Tissue Inhibitor of Metalloproteinases-1 Promotes Liver Metastasis by Induction of Hepatocyte Growth Factor Signaling

Balanced expression of proteases and their inhibitors is one prerequisite of tissue homeostasis. Metastatic spread of tumor cells through the organism depends on proteolytic activity and is the death determinant for cancer patients. Paradoxically, increased expression of tissue inhibitor of metalloproteinases-1 (TIMP-1), a natural inhibitor of several endometalloproteinases, including matrix metalloproteinases and a disintegrin and metalloproteinase-10 (ADAM-10), in cancer patients is negatively correlated with their survival, although TIMP-1 itself inhibits invasion of some tumor cells. Here, we show that elevated stromal expression of TIMP-1 promotes liver metastasis in two independent tumor models by inducing the hepatocyte growth factor (HGF) signaling pathway and expression of several metastasis-associated genes, including HGF and HGF-activating proteases, in the liver. We also found in an in vitro assay that suppression of ADAM-10 is in principle able to prevent shedding of cMet, which may be one explanation for the increase of cell-associated HGF receptor cMet in livers with elevated TIMP-1. Similar TIMP-1-associated changes in gene expression were detected in livers of patients with metastatic colorectal cancer. The newly identified role of TIMP-1 to create a prometastatic niche may also explain the TIMP-1 paradoxon.


Introduction
Metastasis and progressive scattering of tumor cells throughout tissues is the main cause of organ failure and subsequent death of cancer patients.Proteolytic remodeling of components of the extracellular matrix (ECM) in tissue interstitium is a prerequisite of the metastatic process (1).Matrix metalloproteinases (MMP) have long been thought to be major regulators of the pericellular protease-dependent steps of tumor progression (2).Altered expression profiles of some MMPs have been correlated with poor clinical prognosis for several human tumor types (3).Based on these facts, the historical view of MMPs and cancer was that they exerted proinvasive and prometastatic activity by mere ECM remodeling (1).Thus, it was hypothesized that therapeutic broadspectrum inhibition of MMPs would result in antimetastatic activity (1).Indeed, under certain experimental circumstances, overexpression of the endogenous broad-spectrum metalloproteinase [including a disintegrin and metalloproteinase-10 (ADAM-10); ref . 4] inhibitor tissue inhibitor of metalloproteinases-1 (TIMP-1; ref. 5) can reduce tumor cell invasion (6,7).
In contrast to these observations are other results, which are actually in agreement with clinical studies showing a correlation between elevated TIMP-1 expression and poor prognosis in many human cancer types (5,8), where elevated levels of TIMP-1 either had no effect on metastasis (9) or led to increased incidence (10,11) or growth (12) of primary tumors.These effects of TIMP-1 on cancer development have been attributed to its known proproliferative, antiapoptotic (5), or proangiogenic (13) activities, which are necessary but not sufficient for metastasis.
Tumor cell invasion and metastasis are the primary determinants of the survival of cancer patients (13).Signaling pathways leading to the expression of proteolytic enzymes have been suggested to be key mediators of invasive growth (14).Hepatocyte growth factor (HGF/scatter factor) signaling regulates a multitude of downstream prometastatic effector molecules, such as urokinasetype plasminogen activator (uPA), uPA receptor (uPAR), plasminogen activator inhibitor-1 (PAI-1), and MMPs (15)(16)(17).HGF stimulates metastasis of diverse tumor cells (15) and elevated levels of circulating HGF (18) or aberrant activity of cMet (14) has been detected in patients with many cancer types, including non-Hodgkin's lymphoma (19) and colorectal cancer (14).Although elevated TIMP-1 expression in these patients correlates with decreased survival, a link between TIMP-1 and metastasispromoting pathways has previously not been described.Here, we address this issue and report that elevated stromal levels of TIMP-1 induce HGF signaling, leading to enhanced expression of other metastasis-promoting genes.Consequently, susceptibility of the liver for metastasis of tumor cell lines of different origin was shown to be increased, assigning a new role to TIMP-1 as a factor creating a prometastatic niche.
HGF ELISA.Livers were perfused with 100 mL PBS before excision, sample preparation, purification (HiTrap heparin column, Amersham), and ELISA [rat HGF ELISA (Institute of Immunology Co. Ltd.)] were done following the manufacturers' instructions.This kit detects mouse HGF equally well (manufacturers' manual).The ELISA was done in duplicates and results were expressed as picogram HGF per milligram extracted liver.
Patient material.Liver tissue of 14 patients with metastatic colorectal cancer (7 males and 7 females; mean age, 61.0 F 2.3 years) was obtained from patients who had signed informed consent for scientific study of their tumors and surrounding tissues at the Department of Surgery of the Klinikum rechts der Isar or of the University Hospital Heidelberg.Tissue samples were harvested under identical, rigorous procurement protocol and snap frozen in liquid nitrogen immediately after surgical resection.''Normal'' liver tissue was harvested from zones 2 cm distant from visible liver metastases.Three experienced pathologists from two independent institutions used cryodissected sections for confirmation of histologic identity.All samples were stored at À80jC until use.
Inhibition of HGF signaling by a specific kinase inhibitor.Three days before i.v.tumor cell inoculation, four mice were transduced with adenoviral vectors.Mice received 100 mg/kg of 2 inhibitor), with high affinity for cMet [dissolved in 6% (v/v) Tween 80 with two equivalents HCl], i.p. 1 h before and 1 h after tumor cell inoculation.I.p. treatment with the cMet inhibitor was continued daily (100 mg/kg/d).As control, mice were treated with vehicle alone.Three days after tumor cell inoculation, mice were sacrificed and livers were removed and X-Gal stained as described previously (22).Inhibition of cMet by this inhibitor was measured by an ELISA-type assay: Lysate of human colon adenocarcinoma HT29 cells was exposed to a microtiter plate with an antihuman HGF receptor antibody (BAF 358 ATP phosphorylation; R&D Systems).ATP phosphorylation of the receptor kinase (cMet kinase) was detected in the presence or absence of the test compounds using a phosphorylated tyrosine mouse IgG (Santa Cruz Biotechnology).The IC 50 value for cMet inhibition is 2 nmol/L.Kinase inhibition activity against other kinases was measured as described by Fabian et  Knockdown of ADAM-10 in NIH3T3 cells.Oligodeoxynucleotides coding for small hairpin RNA interference (shRNAi) directed against murine ADAM-10 (small interfering RNA sequence: AACCTACGAATGAAGAGG-GAC) were designed and obtained from Eurogenetec.Annealed shRNAis were cloned into pSiren-RetroQ (Clontech).Recombinant retroviruses were generated as described ( 23) by transient cotransfection of 293T cells with pHIT60 (23), pHCMV-G (33), and pSiren-RetroQ coding for anti-ADAM-10 shRNAi.Infection of NIH3T3 cells was done according to standard protocols (23).Transduced cells were selected with 20 Ag/mL puromycin (Clontech).
cMet ELISA.For detection of membrane-bound cMet, NIH3T3 cells transduced with the different viral constructs were resuspended in 0.1 mol/L Tris, 10 mmol/L EDTA, and 0.1% Triton X-114 (AppliChem) at pH 7.5, supplemented with 10 AL Protease Inhibitor Cocktail (Sigma).Subsequently, cells were incubated for 5 min on ice and 5 min at 37jC and centrifuged for 5 min at 300 Â g at 30jC.The lower phase containing the membrane proteins was used for ELISA.For detection of cMet shedding, transduced NIH3T3 cells were incubated with serum-free medium for 36 h.Total and cMet protein in membrane fractions and supernatants were quantified by bicinchoninic acid assay and cMet ELISA (Mouse HGF R/c-MET ECD DuoSet, R&D Systems), respectively, following the manufacturers' instructions.Assays were done in triplicates.
Statistical analysis.Data were tested for normal distribution and normally distributed data were tested by Student's t test, and where the normality test failed, Mann-Whitney test was applied.P < 0.05 was considered significant.

Results
Reduction of experimental macrometastases but induction of scattered micrometastases in the liver by TIMP-1.To achieve elevated TIMP-1 expression in the host, we transduced full-length cDNAs of TIMP-1 by i.v.inoculation of adenoviral vectors into mice.Efficient adenoviral transduction of liver cells was determined by quantitative RT-PCR (Supplementary Fig. S1).Maintenance of transgene overexpression achieved by the adenoviral vectors until the end of all following metastasis assays was verified (Supplementary Fig. S1).To test the effects of elevated stromal TIMP-1 on tumor cell invasion, we used the well-established syngeneic in vivo metastasis assay with lacZ-tagged lymphoma cells (L-CI.5s),which form hepatic metastatic foci following i.v.injection.Colonies that we defined as macrometastases in livers reached the cutoff size of 0.2 mm in this assay at day 6 after tumor cell inoculation.At this time point, the combination of successful extravasation right after tumor cell inoculation and proliferation of the extravasated tumor cells for at least six days resulted in multicellular foci of this size.Detected micrometastases (<0.2 mm) resulted from additionally invaded tumor cells if their number exceeds the 5,000 inoculated tumor cells.
Three days after adenoviral gene transfer, TIMP-1-transduced and control mice were challenged with L-CI.5s tumor cells.The pattern of metastasis in TIMP-1-transduced livers was significantly altered compared with the control virus-transduced livers: Elevated levels of TIMP-1 correlated with significant reduction of the number of macrometastases (by 43%, P = 0.003; Fig. 1A, top left).However, on the level of micrometastases, no decrease but a drastic increase of scattered infiltrating tumor cells was detected in livers with elevated TIMP-1 (Fig. 1A).In one photographic frame (1.3 mm Â 1.5 mm), already 623 micrometastatic foci at the liver surface were visible (Fig. 1A, bottom), showing that the number of scattered micrometastases in the entire liver by far exceeded the number of inoculated cells (5,000).Quantification of the tumor cell tag lacZ revealed an overall increase of total metastasis burden in livers transduced with TIMP-1 cDNA compared with the control (P = 0.044; Fig. 1A, top right), reflecting induction of micrometastasis.
To determine whether inhibition of macrometastasis but promotion of micrometastasis was a function of the metalloproteinase-inhibitory activity of TIMP-1, we investigated the effect of the N-terminal domain of human TIMP-1 (N-TIMP-1) harboring the metalloproteinase-inhibitory bioactivity on liver metastasis.Indeed, elevated levels of N-TIMP-1 significantly suppressed macrometastasis by 51% (P = 0.009; Fig. 1A) and promoted micrometastasis (P = 0.041; Fig. 1A), indicating that N-TIMP-1 is sufficient to evoke the described effects.
We used lacZ-tagged human fibrosarcoma cells (HT1080lacZ-K15) that typically form large pulmonary metastatic foci and only few single tumor cells as micrometastases, but no macrometastases, in livers (26) to exclude that the TIMP-1 effect was cell type dependent.After challenge of CD1 nu/nu mice with HT1080lacZ-K15, significant reduction of lung metastasis by 93% was observed in TIMP-1-transduced animals compared with mice with virus control (P = 0.018; Fig. 1B, top left).Again, in livers with elevated TIMP-1 levels, a significant increase of the metastatic burden was found (P = 0.04; Fig. 1B, top right).No induction of HT1080lacZ-K15 micrometastasis was detected in lungs (data not shown).In the above experiments, no differences in metastasis between control virus-transduced mice and mice treated with vehicle alone were found, excluding effects of viral transduction itself (data not shown).
Effects of TIMP-1 elevation on liver metastasis after tumor cell inoculation.To determine whether TIMP-1 affects liver metastasis on two different levels, [i.e., on the level of initial manifestation and subsequent growth (leading to formation of macrometastases) or subsequent infiltration (micrometastases)], we inoculated L-CI.5s cells into CD1 nu/nu 2 days before adenoviral transfer of TIMP-1 cDNA.The number of macrometastatic colonies was not affected compared with the control (Fig. 1C, top left).However, we still observed induction of micrometastasis compared with the control (Fig. 1C, bottom), resulting in a significantly increased total metastasis burden in livers with elevated TIMP-1 (Fig. 1C, top right).This indicates that micrometastases take advantage of the TIMP-1-modulated host microenvironment.In contrast, macrometastasis formation was only reduced when TIMP-1 was already elevated in the host at the time point of extravasation right after tumor cell inoculation (Fig. 1A).
Proliferation activity of TIMP-1-induced scattered liver micrometastases.To assess if the increased presence of invaded L-CI.5s cells in the liver was a function of TIMP-1-induced proliferation, we determined their proliferative status by examining PCNA expression.Indeed, infiltrating T-cell lymphoma cells in livers with elevated TIMP-1 or N-TIMP-1 were proliferatively active (Fig. 2A), indicating a proproliferative effect of the TIMP-1modulated environment on these cells, accounting for the increase of total metastasis burden in the liver.A proproliferative environment was further indicated by the presence of proliferating hepatocytes in these groups (Fig. 2A, arrows).In the control livers, proliferation was restricted to the macrometastases (Fig. 2A, left).
TIMP-1 induces HGF signaling in the liver.To elucidate a molecular mechanism responsible for the TIMP-1-induced proproliferative and proinvasive environment in the liver, we tested for expression of HGF, also known as scatter factor.Elevated levels of HGF correlated with the TIMP-1-induced scattered metastatic phenotype (Fig. 2B, left).In fact, increase of HGF expression was already a component of the TIMP-1-altered liver environment as it could already be detected 3 days after adenoviral gene transfer (i.e., at the time point of tumor cell challenge; Fig. 2B, right).In addition, cMet, the tyrosine kinase receptor of HGF, was elevated throughout the tissue of metastasis-bearing livers transduced by either TIMP-1 species (Fig. 2C) and in macrometastases (Fig. 2C, insets).
Elevated expression of either TIMP-1 variant led to markedly increased HGF signaling, as detected by immunohistochemical staining of phosphorylated cMet throughout the liver parenchyma (Fig. 2D).In control animals, activated cMet was very low even in macrometastases (Fig. 2D, top).Importantly, this TIMP-1-mediated induction of HGF signaling was already present 3 days after TIMP-1 gene transfer (Fig. 2D, bottom), at the time point of tumor cell challenge, suggesting that a HGF pathway-connected prometastatic niche has been formed.
Causal relationship between HGF signaling and TIMP-1induced scattering of experimental liver metastasis.In TIMP-1-transduced animals, the scattered metastasis phenotype was already apparent 3 days after inoculation of L-CI.5s cells (Fig. 3A, top).To investigate whether HGF signaling was essential for formation of scattered metastasis, we first transduced mice with AdTIMP-1 or Addl70-3 virus, treated the mice with a cMet inhibitor, and challenged with L-CI.5s cells.Kinase inhibition prevented formation of micrometastasis (Fig. 3A, bottom), indicating that HGF signaling was essential for TIMP-1-induced micrometastatic spread.The TIMP-1-induced significant increase of total metastasis burden compared with the control (P < 0.001; Fig. 3B) was significantly reduced by cMet inhibition (P < 0.001; Fig. 3B).The number of multicellular metastatic foci was significantly reduced by elevated levels of TIMP-1 compared with the control (P = 0.048; Fig. 3C), indicating that formation of macrometastasis is not affected by HGF signaling.
Overexpression of TIMP-1 reduces metalloproteinase activity and increases activity of other proteases in the liver.As proteolytic activity is one prerequisite of metastasis, we next assessed whether TIMP-1-related changes in metastasis rely on modulated proteolytic activity.In situ zymography revealed gelatinolytic activity in macrometastatic foci in Addl70-3-transduced livers but not in foci of AdTIMP-1-transduced animals (Fig. 4A, left).As this was inhibitable by the canonical metalloproteinase inhibitor 1,10-phenanthroline (Fig. 4B, left), we showed that the achieved TIMP-1 levels indeed inhibited MMP activity in vivo.No MMP-9-derived proteolytic activity was detected in livers with elevated TIMP-1 (Fig. 4C), indicating the expected functional activity of TIMP-1 to inhibit proMMP-9 activation.
The gelatinolytic activity detectable in TIMP-1-elevated liver parenchyma distant from metastases (Fig. 4A, right) was by and large metalloproteinase independent, as it was only slightly inhibitable by 1,10-phenanthroline (Fig. 4B).Addition of broadspectrum cysteine protease inhibitor E64 (34) or broad-spectrum serine protease inhibitor aprotinin (35) to the in situ zymography DQ-overlay revealed the remaining gelatinolytic activity as partly derived from serine and cysteine proteases (Fig. 4D).Elevated levels of N-TIMP-1 had the same effects (data not shown).These data indicate that increased expression of TIMP-1 in the liver led to an increase of metalloproteinase-independent proteolytic activity.
TIMP-1-induced expression of metastasis-related genes in the liver.Induction of HGF signaling and nonmetalloproteinase proteolytic activity in liver parenchyma by TIMP-1 suggested an altered HGF signaling-related gene expression favoring metastasis.Therefore, we used quantitative RT-PCR to characterize the expression of HGF-inducible genes, HGF-activating proteases, as well as gelatinolytic proteases.Elevation of TIMP-1 led to increase of mRNA expression of uPA, uPAR, PAI-1, tissue plasminogen activator (tPA), matriptase, MMP-9, MMP-2, ADAM-10, cathepsin G, and neutrophil elastase in the tumor-free liver (Fig. 5).Elevated levels of N-TIMP-1 had the same effects (data not shown).
To test whether these TIMP-1-regulated changes in gene expression are of more general relevance, we measured TIMP-1 mRNA expression in human liver tissue samples distant from colorectal metastases.In normal tissue, elevated expression of TIMP-1 (relative mRNA expression: 98.92 F 19.84, n = 8) was associated with increased mRNA expression of uPA, uPAR, matriptase, MMP-9, MMP-2, ADAM-10, cathepsin G, and neutrophil elastase compared with livers with significantly lower TIMP-1 expression (relative mRNA expression: 21.22 F 1.89, n = 6, P < 0.001; Fig. 5).This difference also correlated with shorter time interval between detection of primary tumor and liver metastases [26.83 F 5.81 months (n = 6) versus 13.75 F 2.62 months (n = 8); Inhibition of cMet shedding by TIMP-1.Interestingly, although cMet protein was increased in livers with elevated TIMP-1 (Fig. 2C), mRNA expression of cMet was not up-regulated in these livers, neither at the time point of tumor cell inoculation (Fig. 5) nor in metastasis-bearing livers (Supplementary Fig. S2).Reduced shedding of cMet by TIMP-1-induced inhibition of metalloproteinase activity may account for the accumulation of cMet protein, but it was thus far unknown that TIMP-1 can inhibit its shedding.As this hypothesis cannot be addressed in vivo, we chose a murine nontumorigenic cell line expressing inhibitable metalloproteinases as well as cMet, thus providing all prerequisites for a proof-of-concept study.Overexpression of either TIMP-1 or N-TIMP-1 in NIH3T3 cells (Fig. 6A) resulted in significantly augmented membrane-bound cMet protein in these cells compared with the virus control (both P V 0.026; Fig. 6B), although RNA expression of cMet was not increased (both P > 0.33; Fig. 6C).
Activity of ADAM-10 regulates cMet shedding.Knockdown of ADAM-10 mRNA expression in NIH3T3 cells by 88% (Fig. 6A) led to a significant increase of membrane-bound cMet protein compared with the control (P = 0.018; Fig. 6B), whereas cMet mRNA expression was not enhanced (P = 0.2; Fig. 6C).ADAM-10 knockdown was verified on protein level (Supplementary Fig. S4).A significant reduction of shed cMet in the supernatant of cells with reduced ADAM-10 expression was detected (P = 0.021; Fig. 6D), indicating that depletion of ADAM-10 was sufficient to accumulate cMet on the cell surface.In addition, suppression of ADAM-10 mRNA expression significantly increased cMet signaling (Supplementary Fig. S3).
Together, these data suggest a molecular pathway regulated by TIMP-1 that could lead to a microenvironment with higher susceptibility to metastases.

Discussion
This study shows that host-derived TIMP-1 can promote liver metastasis by induction of HGF signaling.Until recently, the dominant concept of natural inhibitors of tumor-associated proteases focused on their capacity to repress activity of ECM remodeling enzymes, such as metalloproteinases.This view has undergone revision based on increased awareness of ''degradome'' functions of metalloproteinases and the tumor microenvironment and recognition of their participation during many stages of tumor progression (3).
The original hypothesis that TIMP-1 overexpression would yield a reduction in the number of successfully extravasated tumor cells into a target organ was based on numerous reports on the importance of MMPs for metastasis in many tumor models (36).In agreement with these and our previous studies indicating the importance of gelatinase inhibition for formation of macrometastases of the lymphoma model (28,29), elevated host levels of TIMP-1 prevented macrometastases only when it was already increased at the time point of tumor cell inoculation.This confirms the already known effect of metalloproteinase inhibition by TIMP-1 on the level of extravasation.
In contrast to this inhibitory action on initial invasion, we reveal here for the first time an unexpected metastasis-promoting feature of TIMP-1 (i.e., induction of scattered invasion of tumor cells in liver parenchyma).The bulk of these micrometastases cannot derive from augmented extravasation of the L-CI.5s cells because the number of the proliferative active and scattered micrometastases by far exceeded the number of inoculated tumor cells.Previous studies have reported that, even when metalloproteinases are required for metastasis in a particular tumor model, TIMP-1 does not always suppress metastasis (9,37), supporting the idea that TIMP-1 possesses a broad range of biological activities, some of which may be independent of its metalloproteinase-inhibitory function (11,(38)(39)(40).Those variations of antitumorigenic efficacy of TIMP-1 have been suggested to depend on the concentrations of TIMP-1 (10,41) or related to the proproliferative and antiapoptotic features of TIMP-1 (11,12,38).
The new aspect is that we found a proinvasive feature of TIMP-1 when elevated in the liver based on TIMP-1-induced stimulation of HGF signaling and downstream expression of metastasispromoting genes.The N-terminal domain of TIMP-1, which includes the metalloproteinase-inhibitory domain, was sufficient to exhibit this effect, indicating that broad-spectrum inhibition of metalloproteinases is responsible for this scattering.In agreement with this notion, treatment with a synthetic broad-spectrum metalloproteinase inhibitor also leads to liver-specific promotion of metastasis in several xenograft models as well as in the syngeneic T-cell lymphoma model (42,43).In addition, in the present study, promotion of liver metastasis was found, whereas lung metastasis was significantly reduced.Either TIMP-1 was not able to induce HGF signaling in the lung or induction of HGF signaling in the lung tissue has different effects on the formation of a premetastatic niche than in the liver parenchyma.However, several studies revealed that inhibition or overexpression of proteases or protease families evolves different effects on metastasis to liver versus lung (26,(42)(43)(44).This leads to the hypothesis that metastasis to the liver requires a different repertoire of proteases than metastasis to the lung and that the different microenvironments present in these organs react differentially on changed proteolysis.The underlying mechanism of the organ specificity of the prometastatic effect of TIMP-1 will need to be determined in future studies.
Upon elevation of TIMP-1, we observed augmented HGF as well as cMet protein in livers.Simultaneous increase of HGF and functional cell-associated cMet can explain the overall increase of HGF/cMet signaling.Elevated presence of the active form of HGF, which is required for signaling (15), can derive from the observed upregulation of the HGF-activating proteases uPA and matriptase (45,46) in the liver.Active matriptase and uPA (also activated by matriptase; ref. 46) can lead to an additional increase of proteolytic activity by initiating a protease activation cascade, including the activation of plasminogen by uPA and of some pro-MMPs by active plasmin (47).Elevated levels of TIMP-1 in the liver also increased expression of cathepsin G, also a gelatinolytic enzyme (48), as well as expression of neutrophil elastase, which also enhances plasminogen activation (49).Such proteases are likely the source of the increased nonmetalloproteinase-derived proteolytic activity observed in metastasis-bearing liver sections with augmented TIMP-1 levels in situ.Interestingly, the TIMP-1-provoked gene expression signature of the experimental model resembled in some instances (matriptase, uPA, uPAR, PCNA, MMP-9, and cathepsin G) the signature observed in liver samples of colorectal cancers in patients with elevated levels of TIMP-1, which is a hint on general molecular mechanisms provoked by elevated stromal TIMP-1.In the future, stable knockdown of cMet in host and/or tumor cells will further characterize the role of TIMP-1-induced cMet signaling in liver metastasis.

Cancer Research
Cancer Res 2007; 67: (18).September 15, 2007 or N-TIMP-1 and may be a consequence of direct inhibition of ADAM-10 and thereby loss of its sheddase activity (4) or indirectly by inhibition of ADAM-10 activation.
Many tumor cell lines respond to elevated levels of HGF (scatter factor) with increased invasive behavior (15,16).Indeed, the altered host environment fed back to the inoculated tumor cell lines, which expressed functional cMet (Supplementary Fig. S5).Thus, increased scattering of tumor cells in the liver parenchyma seems to be based on a reaction of the tumor cells to the TIMP-1provoked change of homeostasis in the host.Identification of this TIMP-1-provoked prometastatic niche reveals a novel role for TIMP-1 during cancer progression and illuminates the intricacy of protease and protease inhibitor networks that regulate the penultimate step of cancer evolution and are one explanation for the association between tumor aggressiveness and increased levels of TIMP-1 in cancer patients.