN2 levels were reported to be higher in women with metastatic breast cancer (10). The RLN2-induced increase in the secretion of MMP2, MMP7, and MMP9 and the upregulation of MMP2, MMP9, MMP13, and MT1-MMP transcripts resulted in enhanced in vitro invasiveness of the human breast cancer cell lines SK-BR3 and MCF7 (20). A recent study failed to show a correlation between plasma RLN levels and the presence of RLN, RXFP1, or MMP2 in the local tumor environment (76). Instead, these authors reported statistically significant correlations for the presence of RLN and RXFP1 and for RXFP1 and MMP2 in canine mammary tumor tissues and emphasized the importance of a locally active RLN/RXFP1 system for ECM remodeling in tumor tissues (76).

Currently, little information is available on the local effects of RLN2 in tumor cells and tissues. The MMP2/MT1-MMP/TIMP2 complex and α_Vβ₃ integrin were reported to localize to specific caveolin microdomains, invadopodia, at the leading edge of migration (77, 78). This prompted us to elucidate the effects of RLN2 on the local distribution of MT1-MMP at the leading edge of migration in human follicular thyroid carcinoma cells. Pseudopodia assays resemble a modified motility assay with the ability to isolate cell components exclusively derived from the invading membrane tip of motile cells (47). We combined pseudopodia assays with Western blot analysis and confocal fluorescent microscopy to gain quantitative and visual information on the distribution of MT1-MMP at the leading migrating edge. The higher level of immunoreactive MT1-MMP in the RLN2 transfectants caused more MT1-MMP to be concentrated in the small volume of the invadopodia, likely contributing to high local levels of collagenolytic activity with resulting more effective collagen matrix degradation by RLN2 transfectants as compared with the EGFP clones. This “extrinsic” effect of RLN2, which affected the proteolytic activity at the leading migrating edge of tumor cells and resulted in ECM remodeling in the pericellular space, was not the only mechanism used by this hormone to promote tumor cell invasion. Exclusive to the RLN2 transfectants and not observed in EGFP clones, RLN2 also facilitated invasion of thyroid cancer cells by an “intrinsic” mechanism that promoted the formation of actin stress fibers by an as yet unidentified mechanism. Confocal images revealed that these actin stress fibers participated in the formation of the invadopodia in RLN2 transfectants. The remodeling of the actin cytoskeletal architecture and formation of bundles of actomyosin stress fibers are important steps in invadopodia formation and facilitate cell migration and ECM degradation (72, 79).

In summary, the specific upregulation and colocalization of RLN2, MT1-MMP, and MMP2 in human thyroid cancer tissues as well as the contribution of RLN2-induced actin stress fibers to the formation of MT1-MMP–rich invadopodia that cause local collagenolytic remodeling of the surrounding ECM microenvironment identifies RLN2 as a novel and versatile pathogenic factor in thyroid cancer invasiveness.

No potential conflicts of interest were disclosed.

C. H-Vu and T. Klonisch thank the Deutsche Forschungsgemeinschaft (DFG) for their financial support (KL1249/5-1; 5-2). T. Klonisch and S. H-Klonisch thank the Natural Sciences and Engineering Council of Canada (NSERC), the Surgery Research Fund, and the Manitoba Institute for Child Health (MICH) for their generous support. The authors thank Corthera Inc. for their generous gift of recombinant human relaxin.

This work was financially supported by Deutsche Forschungsgemeinschaft (DFG; KL1249/5-1; 5-2), Wilhelm Roux Program, Medical Faculty, Martin Luther University Halle-Wittenberg, Natural Sciences and Engineering Research Council of Canada, and the Manitoba Health Research Council, Canada.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.