Tài liệu Báo cáo khoa học: Collagen I regulates matrix metalloproteinase-2 activation in osteosarcoma cells independent of S100A4 pdf

Results Cell morphology and actin cytoskeleton structure As observed by light microscopy data not shown, pHb-1 and II-11b cells attached to plastic or mono-meric 2D collagen were spread

Collagen I regulates matrix metalloproteinase-2 activation

in osteosarcoma cells independent of S100A4

Renate Elenjord1, Jasmine B Allen2, Harald T Johansen3, Hanne Kildalsen1, Gunbjørg Svineng2, Gunhild M Mælandsmo1,4, Thrina Loennechen1and Jan-Olof Winberg2

1 Department of Pharmacy, University of Tromsø, Norway

2 Department of Medical Biochemistry, University of Tromsø, Norway

3 School of Pharmacy, University of Oslo, Norway

4 Department of Tumor Biology, The Norwegian Radium Hospital, Oslo, Norway

Introduction

The extracellular matrix (ECM) is an intricate network

of macromolecules composed of a wide variety of

locally secreted proteins and polysaccharides which are

closely associated with the surface of the cell that

pro-duced them The ECM can be found in different

forms, from the hard compositions of bone to the soft

structures of connective tissue Collagens are the main

components of ECM, and type I collagen is the most

abundant form in bone and connective tissue [1]

Controlled turnover of ECM is critical for a wide vari-ety of normal physiological processes, such as wound healing and embryogenesis The matrix metalloprotein-ases (MMPs) are considered to be the major enzymes involved in ECM remodelling, and dysregulated MMPs have been implicated in several diseases such as arthritis, cancer and cardiovascular disease [2] The family of MMPs consists of over 20 secreted and mem-brane-bound enzymes which are involved in

degrada-Keywords

collagen I; extracellular matrix; inhibitors of

matrix metalloproteinases; matrix

metalloproteinases; S100A4

Correspondence

J.-O Winberg, Department of Medical

Biochemistry, Institute of Medical Biology,

University of Tromsø, 9037 Tromsø, Norway

Fax: +47 776 45350

Tel: +47 776 45488

E-mail: janow@fagmed.uit.no

(Received 9 December 2008, revised 6 July

2009, accepted 20 July 2009)

doi:10.1111/j.1742-4658.2009.07223.x

This work investigates the effect of cell–collagen I interactions on the syn-thesis and activation of MMP-2, as well as synsyn-thesis of MT1-MMP and TIMP-1, by using an in vitro model with 3D fibrillar and 2D monomeric collagen In order to reveal whether the metastasis-associated protein S100A4 can influence the cell’s response to the two forms of collagen, oste-osarcoma cell lines with high and low endogenous levels of S100A4 were used Attachment of osteosarcoma cells to 3D fibrillar and 2D monomeric collagen resulted in opposite effects on MMP-2 activation Attachment to 3D fibrillar collagen decreased activation of proMMP-2, with a corre-sponding reduction in MT1-MMP By contrast, attachment to monomeric collagen increased the amount of fully active MMP-2 This was caused by

a reduction in TIMP-1 levels when cells were attached to monomeric 2D collagen The effect of collagen on proMMP-2 activation was independent

of endogenous S100A4 levels, whereas synthesis of TIMP-1 was dependent

on S100A4 When cells were attached to monomeric collagen, cells with a high level of S100A4 showed a greater reduction in the synthesis of

TIMP-1 than did those with a low level of STIMP-100A4 Taken together, this study shows that synthesis and activation of MMP-2 is affected by interactions between osteosarcoma cells and collagen I in both fibrillar and monomeric form

Abbreviations

APMA, p-aminophenylmercuric acid; ECM, extracellular matrix; MMPs, matrix metalloproteinases; MT-MMPs, membrane type matrix metalloproteinases; TIMPs, tissue inhibitors of MMPs.

tion and limited proteolysis of extracellular matrix.

Most MMPs are secreted as inactive proenzymes and

latency is maintained by an interaction between a

cystein residue in the prodomain and Zn2+ in the

active site of the catalytic domain Two major types of

endogenous inhibitors regulate the activity of MMPs;

a2-macroglobulin and four tissue inhibitors of

metallo-proteinases, TIMP-1 to TIMP-4 [2] In order to

iden-tify potential drug targets, it is important to know the

role of individual MMPs, their expression pattern and

activation mechanisms The best-described activation

mechanism for MMP-2 is the two-step cell-surface

activation in which proMMP-2 is activated in a

trimo-lecular complex including membrane type 1 MMP

(MT1-MMP) and the inhibitor TIMP-2, giving an

MMP-2 intermediate that is autocatalysed to fully

active MMP-2 [3] Other activators of MMP-2 can

directly activate proMMP-2 to fully active MMP-2, for

example, the lysosomal cysteine proteinase legumain,

which is known to activate proMMP-2 by cleaving the

Asn80–Tyr81 (i.e the autocatalytic site) and Asn82–

Phe83 bonds [4], and MT6-MMP, which can cleave

proMMP-2 at the Asn80–Tyr81 bond [5] Some

activa-tors are suggested primarily to take part in the second

activation step, in which the MMP-2 intermediate is

converted to the fully active MMP-2 Among these are

TIMP-2 [6] and integrins such as aVb3 [7] However,

there is some controversy regarding the role of the

aVb3 integrin, because it is also reported to suppress

collagen I-induced activation of proMMP-2 [8]

Another MMP inhibitor, reversion-inducing

cysteine-rich protein with kazal motifs (RECK) is a cell

mem-brane associated inhibitor of MMP-2 that is able to

inhibit the second activation step of MMP-2 [9]

The small calcium-binding protein S100A4 has been

shown to regulate expression of MMPs and their

inhibitors in several cell lines [10] The protein itself

has no known enzymatic activity, but binds to distinct

intracellular target proteins and regulates specific

func-tions involved in tumour progression such as cell

motility, proliferation and apoptosis [11] Although

S100A4 is strongly associated with the stimulation of

invasion and metastasis, the actual mechanism for the

metastasis-promoting function of S100A4 is not

com-pletely understood The protein seems to have several

functions, both intracellularly and extracellularly By

reducing the S100A4 level in a human osteosarcoma

cell line, and implementing these in mice, the capacity

to metastasize has been shown to decrease [12]

Culti-vation of the same cell lines on plastic also showed

decreased expression and activation of MMP-2 [13,14]

Previously, we have shown that a reduced

endoge-nous level of S100A4 in human osteosarcoma cell lines

resulted in a reduced in vitro and in vivo invasive and metastatic capacity [12,13] Furthermore, we also showed that the reduction in the endogenous level of S100A4 in these cell lines resulted in altered levels of MMP-2, MT1-MMP, TIMP-1 and TIMP-2, as well as active MMP-2 [13,14] Therefore, these cell lines were used in this study to investigate the extent to which synthesis and activation of proMMP-2, as well as syn-thesis of MT1-MMP, TIMP-1 and TIMP-2, are affected by the interaction of the cell with various bio-logical forms of collagen I A fibrillar 3D lattice and a 2D layer of monomeric collagen I will, to a certain extent, mimic the natural environment of osteosarcoma cells and were used in this study as an in vitro model

to study effects of cell–collagen I interactions

Results

Cell morphology and actin cytoskeleton structure

As observed by light microscopy (data not shown), pHb-1 and II-11b cells attached to plastic or mono-meric 2D collagen were spread in a confluent cell layer and hence showed maximum cell–cell contact For cells seeded on a fibrillar 3D collagen matrix, the cells were rounded up and seemed to have a more spherical shape; hence they were separated from most adjacent cells, but were still attached to the surface

Confocal microscopy revealed no differences in actin cytoskeleton organization for cells attached to the dif-ferent surfaces In addition, whether the fibrillar 3D col-lagen gel was attached to or released from plastic, or whether the cells were attached on the top of or inside the fibrillar 3D matrix did not influence the organization

of the actin cytoskeleton (data not shown)

Cell viability

As shown in Fig 1, during 48 h incubation in serum-free medium only minor changes in the number of via-ble cells were observed for both cell lines attached to plastic However, when the cells were attached to monomeric 2D collagen, the number of viable cells increased, whereas for cells attached to fibrillar colla-gen a small reduction in viable cells was observed

S100A4 expression is not affected by the cells binding to collagen I

In order to confirm the difference in S100A4 levels between the two cell lines, western blot analyses were performed on cell lysates from cells attached to plastic

To ensure equal loading, the total amount of cellular

protein in the two cell lines was determined as

described in Materials and methods, and 81 lg of

pro-tein was added to each well The amount of S100A4 in

the II-11b cells was 1.2% of that in the pHb-1 cells

after 2 min exposure of the film and 12% after 5 min

exposure (Fig 2A)

The level of S100A4 protein expression did not

change when cells were attached to monomeric 2D or

fibrillar 3D collagen surfaces pHb-1 cells maintained

high S100A4 expression, whereas the II-11b cells

main-tained low S100A4 expression (Fig 2B) As shown in

Fig 2B, equal amounts of protein were loaded based

on equal amounts of actin

Binding of cells to fibrillar 3D collagen I results in

decreased proMMP-2 activation and reduced

MT1-MMP expression

A decrease in the total amount of MMP-2 (72, 64 and

62 kDa bands), as well as in the active MMP-2 forms

(64 and 62 kDa bands), was observed for both cell lines attached to fibrillar 3D collagen compared with cells attached to plastic (Fig 3A) Because gelatin zymography is a semiquantitative method, the amount

of proMMP-2 was also determined by ELISA As shown in Fig 3B, a large decrease in proMMP-2 was observed for cells attached to fibrillar collagen Reduc-tion in proMMP-2 activaReduc-tion occurred independent of whether the fibrillar 3D collagen was attached to or released from plastic, or whether the cells were attached on top of or inside the fibrillar 3D matrix (data not shown) Some of the synthesized MMP-2 was adsorbed to the fibrillar collagen, and the ratio of active to total MMP-2 was the same as that detected

in the conditioned medium (data not shown) Thus, adsorption may in part explain the reduction in the total amount of MMP-2 in the medium, but it does

0

100

200

Relative cell viability (%)

3 h

48 h

*

*

Fig 1 Cell viability Relative cell viability (mean ± SEM) for pHb-1

and II-11b cells attached to plastic (P), monomeric 2D collagen I

(2D) and fibrillar 3D collagen I (3D) *P < 0.05 for 48 h compared

with 3 h (n = 7).

A

S100A4

Actin

B

Time (min)

2 5

pH β II-11b pH β II-11b S100A4

20

40

Mr

(kDa)

Fig 2 Expression of S100A4 in pHb-1 and II-11b cells (A)

Deter-mination of S100A4 by western blotting of cell lysates from pHb-1

and II-11b cells attached to plastic, using a total protein

concentra-tion of 81 lgÆmL)1, and the blot was exposed to the film for 2 or

5 min (B) Western blot of cell lysates from pHb-1 and II-11b cells

attached to plastic (P), monomeric 2D collagen I (2D) and fibrillar

3D collagen I (3D) Actin was used as loading control St: molecular

mass standard.

pHβ-1 II-11b

P 2D 3D P 2D 3D

3

2

1

0

Mr

(kDa)

72

62

*

*

64

0 0.5 1.0

B

*

1.5

A

Fig 3 Expression of MMP-2 in serum-free media from pHb-1 and II-11b cells (A) Gelatin zymography of harvested media from pHb-1 and II-11b cells attached to plastic (P), monomeric 2D collagen I (2D) or fibrillar 3D collagen I (3D) Typical zymograms showing proMMP-2 (72 kDa), intermediate MMP-2 (64 kDa) and fully acti-vated MMP-2 (62 kDa) Box-plots illustrate the ratio of actiacti-vated to total MMP-2 Open boxes denote pHb-1 cells while filled boxes denote II-11b cells Lines inside the boxes indicate median values, and dotted lines illustrate mean values (n = 12) (B) Harvested media from pHb-1 and II-11b cells attached to plastic (P), mono-meric 2D collagen I (2D), or fibrillar 3D collagen I (3D) were analy-sed for proMMP-2 expression by ELISA Relative values (± SD) are adjusted for cell viability Open bars denote pHb-1 cells and filled bars denote II-11b cells (n = 3) *P < 0.05 compared with cells attached to plastic.

not explain the reduction in active forms Because both

cell lines showed reduction in active forms, S100A4

did not influence this alteration in activation

Western blots of cell lysates showed reduced

expres-sion of MT1-MMP for both cell lines when attached

to fibrillar 3D collagen compared with cells attached

to plastic (Fig 4A) The level was reduced to 34% and

33% in pHb-1 and II-11b cells, respectively

Cell binding to monomeric 2D collagen I

increased the amount of fully activated MMP-2

For cells attached to both plastic and monomeric 2D

collagen, pHb-1 cells produced more of the activated

forms of MMP-2 than did II-11b cells (Fig 3A) For

both cell lines, attachment to 2D monomeric collagen

increased the amount of fully activated MMP-2

(62 kDa) and decreased the amount of the intermediate

activated form (64 kDa) compared with cells attached

to plastic (Figs 3A and 5) Although the amount of

activated forms varied between experiments, the

64 kDa intermediate form was always weaker when

cells were attached to monomeric 2D collagen than

when cells were attached to plastic (Fig 5) However,

the ratio of activated forms to total MMP-2 was

approximately the same for cells attached to monomeric

collagen and to plastic (Fig 3A) As shown by ELISA

in Fig 3B, a decrease in proMMP-2 was observed when

cells were attached to momomeric collagen

Western blots of cell lysates showed only small changes in MT1-MMP expression for both cell lines (+3% for pHb-1 and)10% for II-11b) when attached

to monomeric 2D collagen compared with cells attached to plastic (Fig 4A) Neither of the cell lines showed any difference in TIMP-2 expression when comparing cells attached to plastic and to monomeric collagen (Fig 4B) This indicates that TIMP-2 was not involved in the observed difference in activation

Cell binding to monomeric 2D collagen I results

in an S100A4-dependent decrease in TIMP-1 expression

Approximately twice as much TIMP-1 was secreted into the medium from pHb-1 cells compared with II-11b cells when attached to plastic (5.0 versus 3.1 lgÆmL)1Æ106cells)1) (Fig 6) The difference was sus-tained when cells were attached to fibrillar 3D colla-gen However, when cells were attached to monomeric 2D collagen, pHb-1 cells produced significantly less TIMP-1 (2.8 lgÆmL)1Æ106cells)1), whereas the level

MT1

Actin

40

50 50 60

(kDa)

Actin

TIMP-2

20

40

(kDa)

A

Fig 4 The expression of MT1-MMP and TIMP-2 in pHb-1 and

II-11b cells Determination of MT1-MMP (A) and TIMP-2 (B) by

wes-tern blot of cell lysates from pHb-1 and II-11b cells (A) Cells were

attached to plastic (P), monomeric 2D collagen I (2D) and fibrillar

3D collagen I (3D) Quantification of two blots gave mean values

for pHb-1 cells: P = 100%, 2D = 103%, 3D = 34% and for II-11B

cells: P = 100%, 2D = 90%, 3D = 33% (B) Cells were attached to

plastic (P) and monomeric 2D collagen I (2D) Actin was used as

loading control St: molecular mass standard.

pHβ-1

P 2D

II-11b

P 2D

72

62

64

72

62

64

Mr

(kDa)

Fig 5 The effect of monomeric 2D collagen I on MMP-2 activa-tion Gelatin zymography of harvested media from pHb-1 and II-11b cells attached to plastic (P) and monomeric 2D collagen I (2D) Typi-cal zymograms showing proMMP-2 (72 kDa), intermediate MMP-2 (64 kDa) and fully activated MMP-2 (62 kDa).

0

2

4

6

–1 ·10

6 cells

*

*

Fig 6 The effect of collagen I on TIMP-1 synthesis from pHb-1 and II-11b cells The cells were either attached to plastic (P), mono-meric 2D collagen I (2D) or on the top of fibrillar 3D collagen gel I (3D) Harvested media were analysed for TIMP-1 expression by ELISA Open bars denote pHb-1 cells and filled denote for II-11b cells Mean values ± SEM are adjusted for cell viability (n = 6).

*P < 0.05 compared with cells attached to plastic.

was reduced to 2.3 lgÆmL)1Æ106cells)1 for the II-11b

cell line (Fig 6)

TIMP-1 prevents the second step in the

MT1-MMP-induced activation of proMMP-2

Previously, it was shown that TIMP-1 prevents the

p-aminophenylmercuric acid (APMA)-induced

auto-activation of proMMP-2 and other MMPs [15–17] In

order to test whether TIMP-1 inhibits the second step

in the MT1-MMP-induced activation of proMMP-2,

commercial recombinant proMMP-2 was activated for

24 h at 37C, either with membranes isolated from

colchicines-stimulated pHb-1 cells containing a high

level of MT1-MMP [18] or with a commercial

recom-binant soluble form of MT1-MMP containing only the

catalytic domain As shown in Fig 7, both the

mem-brane fraction and the recombinant soluble

MT1-MMP activated proMT1-MMP-2 As expected, TIMP-1 did

not inhibit the first step where MT1-MMP cleaves the

72 kDa proMMP-2 form into the 64 kDa intermediate

However, TIMP-1 inhibited the second step that is an

autoactivation of the intermediate 64 kDa form to the

fully active 62 kDa enzyme Further, autoactivation of

the active 62 kDa form to a C-terminally truncated

45 kDa form was also inhibited by TIMP-1

Investigation of mechanisms that may explain

the increased activation of MMP-2 when cells are

attached to monomeric collagen I

To investigate whether TIMP-1 affected MMP-2

acti-vation, recombinant TIMP-1 was added to cells

attached to monomeric 2D collagen As shown in

Fig 8A, the intermediate 64 kDa band is stronger in the presence of TIMP-1 than in the absence of the inhibitor In order to determine whether other previ-ously described mechanisms are also involved, several experiments were performed First, we studied whether

an interaction between either the pro (72 kDa) or intermediate (64 kDa) MMP-2 and the underlying col-lagen layer caused the formation of a fully activated

62 kDa form of the enzyme Harvested media from cells attached to plastic were incubated for 24 h at

37C in culture wells with or without monomeric col-lagen Neither of these two conditions altered the rela-tive amount of the activated forms, indicating that binding of the pro (72 kDa) or intermediate (64 kDa) forms of MMP-2 to collagen did not result in enhanced autoactivation (data not shown) Second, we wanted to investigate whether the cysteine proteinase legumain is involved in the activation The presence of this proteinase was shown in both cell lines, but incu-bation on monomeric 2D collagen did not change its level (data not shown) Third, treatment of cells attached to plastic or monomeric 2D collagen with the cysteine proteinase inhibitors egg white cystatin, E-64

or E-64d showed no effect on the synthesis and activa-tion of proMMP-2 (Fig 8B) Fourth, in order to determine whether the observed activation took place intracellularly or extracellularly, surface proteins of cells attached to plastic and 2D collagen were labelled with biotin and removed as described in Materials and methods The unlabelled intracellular fraction of pro-teins were analysed by gelatin zymography No active MMP-2 was detected, demonstrating that the activa-tion occurred outside the cell (data not shown) Fifth,

24 h 37 °C

rMT1-MMP 0

0.5 0.5

0 1.0 2.0 1.0 2.0

Membr.

[TIMP-1]

[MMP-2]

72

62

64

Fig 7 Activation of proMMP-2 by isolated cell membranes and

MT1-MMP in the presence of TIMP-1 Human recombinant

proM-MP-2 (3 lgÆmL)1; 42 n M ) was incubated for 24 h at 37 C with

membranes isolated from colchicine treated pHb-1 cells or

recombi-nant human MT1-MMP catalytic domain in the presence of

increas-ing concentrations of TIMP-1 as described in Materials and

methods and analysed by gelatin zymography As controls, the

proMMP-2 alone was either dirctly added to loading buffer without

incubation or after 24 h incubation at 37 C.

TIMP-1 (ng·well–1) 0 0 75

P 2D

A

C E64 E64d Cys

2D

B

72

62

64

72

62

64

Fig 8 The effect of added inhibitors on MMP-2 activation Typical gelatin zymograms showing (A) the effect of TIMP-1 on the activa-tion of MMP-2 using pHb-1 cells attached to monomeric 2D colla-gen (2D) compared with pHb-1 cells attached to plastic (P), (B) the effect of cystein proteinase inhibitors E-64, E-64d and cystatin (Cys) on the activation of MMP-2 using pHb-1 cells attached to monomeric 2D collagen (2D) C, control without added inhibitors.

we investigated whether new activators on cell

mem-branes from cells attached to monomeric 2D collagen

could be responsible for the increased activation of

proMMP-2 Isolated cell membranes from cells

attached to plastic and monomeric 2D collagen did

not show any difference in their capacity to activate

proMMP-2 at pH 8.0 or at pH 5.8, indicating no new

activators at the cell membrane of cells attached to the

monomeric collagen

Discussion

The two osteosarcoma cell lines used in this study have

characteristics suggesting they are of osteoblastic origin

[19,20] In bone, active osteoblasts are embedded in a

3D ECM, whereas fully mature osteoblasts flatten out

and line quiescent bone surfaces Collagen I is the

main component of ECM, and interactions between

the osteosarcoma cell lines and the 3D fibrillar

colla-gen I network or the 2D layer of monomeric collacolla-gen

I will, to a certain extent, mimic the in vivo situation

for these cells Hence, this model can reveal to what

extent cell–collagen I interactions will affect synthesis

and activation of MMPs and their inhibitors Further,

this model will discover whether the endogenous level

of the metastasis-associated protein S100A4 is of

importance for how the cells respond to interactions

with these two forms of collagen I

Included among cells that have been shown to

increase the activation of proMMP-2 when they

inter-act with fibrillar 3D collagen I are various human

tumour cell lines [21,22], human skin fibroblasts

[21,23–27], human umbilical vein and neonatal foreskin

endothelial cells [28], human fetal lung fibroblasts [29],

human hepatic stellate cells [30], rat capillary

endothe-lial cells [31] and rat cardiac fibroblasts [32] In most

cases, the increased activation of proMMP-2 is shown

to be associated with an increase in MT1-MMP

Fur-thermore, cells have a changed morphology when

attached to fibrillar 3D collagen compared with the

same cells attached to a 2D surface such as monomeric

collagen or plastic [21–23,29–31,33] This was also

shown for the osteosarcoma cells in our study where

the cells appeared more rounded in shape In human

skin fibroblasts there is an increase in proMMP-2

acti-vation that is independent of the lattice contraction

[23] It has also been shown that cells grown on

fibril-lar 3D collagen I attached to a plastic surface contain

actin stress fibres, whereas stress fibres are missing

when collagen is released from the surface Only under

conditions where the cells lack stress fibres, do

fibro-blasts produce increased amounts of activated MMP-2

[29] A lack of stress fibres is also necessary for the

increase in the activation of proMMP-2 in smooth muscle endothelial cells [34] In several of the studies referred to above, it has also been shown that treat-ment of cells attached to a planar substrate (such as monomeric collagen or plastic) with compounds that dissolve actin stress fibres (cytochalasin D, vascular endothelial growth factor), results in increased activa-tion of proMMP-2 However, treating cells with com-pounds that dissolve the tubulin network (colchicine, nocodazole) did not induce proMMP-2 activation In contrast to this, we have previously shown that colchi-cine-induced rearrangements of the microtubule network in osteosarcoma cell lines increase activation

of proMMP-2 along with an increased level of MT1-MMP [18] This study shows that the interaction between osteosarcoma cells and fibrillar 3D collagen reduces the activation of proMMP-2 because of a decrease in MT1-MMP The reduction was indepen-dent of whether the 3D collagen lattice was attached

to plastic or not In contrast to the cell lines discussed above, the actin cytoskeleton in the osteosarcoma cells was not affected by the surface the cells were attached

to Hence, the reduction in MT1-MMP and active forms of MMP-2 could not be attributed to changes in the actin cytoskeleton This adds to previous investiga-tions showing that these osteosarcoma cell lines respond differently to various stimuli compared with other cells

We also show that the osteosarcoma cells produce an increased amount of fully active MMP-2 when bound

to 2D monomeric collagen (Figs 3A and 5) which is another example of a different characteristic trait of these cells compared with fibroblasts and endothelial cells No drastic change in the amount of MT1-MMP was observed in osteosarcoma cells attached to mono-meric collagen compared with plastic Although MT1-MMP here may participate in the activation of proMMP-2, it cannot account for the increased amount

of fully activated enzyme MT1-MMP induces the con-version of proMMP-2 to the intermediate 64 kDa form

by cleaving the Asn37–Leu38 bond [35,36], whereas the

64 kDa intermediate is further processed to the fully activated 62 kDa species in an autoactivation step In this study, various experiments were performed to determine whether one of the following mechanisms was responsible for the increase in fully activated MMP-2 when cells were attached to monomeric colla-gen: (a) increased expression of an activator enzyme that cleaves proMMP-2 in or near the autocatalytic site (Asn80–Tyr81), (b) increased expression of a factor that stimulates the second step of the MT1-MMP induced activation, or (c) reduced expression of an inhibitor of the second step of the activation

The two osteosarcoma cell lines used in this study

produce legumain, a potential intracellular activator of

proMMP-2 However, because we were not able to

detect any intracellular fully activated MMP-2, nor

could we inhibit MMP-2 activation with E-64, we can

rule out legumain or other lysosomal cysteine

protein-ases as activators when the cells are attached to

mono-meric 2D collagen Furthermore, there was no

difference in activation of exogenously added

recombi-nant proMMP-2 between isolated cell membranes from

cells attached to plastic and monomeric 2D collagen,

irrespective of the pH used This result excludes the

existence of some enzyme in membranes from cells

attached to monomeric collagen that cleaves

proMMP-2 either in or close to the autoactivation site Nor

could the activation be explained by autoactivation

caused by a direct binding of either the 72 kDa

pro-form or the 64 kDa intermediate to the cell membrane

Lafleur et al [6] have shown that TIMP-2, in addition

to being involved in the first step in the MT1-MMP

activation of MMP-2, can also take part in the second

autoactivation step and function as an activator by

promoting conversion of the 64 kDa intermediate to

the fully active 62 kDa form In this study, we did not

find any difference in the levels of TIMP-2 when cells

were attached to plastic compared with monomeric 2D

collagen (Fig 4B), hence TIMP-2 is not the cause of

the observed difference in activation level of MMP-2

Our results rule out the two first alternatives, (a) and

(b), as an explanation for the increased activation

when cells are attached to monomeric 2D collagen

However, alternative (c), reduced expression of an

inhibitor of the second step of the activation, seemed

to be an explanation We have shown that TIMP-1 is

a regulator of the second step in the activation of

proMMP-2 using both recombinant MT1-MMP and

isolated cell membranes rich in MT1-MMP (Fig 7)

This is consistent with previous observations that

TIMP-1 inhibits autoactivation of several MMPs such

as: the Ca2+-induced intramolecular autoactivation of

proMMP-9 covalently linked to the core protein of a

chondroitin sulfate proteoglycan [37]; the

APMA-induced autoactivation of MMP-9 to the 80 kDa

inac-tive intermediate and the 68 kDa acinac-tive species, where

TIMP-1 prevented the formation of the latter species

[15,38,39]; APMA-induced autoactivation of

proMMP-2 [15]; APMA-induced autoactivation of proMMP-3

and N-terminally truncated proMMP-3 [15,40]; and

APMA-induced autoactivation of proMMP-1 and

proMMP-8 [16,17] At the cellular level, we have

shown that exogenously added TIMP-1 increased the

intermediate 64 kDa form of MMP-2 (Fig 8A),

indi-cating that the decreased level of TIMP-1 is the main

cause of increased activation of MMP-2 when cells were attached to monomeric 2D collagen Altogether, our results show that endogenously produced TIMP-1 can act as a modulator of the MT1-MMP-induced activation of proMMP-2

The interaction between cells and fibrillar or mono-meric collagen, respectively, showed opposite effects

on proMMP-2 activation This effect was independent

of the endogenous level of S100A4 in the two cell lines

By contrast, the expression of TIMP-1 was dependent

on the cell endogenous level of S100A4 When cells were attached to plastic and fibrillar 3D collagen, those with a high endogenous level of S100A4 pro-duced approximately twice as much TIMP-1 as those with a reduced S100A4 level However, when cells were attached to monomeric 2D collagen, the production of TIMP-1 from cells with a high level of S100A4 was reduced to approximately the same amount as from cells with a low S100A4 level This suggests that the interaction between cells and monomeric 2D collagen causes a block in the S100A4-induced pathway that upregulates TIMP-1 expression Taken together, our results show that osteosarcoma cells interact with two types of collagen I found in vivo, and the form of the collagen determines the cells synthesis and activation

of MMP-2 as well as the synthesis of MT1-MMP and TIMP-1

Previously, it has been shown that the reduced level

of S100A4 in the II-11b cells compared with pHb-1 cells resulted in a large reduction of in vivo and in vitro invasive capacity, as well as in vitro motility [12,13] The reduction in S100A4 also resulted in a decrease in the expression of MT1-MMP, TIMP-1 and MMP-2 at both the mRNA and protein levels, in addition to a decreased amount of activated MMP-2 [13,14] MMPs and TIMPs are associated with cell invasion and metastasis, although their role is dual [41–43] Both MMPs and TIMPs, as well as the in vivo substrates of

a given MMP, can prevent or facilitate the invasion and metastasis process, depending on the time and localization of their expression The N-terminal part of TIMPs is involved in binding to the active site of MMPs and hence prevents their action, whereas the C-terminal part can bind to proteins in the cell mem-brane and modulate cell growth and viability indepen-dent of MMPs [44,45] One of the aims of current research on MMPs and TIMPs in cancer is to establish the localization and timeframe for their expression, as well as the identification of the in vivo substrates of individual MMPs It is thus important to discover how each ECM component in the microenvironment of a given cancer type affects expression and activation of MMPs and TIMPs Our investigation shows that two

forms of the main ECM component in the

microenvi-ronment of osteosarcoma cells differently affect their

expression of MMPs and TIMPs as well as their

acti-vation of MMP-2 In addition, it adds to the earlier

investigations by showing that the structure of the

sur-rounding matrix components will be of importance for

the effect of S100A4 on MMP and TIMP regulation

and thereby on the cell’s possibility to promote

inva-sion and metastasis

Materials and methods

Materials

DMEM containing Hams F12 medium, penicillin,

strepto-mycin, geneticin disulfate salt (G418), gelatin 300 Bloom

type A (from porcine skin), egg white cystatin (C-0408),

BSA, Igepal CA-630, bactrerial collagenase and Sigma

serum replacement 1 were all from Sigma-Aldrich (St

Louis, MO, USA) Fetal bovine serum was from Biochrom

AG (Berlin, Germany), l-glutamine was from Gibco BRL

Life Technologies (Paisley, UK), nonessential amino acids

(100·) were from PAA Laboratories GmbH (Pasching,

Austria), sterile rat-tail collagen I was from Roche

Diag-nostics GmbH (Basel, Switzerland), poly(vinylidene

difluo-ride) membranes were from Millipore (Bedford, MA,

USA), Alexa Fluor 568-labelled phalloidin (A12380), Magic

gels were from Invitrogen Life Technologies (Carlsbad, CA,

USA), Biotrac TIMP-1 and MMP-2 ELISA were from

Amersham Biosciences (Little Chalfont, UK) and E-64

(N-1645) and E-64d (N-1650) were from Bachem

(Buben-dorf, Switzerland) Western Blotting Luminol Reagent was

from Santa Cruz Biotechnology (Santa Cruz, CA, USA),

EZ-Link Sulfo-NHS-LC-LC biotin, streptavidin agarose

resins and halt protease inhibitor cocktail were from Pierce

Biotechnology (Rockford, IL, USA) The following

anti-bodies were used; GAPDH rabbit mAb and pan-actin

rabbit polyclonal from Cell Signaling Technology (Danvers,

MA, USA), MT6-MMP mouse mAb from R&D systems

(Minneapolis, MN, USA), TIMP-2 and MT1-MMP rabbit

polyclonal from Panomics (Redwood City, CA, USA),

S100A4 rabbit polyclonal from Abcam (Cambridge, UK)

and RECK mouse mAb from BD Biosciences (San Jose,

CA, USA) Anti- mouse and anti-rabbit IgG horseradish

peroxidase-linked antibodies were from Cell Signalling

Technology TIMP-1 was from Oncogene Research

Prod-ucts (Boston, MA, USA), purified human recombinant

proMMP-2 and MT1-MMP (catalytic domain) were from

Calbiochem (San Diego, CA, USA) Solution cell

prolifera-tion assay (Cell Titer 96AQueous One) was from Promega

(Madison, WI, USA) Paraformaldehyde was purchased

from Merck (Darmstadt, Germany) and Triton X-100 from

BDH Biochemicals Ltd (Poole, UK)

Cell cultures

The highly metastatic osteosarcoma cell line, OHS, was established from a bone tumour biopsy from a patient trea-ted at the Norwegian Radium Hospital [46] The OHS cell line was transfected with a vector encoding a S100A4-spe-cific ribozyme or, as a control, with the vector alone [12] The ribozyme-transfected clone was designated II-11b, and the control cell clone transfected with the vector alone was designated pHb-1 The II-11b cell line had a reduced level

of S100A4 and a decreased metastatic capacity, whereas the pHb-1 cell line maintained the S100A4 expression level and metastatic properties of the parental OHS cell line [12] Transfectants were subcultivated in a 1 : 1 mixture of DMEM and Hams F12 medium (basal medium) containing

l-glutamine, nonessential amino acids (100· dilution),

37C

Preparation of wells for cell experiments

Cells (6.0· 104) were, in addition to plastic, attached to monomeric 2D collagen I and on top and inside fibrillar lattices of 3D collagen I, all in 0.33 cm2wells Ten

spread into the wells to prepare monomeric collagen I The wells were dried for 2 h, followed by washing in serum-free medium (culture medium in which fetal bovine serum was replaced by 2% serum replacement) and left with 50 lL serum-free medium for 20 min 3D fibrillar collagen I gels were prepared by adding 50 lL neutralized collagen I solu-tion (7 : 1 : 1 : 1 of each 3 mgÆmL)1collagen I in 0.2% ace-tic acid, 10· serum-free medium, 1.0 m Hepes, pH 7.3, and 0.33 m NaOH, respectively) to the wells After 2 h of

medium for 20 min and the medium was removed before cells were added in a new aliquot of medium For cell attachment inside 3D collagen I gels, cells were mixed with

50 lL neutralized collagen I solution and left for 2 h while polymerization took place

Cell viability assay

To determine the viability of the cells, trypsinized cells were suspended in serum-containing medium In order to remove serum, cells were washed three times with serum-free med-ium prior to seeding on the different plate surfaces as described above A 100-lL cell suspension was added to plastic and monomeric 2D collagen, whereas 50 lL serum-free medium with or without cells was added to the 3D

absorbance at 490 nm was determined according to the

instructions of the manufacturer’s protocol Prior addition

of cell viability reagent to wells containing fibrillar 3D

colla-genase in 0.1 m Hepes pH 7,5 containing 0.9% NaCl) was

activity For each surface cells were attached to, standard

curves were made to assure a linear response between cell

number and absorbance

Phalloidin staining of actin filaments

Cells were cultured on glass covers slides, on monomeric

2D collagen I, on top of or inside fibrillar 3D collagen I for

24 h in serum-free media Cells were fixed in a final

concen-tration of 4% paraformaldehyde for 10 min on ice, washed

labelling with 1 unit of Alexa Fluor 568-conjugated

phalloi-din in 1% BSA for 30 min at room temperature Cells were

obtained using a 40· water objective on a LSM 510 META

confocal laser-scanning inverted microscope (Carl Zeiss

International, Go¨ttingen, Germany)

Production of conditioned media for MMP and

TIMP determination

To determine the secretion of MMP-2 and TIMP-1 into the

media, trypsinized cells were suspended in serum-containing

medium In order to remove serum, cells were washed three

times with serum-free medium prior to seeding on the

dif-ferent plate surfaces, as described above A 100 lL cell

sus-pension was added to plastic and monomeric 2D collagen,

whereas 50 lL serum-free medium with or without cells

was added to the 3D collagen gels After 48 h incubation in

harvested Prior to freezing, the harvested media was

test whether TIMP-1 or legumain and other lysosomal

cys-teine proteinases were involved in proMMP-2 activation,

cells were attached to plastic and 2D collagen I surfaces

10 lm E-64, 10 lm E-64d or 1 lm egg white cystatin)

Isolation of cell membranes

Cells (1.4· 107

Petri dishes, uncoated or coated with 2D monomeric

the conditions described above Plasma membranes were

prepared as previously described [18,47] Production and

attached to plastic and treated with colchicine under serum free conditions were performed as previously described [18]

Isolation of cell lysates

To compare the level of S100A4 in the two cell lines

4000 g Cells were then sonicated and the lysate centrifuged

The amount of cellular protein was detected by the Brad-ford method (Bio-Rad, Hercules, CA, USA), using BSA as

a standard

Activation of MMP-2 by cell membranes

Membrane-mediated activation of human MMP-2 was

analysed by gelatin zymography Activation experiments using cell membranes from colchicine stimulated pHb-1 cells were performed as described previously [14,18], with and without recombinant TIMP-1 present

Activation of MMP-2 by recombinant MT1-MMP

Activation of MMP-2 by MT1-MMP was performed by

with human recombinant MT1-MMP catalytic domain

were applied to gelatin zymography

Biotinylation of cell-surface proteins

Cells (2.0· 106

) were seeded in serum-free media in six-well plates either uncoated or coated with monomeric 2D

and kept under the conditions described above Condi-tioned media were removed and wells were washed with cold NaCl⁄ Pi(pH 8) To release cells, NaCl⁄ Pi was added

16 lL of 10 mm biotinreagent solution was added The reaction mixture was incubated at room temperature for

30 min Cells were washed with NaCl⁄ Pi containing

100 mm glycine to remove excess biotin reagent and

coc-tail) To remove biotinylated surface proteins, cell lysate

was added to a streptavidin agarose resin and incubated for

30 min After centrifugation, only unlabelled intracellular

proteins were found in the supernatant

Gelatin zymography

Conditioned medium was mixed with loading buffer

blue and 50% glycerol) and loaded onto a 10% gelatin gel

To determine the amount of gelatinase in the harvested 3D

collagen I gels, an equal volume of gel and loading buffer

were mixed and left for 30 min at room temperature prior

was applied to the zymography gel To determine whether

gelatinases bind to monomeric 2D collagen, 10% dimethyl

sulfoxide in serum-free media was added to wells after

removal of conditioned medium, and the extract was added

loading buffer and applied to the gelatin gel Gelatin

zymography gels were run, washed and stained as described

previously [14] Gelatinolytic activity was evident as

trans-parent zones in the blue gels The area of the cleared zones

was analysed using the genetools program from SynGene

(Cambridge, UK)

Western blot analysis

After removal of conditioned media, cells were lysed in gel

loading buffer containing 0.1 m dithiothreitol, and boiled

for 5 min Samples were electrophoresed on acrylamide

gra-dient gels (4–12%) and proteins were transferred to a

poly(vinylidene difluoride) membrane by electroblotting

After blocking nonspecific binding sites with non-fat milk

(5% solution), blots were incubated for 1 h at room

tem-perature with primary antibodies against S100A4,

MT1-MMP or TIMP-2 After washing, the blots were incubated

for 1 h at room temperature with horseradish

peroxidase-conjugated secondary antibodies diluted in blocking

solu-tion, and developed using a western blotting luminol

reagent The membranes were washed, blocked and

rep-robed for the detection of actin or GAPDH The amount

of protein in the detected spots were analysed on either the

Prod-ucts (Cambridge, UK) or the genetools program

ELISA

The levels of TIMP-1 and MMP-2 were determined from

serum-free conditioned media, according to manufacturer’s

instructions The TIMP-1 assay recognizes total human

TIMP-1, i.e free TIMP-1 and TIMP-1 bound to MMPs The MMP-2 assay recognizes proMMP-2 and proMMP-2 bound to TIMP-2, but not the active form of MMP-2

Legumain activity

Legumain was measured by recording the cleavage of the substrate Z-Ala-Ala-Asn-NHMec (Department of

described [48,49] Twenty microlitres of cell lysate were added to black 96-well microtiter plates (Costar) After the addition of 100 lL buffer and 50 lL substrate solution (10 lm final concentration), a kinetic measurement based

on increase in fluorescence over 10 min was performed

performed in triplicate

Statistics

Statistical analyses were performed using the student t-test

mean ± SD (gelatin zymography, western blotting and ELISA data) A P-value < 0.05 was considered significant Analyses were based on three or more independent cell cul-ture experiments Conditioned medium from each experi-ment was run in duplicate on gelatin zymography, ELISA and western blots

Acknowledgements

This work was supported in part by grants from The Norwegian Cancer Society and the Erna and Olav Aakre Foundation for Cancer Research We are grate-ful to Dr Peter McCourt for linguistic revision of the manuscript

References

1 Gelse K, Poschl E & Aigner T (2003) Collagens – struc-ture, function, and biosynthesis Adv Drug Deliv Rev 55, 1531–1546

2 Murphy G & Nagase H (2008) Progress in matrix metalloproteinase research Mol Aspects Med 29, 290–308

3 Bjorklund M & Koivunen E (2005) Gelatinase-mediated migration and invasion of cancer cells Biochim Biophys Acta 1755, 37–69

4 Chen JM, Fortunato M, Stevens RA & Barrett AJ (2001) Activation of progelatinase A by mammalian legumain, a recently discovered cysteine proteinase Biol Chem 382, 777–783

5 Nie J & Pei D (2003) Direct activation of pro-matrix