Content uploaded by Pia Nyberg
Author content
All content in this area was uploaded by Pia Nyberg on Oct 28, 2016
Content may be subject to copyright.
Generation of biologically active endostatin fragments from human
collagen XVIII by distinct matrix metalloproteases
Ritva Heljasvaara
a
, Pia Nyberg
b
, Jani Luostarinen
a
, Mataleena Parikka
b
, Pia Heikkila¨
c
,
Marko Rehn
a
, Timo Sorsa
c
, Tuula Salo
b
, Taina Pihlajaniemi
a,
*
a
Collagen Research Unit, Biocenter Oulu and Department of Medical Biochemistry and Molecular Biology,
University of Oulu, PO Box 5000, FIN-90014 Oulu, Finland
b
Departments of Diagnostics and Oral Medicine, University of Oulu, FIN-90014 Oulu, Finland
c
Department of Oral and Maxillofacial Diseases, Helsinki University Central Hospital, Institute of Dentistry,
University of Helsinki, FIN-90014 Helsinki, Finland
Received 31 January 2005, revised version received 4 March 2005
Available online 28 April 2005
Abstract
Endostatin, a potent inhibitor of endothelial cell proliferation, migration, angiogenesis and tumor growth, is proteolytically cleaved from
the C-terminal noncollagenous NC1 domain of type XVIII collagen. We investigated the endostatin formation from human collagen XVIII by
several MMPs in vitro. The generation of endostatin fragments differing in molecular size (24 –30 kDa) and in N-terminal sequences was
identified in the cases of MMP-3, -7, -9, -13 and -20. The cleavage sites were located in the protease-sensitive hinge region between the
trimerization and endostatin domains of NC1. MMP-1, -2, -8 and -12 did not show any significant activity against the C-terminus of collagen
XVIII. The anti-proliferative effect of the 20-kDa endostatin, three longer endostatin-containing fragments generated in vitro by distinct
MMPs and the entire NC1 domain, on bFGF-stimulated human umbilical vein endothelial cells was established. The anti-migratory potential
of some of these fragments was also studied. In addition, production of endostatin fragments between 24 –30 kDa by human hepatoblastoma
cells was shown to be due to MMP action on type XVIII collagen. Our results indicate that certain, especially cancer-related, MMP family
members can generate biologically active endostatin-containing polypeptides from collagen XVIII and thus, by releasing endostatin
fragments, may participate in the inhibition of endothelial cell proliferation, migration and angiogenesis.
D2005 Elsevier Inc. All rights reserved.
Keywords: Endostatin; Collagen XVIII; Matrix metalloproteases; Endothelial cell proliferation; Endothelial cell migration; Hepatoblastoma
Introduction
Angiogenesis, the formation of new capillary blood
vessels, plays an essential role in normal and pathological
processes, such as embryogenesis, wound healing and tumor
growth. Solid tumors cannot grow beyond a few millimeters
in diameter without generation of tumor vasculature, and the
disturbance of the delicate balance between the pro- and
antiangiogenic factors induces tumor neovascularization [1].
Many endogenous angiogenesis inhibitors are cryptic frag-
ments of larger extracellular matrix molecules that are not
antiangiogenic as intact molecules [2]. Endostatin, a 20-kDa
C-terminal proteolytic fragment derived from type XVIII
collagen [3,4] was originally isolated and characterized as a
specific inhibitor of endothelial cell proliferation from
conditioned murine hemangioendothelioma (EOMA) cell
media [5]. Thereafter, several studies have demonstrated that
systemic administration of recombinant endostatin, produced
0014-4827/$ - see front matter D2005 Elsevier Inc. All rights reserved.
doi:10.1016/j.yexcr.2005.03.021
Abbreviations: EOMA, hemangioendothelioma; MMP, matrix metal-
loprotease; ECM, extracellular matrix; BM, basement membrane; TIMP,
tissue inhibitor of matrix metalloproteinase; HepG2, hepatoblastoma cell line;
NC1, C-terminal noncollagenous domain 1of collagen XVIII; rhNC1,
recombinant human NC1; HUV-EC-C, human umbilical vein endothelial cell
line; APMA, p-aminophenylmercuric acetate; SFCM, serum-free culture
medium; HepG2-CM,HepG2-conditioned serum-free culture medium;bFGF,
basic fibroblast growth factor; VEGF, vascular endothelial growth factor.
* Corresponding author. Fax: +358 8 5375810.
E-mail address: taina.pihlajaniemi@oulu.fi (T. Pihlajaniemi).
Experimental Cell Research 307 (2005) 292 – 304
www.elsevier.com/locate/yexcr
and delivered by varying means, efficiently blocks angio-
genesis and suppresses primary tumor and metastasis growth
in experimental animal models (reviewed in [6 – 8]). Unex-
pectedly, the mice lacking type XVIII collagen do not have
abnormalities in transplantable tumor growth compared to the
normal mice indicating that the physiological level of
endostatin (40–100 ng/ml in mouse plasma) is not sufficient
to protect from tumors [9,10]. The exact molecular mecha-
nism(s) by which endostatin inhibits endothelial cell pro-
liferation, migration and angiogenesis have remained largely
unclear. Two recent studies shed light on this issue suggesting
that endostatin downregulates several cell cycle and signaling
pathways associated with antiangiogenic activities [11,12].
Among other effects, endostatin has been shown to reduce
endothelial cell motility by interfering with bFGF-induced
signal transduction [13], block the VEGF-mediated signaling
by directly interacting with the VEGF-R2 receptor [14],
inhibit endothelial cell migration by binding to integrin a5h1
and thereby disrupting the cell-matrix adhesion [15 – 17] and
impair the mobilization of the endothelial progenitor cells
[18].
The identification of multiple forms of endostatin in
human plasma [19 – 21] and mouse tissue extracts [20,22],
with molecular masses varying between 18 and 38 kDa,
suggests that several proteolytic cleavage mechanisms exist
for their generation from the native precursor type XVIII
collagen rather than just a single mechanism. The production
of mouse endostatin by EOMA cells was shown to involve
two steps [23]: a metal-dependent initial step results in the
formation of larger endostatin-containing fragments, which
are then further cleaved by elastase activity to a fragment with
the N-terminal sequence identical to that reported for func-
tional endostatin [5]. In addition, cathepsin L can generate
endostatin at low pH irrespective of metalloprotease activity
[24]; the same paper also indicating metalloprotease involve-
ment in the C-terminal processing of collagen XVIII. The
capability of elastase and several MMPs to process recombi-
nant human NC1 domain of collagen XVIII in vitro was later
confirmed [25], and MMP-7 was proved to cleave efficiently
corneal collagen XVIII to produce a 28-kDa endostatin-
containing fragment [26]. Cathepsins and elastase were also
shown to degrade recombinant human endostatin [25].
However, the specific cleavage sites and the biological
activity of the MMP-generated endostatin fragments were
not fully characterized.
Matrix metalloproteases (MMPs), a family of genetically
distinct but structurally related zinc-dependent extracellular
and membrane-associated endopeptidases, can collectively
degrade essentially all extracellular matrix (ECM) and
basement membrane (BM) components during angiogenesis,
and they also affect neovascularization by regulating
endothelial cell adhesion, proliferation and migration [27 –
29]. Both endogenous tissue inhibitors of matrix metal-
loproteases (TIMPs) and synthetic MMP inhibitors block
angiogenesis in vitro and in vivo [27,28]. Although MMPs
have traditionally been considered to be pro-angiogenic and
pro-tumorigenic, certain MMPs may also participate in the
inhibition of neovascularization by converting plasminogen
to angiostatin, which is another potent antiangiogenic protein
[30–32].
We set out to study the capacity of a number of human
MMPs to generate endostatin fragments from human collagen
XVIII in vitro. Full-length collagen XVIII isolated from a
human hepatoblastoma cell line HepG2 and a recombinant C-
terminal noncollagenous domain NC1 of human collagen
XVIII (rhNC1) were used as substrates. MMP-3, -7, -9, -13
and -20 were shown to generate endostatin-containing frag-
ments both from full-length collagen XVIII and from rhNC1.
Collagenases MMP-1 and MMP-8, as well as MMP-2 and
MMP-12, failed to show any significant activity against
collagen XVIII or rhNC1. MMP inhibitors were used to
demonstrate that formation of endostatin fragments in HepG2
cell system is dependent on MMP action on type XVIII
collagen. Furthermore, endothelial cell proliferation and
migration were measured to demonstrate the biological
activity of the E. coli-derived human endostatin and several
longer endostatin-containing fragments including the entire
NC1. Our results indicate that certain, mainly cancer-related,
MMPs may participate in the inhibition of endothelial cell
proliferation and angiogenesis by generating antiangiogenic
endostatin-containing peptides from collagen XVIII.
Experimental procedures
Materials
A human hepatoblastoma cell line HepG2 (HB8065) and a
human umbilical vein endothelial cell line HUV-EC-C (CRL-
1730) were purchased from the American Type Culture
Collection. A permanent human umbilical vein-derived
endothelial cell line EA.hy926 was kindly provided by Dr.
Cora-Jean Edgell (University of North Carolina, NC).
Human MMP-1, MMP-3, MMP-7 and proMMP-13 were
obtained from Chemicon International Inc. (Temecula, CA),
human proMMP-9 from Oncogene Research Products (Cam-
bridge, MA) and human MMP-12 from Elastin Products
Company, Inc. (Owensville, MO). In addition, proMMP-2
and proMMP-9 were purified from human gingival fibro-
blasts and human neutrophils, respectively, as described
elsewhere [33,34]. Human proMMP-8 was purified from
extracts of polymorphonuclear neutrophils by a previously
published protocol [35], human proMMP-3 was purified
from human synovial fibroblasts by a published protocol [36]
and human proMMP-13 was prepared as described elsewhere
[37]. The active human recombinant MMP-20 was a
generous gift from Dr. John D. Bartlett (Department of
Biomineralization, Forsyth Dental Center, Boston, MA), and
the rat proMMP-7 from Dr. J.F. Woessner Jr. (Department of
Biochemistry and Molecular Biology, University of Miami,
Florida). p-Aminophenylmercuric acetate (APMA) was
obtained from Sigma Chemicals Co (St. Louis, MO). MMP
R. Heljasvaara et al. / Experimental Cell Research 307 (2005) 292– 304 293
inhibitors marimastat (BB-2516) and MMP Inhibitor III were
purchased from British Biotech Pharmaceuticals Ltd
(Oxford, UK) and Calbiochem (San Diego, CA), respec-
tively. Recombinant human TIMP-1 was obtained from
Oncogene Research Products (San Diego, CA).
Cell culture
HepG2 cells were maintained in Dulbecco’s modified
Eagles’s medium (DMEM, Gibco BRL) supplemented with
10% fetal bovine serum (FBS), 2 mM l-glutamine and 1 mM
sodium pyruvate in a 5% CO
2
atmosphere at 37-C. When the
plates were 90% confluent, the cells were washed twice with
PBS and serum-free culture medium (SFCM) was added.
HepG2-conditioned SFCM (HepG2-CM) was collected after
72 h and used for collagen XVIII purification and to
demonstrate the production of MMPs by HepG2 cells.
HUV-EC-C cells were cultivated in modified Kaighn’s
F12K medium (Irvine Scientific, Santa Ana, CA) supple-
mented with 0.1 mg/ml heparin and 0.1 mg/ml endothelial
cell growth supplement containing bFGF (Sigma, St. Louis,
MO) and 10% FBS. EA.hy926 cells were cultured in DMEM
complemented with 10% FBS, 2 mM glutamine and HAT
additive containing 5 mM hypoxanthine, 20 AM aminopterin
and 0.8 mM thymidine (Sigma, St. Louis, MO).
Antibodies
Three polyclonal antibodies against human collagen
XVIII were used: anti-all huXVIII against the N-terminal
noncollagenous region of human collagen XVIII [38] and
ES2a and HES.6 against the C-terminal endostatin portion of
the molecule. The ES2a antiserum was raised against the
synthetic peptide EAPSATGQASSLL, derived from human
endostatin according to the manufacturer’s protocol (Inno-
vagen, Lund, Sweden). The rabbit polyclonal antibody
HES.6 to recombinant human endostatin was raised by
conventional methods. The production and purification of
recombinant human endostatin have been described else-
where [15]. Purified antigen in complete Freund’s adjuvant
(Sigma, St. Louis, MO) was injected into a rabbit subcuta-
neuosly followed by booster injections at intervals of 2
weeks. All the polyclonal antisera were affinity-purified on
columns with the corresponding antigen coupled to CNBr-
activated Sepharose 4B (Amersham Pharmacia Biotech,
Uppsala, Sweden) as described elsewhere [38]. The specifici-
ties of the ES2a and HES.6 antisera were verified by Western
blotting against recombinant human endostatin, human
collagen XVIII isolated from HepG2 cells and recombinant
human endostatin XV (Supplementary Fig. 1).
Expression and purification of C-terminal fragments of
human collagen XVIII
The cloning of the recombinant human endostatin used in
this work has been described earlier [15]. The cDNAs
encoding the C-terminal NC1 domain of human collagen
XVIII (rhNC1) and the 28-, 25- and 24-kDa endostatin-
containing fragments starting at residues Trp
89
,Tyr
107
and
Tyr
116
, respectively, were amplified by PCR from a human
a1 (XVIII) cDNA clone HP19.3 [39]. The forward primers
used were 5V-TCGGTACCTCAGGGGTGAGGCTCTGG-
3V(rhNC1), 5VCGACGTGGATCCTGGCGGGCAGA-
GGATCCTGGCGGGCAGATGACATCCTG-3V(Trp
89
),
5V-CGACGTGGATCCTACCCCGGAGCCCCGCACCAC-
3V(Tyr
107
)and5V-CTAGGGATCCTACGTGCACCTG-
CGGCCG-3V(Tyr
116
) and the reverse primers were 5V-
ATAAGCTTACTTGGAGGCAGTCATGAA-3V(rhNC1)
and 5V-CTAGAAGCTTCTACTTGGAGGCAGTCAT-
GAA-3V(Trp
89
,Tyr
107
and Tyr
116
). The PCR fragments
were cloned into the KpnI/Hin dIII (rhNC1) or BamHI/
HindIII (Trp
89
,Tyr
107
and Tyr
116
) site of the expression
vector pQE-30 (Qiagen), which contains an N-terminal His-
tag to facilitate purification. The identity of the PCR-
amplified fragments was verified by sequencing. The
resulting clones were transformed into the E. coli strain
M15(pRep4), and recombinant proteins were expressed
according to the protocol suggested by Qiagen. For
proliferation assay, the three His-tagged recombinant
proteins were purified according to a previously described
protocol [15]. When purifying rhNC1 for N-terminal
microsequence analysis, the heparin Sepharose and Poly-
myxin columns were omitted.
Isolation of human collagen XVIII from HepG2 cells
Conditioned SFCM from HepG2 cell cultures was
collected after 72 h of incubation and subjected to
heparin Sepharose purification (Amersham Pharmacia
Biotech, Uppsala, Sweden). Collagen XVIII was eluted
from the column with 20 mM Tris buffer, pH 7.5,
containing 1 M NaCl. The heparin affinity-purified
collagen XVIII was concentrated by ultracentrifugation
(Ultrafree, MWCO 30 kDa, Millipore, Billerica, MA),
and the buffer was changed to 50 mM Tris–HCl, pH 7.8,
0.2 M NaCl, 1 mM CaCl
2
using a desalting chromatog-
raphy column (Bio-Rad, Hercules, CA). The total protein
concentration was measured using the Roti
\
-Quant
protein assay (Carl Roth GmbH+Co, Karlsruhe, Ger-
many). CompleteiEDTA-free protease inhibitor cocktail
(Boehringer Mannheim) was used throughout the collagen
XVIII isolation procedure to prevent proteolysis of the
sample.
Processing of human collagen XVIII and the rhNC1 domain
by MMPs
Aliquots of heparin Sepharose-enriched human colla-
gen XVIII from HepG2 cells (total protein 8 Ag) or
rhNC1 (3–5 Ag) were incubated with various MMPs
(molar enzyme/substrate ratios between 1:3 and 1:1000 as
indicated) from 30 min up to 72 h at 37-Cin50mM
R. Heljasvaara et al. / Experimental Cell Research 307 (2005) 292– 304294
Tris– HCl, pH 7.8, containing 0.2 M NaCl and 1 mM
CaCl
2
. When available, MMPs from two different sources
were used for the digestions. Latent zymogen forms of
MMPs were activated with 0.5 mM APMA, and 20 mM
EDTA was added to inhibit MMP cleavage. The cleavage
products were analyzed by Western blotting with anti-
bodies against human endostatin. The catalytic activities
of the pure human and rat MMPs used were ascertained
by assaying their degradative action against native type I
collagen, gelatin, h-casein and laminin-5 g2-chain as
described [33,40 – 42]. After the desired incubation time,
the digestions were terminated by adding SDS sample
buffer containing h-mercaptoethanol.
Western blot analysis
Proteins isolated from HepG2 cell culture media or
recombinant rhNC1 produced in E. coli were separated
on a 7–12% SDS-PAGE under reducing conditions,
electrotransferred to a nitrocellulose membrane (Protran,
Schleicher and Schuell, Keene, NH) and probed with
rabbit polyclonal antibodies against human collagen
XVIII followed by a horseradish peroxidase-conjugated
goat anti-rabbit antibody (Bio-Rad). After washing the
membrane extensively, the proteins that were reactive to
human collagen XVIII antibodies were visualized with
ECL Western blotting detection reagents (Amersham
Pharmacia Biotech, Uppsala, Sweden).
Extraction of RNA from HepG2 cells and RT-PCR analysis
of MMPs
For RT-PCR analysis of MMPs, total RNA was extracted
from HepG2 cells maintained in DMEM. The RNA
extraction and purification took place according to the
instructions accompanying the Trizol\kit (Gibco BRL,
Gaithersburg, MD). For cDNA synthesis, 4 ng of the total
RNAwas reverse transcribed using oligo(dT) as a primer. The
specific primers for amplifying MMP-3 were 5V-AGTC-
TTCCAATCCTACTGTT (forward) and 5V-GTATCCT-
TTGTCCATTGTTC (reverse), and those for amplifying
MMP-7 were 5V-GGTCACCTA C A G G AT C G TAT C ATAT
(forward) and 5V-CATCACTGCATTAGGATCAGAGGAA
(reverse). MMP-9 was amplified with primers 5V-ACCG-
CTATGGTTACACTCGG (forward) and 5V-GCAGGC-
AGAGTAGGAGCG (reverse). The primers for MMP-13
amplification were 5V-AGATAAGTGCAGCTGTTCAC
(forward) and 5V-TCATTGACAGACCATGTGTC
(reverse) together with the nested primers 5V-AGCA-
TCTGGAGTAACCGT (forward) and 5V-TCAATGTG-
GTTCCAGCCA (reverse). MMP-20 was amplified using
5V-GGTGCTCCCTGCATCTGG (forward) and 5V-CC-
TCCCAGGCCTTCTCCA (reverse) and the nested pri-
mers 5V-CTCCCTAGTTGCAGCAGCCTC (forward) and
5V-CCATCGAATGGATAGGAA (reverse). The annealing
temperatures were as follows: 60-C for the MMP-20
primers, 55-C for the nested MMP-20 primers, 54-C for
the nested MMP-13 primers and 58-C for the MMP-7
primers. Touchdown PCR was performed for MMP-13 at
annealing temperatures of 50-C, 52-C and 54-C. MMP-9
was amplified by touchdown PCR using annealing
temperatures of 56-C, 58-C and 60-C. 18S ribosomal
RNA was used as a control for RNA integrity. The PCR
products were run on 1% agarose gel containing 1 Ag/ml
ethidium bromide. The MMP-9 and MMP-20 PCR
products were purified by the QIAEX II agarose gel
extraction protocol (Qiagen GmbH, Hilden, Germany)
and determined by automated sequencing (ABI 310
DNA Sequencer, Applied Biosystems Inc., Foster
City, CA).
N-terminal microsequence analysis
MMP cleavage products separated on a 12% SDS-
PAGE under reducing conditions were blotted onto a
PVDF membrane and Coomassie blue-stained bands were
excised from the membrane. The N-terminal ends of the
cleavage products were determined by automated Edman
degradation with a 492 Prociseiprotein sequencer
(Applied Biosystems Inc., Foster City, CA).
Endothelial cell proliferation and migration assays
The proliferation assay was performed as previously
described [5] with a few modifications. A HUV-EC-C
cell suspension was plated on 24-well plates at
11,500– 12,500 cells/well and incubated (37-C, 5%
CO
2
) for 24 h. The medium was replaced with 0.5
ml of fresh medium supplemented with 5% fetal bovine
serum and different concentrations (0, 1, 5, 10 and 20
Ag/ml) of various C-terminal fragments of collagen
XVIII in triplicate wells. After 72 h, cells were
dispersed in 0.05% trypsin and counted in a Bu¨rker-
Tu¨rk chamber.
HUV-EC-C and EA.hy926 cell migration was studied
using 8.0-Am pore size and 6.5-mm diameter Transwell
inserts (Costar, Cambridge, MA, USA) equilibrated in
serum-containing medium for 2 h before use. Cells
were preincubated at 37-C in a humidified 5% CO
2
atmosphere for 30 min in the presence of 5 or 20 Ag/
ml endostatin fragments. For migration assay, 600 Alof
the serum and VEGF (2 ng/ml) containing medium was
added to the lower compartment of the migration
apparatus and 20,000 cells in a volume of 100 Alof
serum-containing medium were plated on the Transwell
filter. After culturing for 20 h, the cells were fixed in
methanol, washed and stained in toluidine blue. The
cells were removed from the upper surface of the
membrane with a cotton swab, and the cells that
migrated through the membrane were counted by
scanning (Bio-RAd GS-700 Imaging Densitometer,
Bio-Rad).
R. Heljasvaara et al. / Experimental Cell Research 307 (2005) 292– 304 295
MMP inhibitor assays
HepG2 cells were plated in 6-well plates and cultivated
in DMEM supplemented with 10% FBS until the plates
were confluent. The cells were rinsed twice with PBS and
then incubated with SFCM for 3 h. After that, the cells were
washed once with SFCM and the incubation was started
with fresh SFCM containing indicated increasing concen-
trations of MMP inhibitor marimastat (BB-2516), human
recombinant TIMP-1 or MMP Inhibitor III. Since marima-
stat was dissolved in DMSO, 1% DMSO was added to
some controls. After 24 h incubation, the inhibitors
were replenished and the HepG2-CM was collected
after 48 h. 750 Al aliquots of the HepG2-CM were
concentrated by ultracentrifugation (MWCO 10 kDa),
separated by 15% SDS-PAGE under reducing conditions
and identified by Western blotting with the polyclonal
HES.6 antibody to human endostatin.
Collagen degradation assay
The degradation of native type I collagen by proteases
present in HepG2-conditioned cell culture medium was
studied by a collagen degradation assay. HepG2-CM was
collected after 72 h culture. The substrate, 1.5 AM native
human skin type I collagen [43], was incubated for 12 h with
HepG2-CM or with the same conditioned medium pretreated
with indicated amount of recombinant human TIMP-1 for 1 h
at 37-C. Collagen I substrate incubated for the same time in
the incubation buffer [43] alone was used as a control. The
proteins were separated by 10% SDS-PAGE and stained with
Coomassie brilliant blue.
Zymography
To detect gelatinases, 50 Al of 30-fold concentrated
SFCM from HepG2 cell cultures without added protease
Fig. 1. Processing of HepG2-derived human collagen XVIII and recombinant NC1 by MMPs. Heparin affinity-purified human collagen XVIII isolated from
conditioned HepG2 cell media (A and B) or the E. coli-derived recombinant NC1 domain of human collagen XVIII (C and D) was incubated with the various
MMPs (enzyme/substrate ratios 1:10 or 1:15, as indicated in parenthesis) for 24 h at 37-C. The cleavage products were separated by 12% SDS-PAGE under
reducing conditions and identified by Western blotting with the polyclonal HES.6 (A and B) or ES2a (C and D) antibody to human endostatin. (A) Control,
HepG2-derived human collagen XVIII incubated for 24 h at 37-C without MMPs (lane 1) and incubated with MMP-1 (1:15, lane 2), MMP-2 (1:15, lane 3),
MMP-3 (1:15, lane 4), MMP-7 (1:15, lane 5) and MMP-8 (1:15, lane 6). (B) Control as above (lane 1) and collagen XVIII incubated with MMP-9 (1:15, lane
2), MMP-12 (1:15, lane 3), MMP-13 (1:15, lane 4) and MMP-20 (1:15, lane 5). (C) Control, recombinant human NC1 (rhNC1) incubated for 24 h at 37-C
without MMPs (lane 1), rhNC1 incubated with MMP-1 (1:15, lane 2), MMP-2 (1:15, lane 3), MMP-3 (1:10, lane 4), MMP-7 (1:15, lane 5) and MMP-8 (1:15,
lane 6). (D) Control (lane 1), rhNC1 incubated with MMP-9 (1:10, lane 2), MMP-12 (1:15, lane 3), MMP-13 (1:15, lane 4) and MMP-20 (1:15, lane 5).
Molecular masses (kDa) of the cleavage products are shown on the right.
R. Heljasvaara et al. / Experimental Cell Research 307 (2005) 292– 304296
inhibitors was analyzed by zymography as described
earlier [44].
Results
Processing of HepG2-derived human collagen XVIII and
recombinant NC1 by MMPs
Human collagen XVIII was enriched by heparin affinity
chromatography from conditioned serum-free HepG2 cell
media. Aliquots of this material (total protein 8 Ag) were
incubated with various MMPs (molar enzyme/substrate ratio
between 1:3 and 1:15 as indicated) for 24 h at 37-C, and the
cleavage products were separated out on 7 – 12% SDS-
PAGE under reducing conditions and identified by Western
blotting with polyclonal antibodies against the N-terminal
(anti-all huXVIII) and C-terminal (HES.6) portions of the
collagen XVIII molecule. The full-length glycosylated,
smear-like, over 200 kDa collagen XVIII molecule was
detected on a 7% polyacrylamide gel with both the anti-all
huXVIII and the HES.6 antibodies (not shown), and it
migrated as a sharp band near the border of the stacking and
separating gels in 12% SDS-PAGE (Fig. 1). Although a
protease inhibitor cocktail was used throughout the purifi-
cation procedure, certain endogenous proteases from the
HepG2 cells digested the collagen XVIII, producing faint
bands between 24 and 38 kDa that were reactive with the
HES.6 antibody (Figs. 1A and B, lane 1). Degradation of the
full-length collagen XVIII and a significant increase in the
occurrence of small endostatin-containing polypeptides in
the interval 24–30 kDa were nevertheless detected after the
incubation of collagen XVIII with MMP-3, -7, -9, -13 and
-20 (Fig. 1A, lanes 4–5; Fig. 1B, lanes 2, 4 and 5),
whereas MMP-1, -2, -8 and -12 resulted in only very little,
if any, degradation of collagen XVIII as compared with a
control incubation for 24 h at 37-C without MMPs (Fig.
1A, lanes 2, 3 and 6; Fig. 1B, lane 3). Even prolonged
incubations of up to 72 h or the use of increased amounts
of these MMPs did not result in the appearance of
endostatin fragments. The addition of EDTA to the
incubation mixtures (final concentration 20 mM) com-
pletely abolished the proteolytic cleavage, confirming that
it is MMP-dependent (not shown). Attempts to purify and
sequence the endostatin-related digestion products resulting
from the HepG2-derived full-length collagen XVIII were
unsuccessful due to the small amount of protein present.
To study in detail the degradative action of MMPs on the
NC1 domain of collagen XVIII, we cloned and expressed
the recombinant human NC1 (rhNC1) in E. coli and purified
it with the aid of a His-tag inserted into the N-terminus of
the recombinant protein. 3 – 5 Ag of the purified rhNC1 was
incubated with MMPs (molar enzyme/substrate ratios 1:10
or 1:15) for 24 h at 37-C and the resulting cleavage products
were subjected to 12% SDS-PAGE and analyzed by Western
blotting with a polyclonal peptide antibody ES2a (Figs. 1C
and D). The results of these digestions confirmed the results
obtained with the endogenous HepG2-derived collagen
XVIII. MMP-3, -7, -9, -13 and -20 clearly degraded the
rhNC1, yielding endostatin-containing fragments of 20-30
kDa, whereas MMP-1, -2, -8 and -12 did not show any
activity against rhNC1, even when used in molar excess
amounts (not shown). With MMP-7, the 38-kDa NC1 was
transformed into three cleavage products that were reactive
with the ES2a antibody, the major bands having molecular
masses of 30 kDa and 27 kDa (Fig. 1C, lane 5). MMP-20
cleaved NC1 to a prominent 24-kDa polypeptide detectable
with ES2a (Fig. 1D, lane 5), and MMP-13 generated two
fragments of approximately 24 and 28 kDa (Fig. 1D, lane
4). MMP-3 and MMP-9 resulted only in a partial digestion
of NC1 during the 24 h incubation when MMP-3 produced
several ES2a-positive polypeptides between 20 kDa and 30
Fig. 2. Production of MMPs by HepG2. (A) RT-PCR analysis of MMP-3 (lane 2), MMP-7 (lane 3), MMP-9 (lane 4), MMP-13 (lane 5) and MMP-20 (lane 6)
and control 18S ribosomal RNA (lane 7) in HepG2 cells. Sizes of the PCR products (bp) are shown on the right. (B) Zymography analysis of gelatinases in
HepG2 cell culture. HepG2-CM was collected, concentrated and analyzed by gelatine zymography. (C) Degradation of type I collagen by conditioned HepG2
cell media. HepG2-CM was collected after 72 h in culture and part of it was pretreated with recombinant human TIMP-1 for 1 h at 37-C. 1.5 AM of native
human skin type I collagen was incubated for 12 h with buffer alone (lane 1), with HepG2-CM (lane 2) or with HepG2-CM pretreated with 35 AM (lane 3) of
TIMP-1. Coomassie brilliant blue staining was used to visualize the intact collagen I monomers (denoted by a) and the characteristic 3/4-degradation products
(denoted by aA).
R. Heljasvaara et al. / Experimental Cell Research 307 (2005) 292– 304 297
kDa (Fig. 1C, lane 4) and MMP-9 generated two endostatin-
related fragments with molecular masses of 28 kDa and 24
kDa (Fig. 1D, lane 2). The calculated sizes of the major
cleavage products corresponded to the endostatin fragments
released from HepG2-derived collagen XVIII by MMPs.
The cleavage of the 20-kDa endostatin fragment by MMP-3
was detected only when recombinant NC1 but not HepG2-
derived native collagen XVIII was used as a substrate (Fig.
1A, lane 4) which may reflect defects in the folding of the
rhNC1. The addition of a metal chelator (20 mM EDTA)
inhibited the degradative action of the MMPs on NC1 (not
shown). Furthermore, a detailed time course study and
series of enzyme dilutions for each MMP were performed.
Particularly, MMP-20, but also MMP-7 and -9, proved to be
potent proteases in digesting endostatin-containing peptides
from the HepG2-derived human collagen XVIII; already
after 30 min incubation at 37-C cleavage products between
24–38 kDa reactive to endostatin antibody started to
accumulate (not shown). In our experiments, MMP-3 and
MMP-13 showed considerably less activity against collagen
XVIII (not shown).
Production of endostatin fragments by the HepG2 cell
system
To study more closely the endogenous cleavage of
human collagen XVIII that occurred in the HepG2 cell
system, we ascertained the synthesis of several MMPs by
this cell type. mRNA expression of five MMPs (MMP-3, -7,
-9, -13 and -20) that degraded collagen XVIII and produced
endostatin fragments in vitro was examined using RT-PCR.
All of them were demonstrated to be expressed by HepG2
cells (Fig. 2A). The identities of their PCR products were
confirmed by DNA sequencing. To further examine the
production of active MMPs by HepG2 cells, zymography
analysis was performed to demonstrate gelatinases in the
HepG2 cell medium. This experiment revealed a strong 92-
kDa and a weaker 77-kDa gelatinolytic band representing
the latent and active forms of MMP-9, respectively (Fig.
2B). Furthermore, a weak 63-kDa band was observed. The
presence of catalytically competent collagenases in the
HepG2-CM was verified by the collagen degradation assay,
which showed the in vitro processing of native type I
collagen to characteristic 3/4-cleavage products in the
control media (Fig. 3C, lane 2) whereas the HepG2-CM
pretreated with TIMP-1 (35 AM) failed to generate these
fragments (Fig. 3C, lane 3). Furthermore, antibodies against
MMP-3, -7, -13 and -20 recognized several immunoreactive
species that corresponded to the reported molecular weights
of latent and active forms of these MMPs (not shown).
These results indicate that the HepG2 cell system contains
active forms of endogenous MMPs that can generate
endostatin fragments in culture.
We also used the broad-spectrum MMP inhibitor
marimastat (BB-2516) to block the function of major
metalloproteinase subtypes in HepG2 cells. The addition
of marimastat (1 – 50 AM) in the cell culture medium
impaired the formation of endostatin fragments between
24–30 kDa indicating that their production is likely due to
MMP action on type XVIII collagen (Fig. 3A). The
intensities of the endostatin-containing polypeptides formed
in HepG2 cell culture in the absence and presence of
marimastat were quantified. 1 – 50 AM marimastat was
found to decrease the formation of endostatin-containing
fragments 25–60% (Fig. 3B). Furthermore, TIMP-1, an
endogenous inhibitor of MMPs, and the MMP Inhibitor III,
which blocks the activity of MMP-1, -2, -3, -7 and -13,
reduced the formation of these fragments by HepG2 cells
(not shown).
Cleavage sites of the human type XVIII collagen NC1
domain by MMPs
In order to determine the specific cleavage sites
generated in NC1 by MMP-3, -7, -9, -13 and -20,
approximately 10 – 15 Ag of rhNC1 was incubated with
Fig. 3. Inhibition of collagen XVIII processing by MMP inhibitor marimastat. (A) Inhibition of endogenous MMP cleavage of human collagen XVIII by
HepG2 cells. SFCM from HepG2 cell cultures was collected after 48 h incubation in the presence of increasing amounts of MMP inhibitor marimastat. Aliquots
of the SFCM were concentrated by ultracentrifugation, separated by 15% SDS-PAGE under reducing conditions and electroblotted onto nitrocellulose. Type
XVIII collagen fragments containing endostatin were identified by Western blotting with the polyclonal HES.6 antibody to human endostatin. Molecular
masses (kDa) of the endogenous cleavage products are shown on the left. (B) The intensities of the endostatin fragments were measured using Quantity One
software (Bio-Rad). The inhibition in endostatin fragment formation by marimastat is presented as % of the control sample. The columns in panel (B) represent
means of two separate experiments TSD.
R. Heljasvaara et al. / Experimental Cell Research 307 (2005) 292– 304298
these MMPs for 24 h at 37-C. MMP cleavage products
excised from a Coomassie blue-stained PVDF membrane
were sequenced by automated Edman degradation. When
available, MMPs from two different sources were used for
these experiments, and each cleavage product was
sequenced at least twice. MMP-3 (molar enzyme/substrate
ratios 1:10, 1:20 and 1:30) generated two prominent
cleavage products of 20 and 24 kDa from rhNC1 with
cleavage sites at His
130
–Ser
131
and Ser
115
–Tyr
116
, respec-
tively. Despite several attempts, the N-termini of the other
ES2a-positive fragments produced by MMP-3 (Fig. 1C, lane
4) could not be determined. Human and rat MMP-7 (molar
enzyme/substrate ratios of 1:20 and 1:30) repeatedly
generated two fragments, a major 30-kDa fragment and a
minor 27-kDa fragment corresponding to cleavage at Leu
70
and Ile
94
. In addition, an MMP-7-derived minor fragment of
24 kDa was identical to the N-termini of the corresponding
MMP-3 and MMP-20 cleavage products. MMP-9 (molar
enzyme substrate ratio 1:10) resulted in cleavage of Pro
106
–
Tyr
107
and Pro
88
–Trp
89
bonds in rhNC1 with fragment sizes
of about 25 and 28 kDa. Incubation of rhNC1 with MMP-13
(molar enzyme/substrate ratios of 1:15, 1:25, 1:30 and 1:50)
resulted in 24- and 28-kDa fragments starting at residues
Tyr
116
and Trp
89
, respectively. Furthermore, a slightly
smaller 24-kDa fragment resulting from MMP-13 cleavage
at residues His
113
–Ser
114
was identified. MMP-20 (molar
enzyme/substrate ratios of 1:40 and 1:100) generated a
single Coomassie-visible product of 24 kDa, the cleavage
site residing between Ser
115
and Tyr
116
in rhNC1. The
results of the N-terminal amino acid sequencing are
summarized in Table 1.
Endothelial cell proliferation and migration
The anti-proliferative activity of five C-terminal frag-
ments from human collagen XVIII, namely the 20-kDa
endostatin, the 24-, 25- and 28-kDa endostatin-containing
polypeptides starting at residues Tyr
116
,Tyr
107
and Trp
89
,
respectively, and the full-length NC1 domain of 38 kDa,
was measured using bFGF-stimulated HUV-EC-C cells.
Depending on experiment, 11,500–12,500 cells were
plated on each well, and after 24 h incubation, the
culture medium was replaced with fresh medium con-
taining different concentrations (0 – 20 Ag/ml) of various
endostatin fragments of type XVIII collagen. After 72 h
incubation, cells were counted and the proliferation of the
cells was compared to control samples (% of control). All
five polypeptides inhibited the proliferation of HUV-EC-
C cells in vitro in a dose-dependent fashion (Fig. 4A).
The inhibition with increasing endostatin fragment con-
centrations (0 –20 Ag/ml) ranged between 25 – 80% of the
controls. The 25- and 24-kDa fragments followed nicely
the inhibition profile of recombinant human endostatin
whereas the two longer fragments, namely the 38-kDa
NC1 domain and the 28-kDa fragment, had more
pronounced effect on the proliferation of HUV-ECs at
high concentrations (74 – 80%). Some fragments (endo-
statin, 28- and 25-kDa fragments) were tested also in 24
h proliferation assay, and they showed essentially similar
inhibition curves as in the 72 h assay (not shown).
During the proliferation assays, the longer endostatin-
containing fragments started to degrade to the stable
endostatin form. Densitometric analysis of two of these
fragments, namely the 25- and 28-kDa polypeptides,
indicated that approximately 40 to 50% of them were
further processed during incubation, whereas recombinant
endostatin remained intact (not shown). In the individual
24 h and 72 h proliferation assays, where endostatin was
tested in parallel with the 25- or 28-kDa fragment, the
two longer fragments appeared to inhibit HUV-EC-C cell
proliferation slightly more efficiently (9 – 18%) at various
concentrations (not shown). Thus, we reasoned that the
anti-proliferative effects caused by the longer endostatin-
containing fragments could not only be a consequence of
their processing to endostatin but also indicate activity of
longer fragments. Overall, the various endostatin-contain-
ing fragments of collagen XVIII possessed an ability to
inhibit endothelial cell proliferation. Significant differ-
ences between the fragments were not detected, and
seemingly the inhibition efficiency was not directly
related to the fragment length.
We also tested the E. coli-derived human endostatin
and the 24- and 28-kDa fragments in an in vitro VEGF-
induced endothelial cell migration assay using Transwell
filters. 20 000 cells, preincubated for 30 min in the
presence of 5 or 20 Ag/ml endostatin fragments, were
plated on the filter, and VEGF (2 ng/ml) was used as a
chemoattractant. After 20 h incubation, the cells that
migrated through the Transwell membrane were counted
by scanning. We saw that the 20-kDa endostatin was
slightly more efficient in preventing HUV-EC-C cell
migration than the 24-kDa polypeptide (Fig. 4B),
whereas endostatin and the 28-kDa fragment inhibited
EA.hy926 cell migration with the same potency (Fig.
4C). The maximal inhibition of these endostatin frag-
ments on HUV-EC-C and EA.hy926 cell migration was
Table 1
Endostatin fragments generated from human collagen XVIII-derived
recombinant NC1 domain by distinct MMPs
Endostatin fragment
MMP Size (kDa) N-terminal sequence Cleavage site in NC1
MMP-3 24 YVHLRPARXT Ser
115
–Tyr
116
20 SHRXFQPVL His
130
– Ser
131
MMP-7 30 LHDSNPYPXR Gln
69
– Leu
70
27 ILASPPXL Asp
93
– Ile
94
24 XVHXRPARXT Ser
115
–Tyr
116
MMP-9 28 WRADDILASP Pro
88
–Trp
89
25 YPGAXXX Pro
106
–Tyr
107
MMP-13 28 XRADDILASP Pro
88
–Trp
89
24 YVHLRPARPT Ser
115
–Tyr
116
24 SSYVHLRPART His
113
– Ser
114
MMP-20 24 YVHLRPARPT Ser
115
–Tyr
116
R. Heljasvaara et al. / Experimental Cell Research 307 (2005) 292– 304 299
approximately 30% and 26 – 40% at the concentration of
20 Ag/ml.
Discussion
The present study demonstrates in vitro cleavage of the
C-terminal noncollagenous NC1 domain of human collagen
XVIII by several MMPs and the generation of biologically
active endostatin fragments of distinct sizes and N-terminal
sequences. Of the nine MMPs examined, NC1 was clearly
degraded by cancer-related MMP-3, -7, -9, -13 and -20 [45],
and N-terminal microsequencing of the digestion products
indicated that the cleavages occurred within the protease-
sensitive hinge region of NC1, as judged by reference to the
domain model of Sasaki et al. [20]. While only the Pro
106
–
Tyr
107
cleavage site by MMP-9 strictly follows the general
consensus sequence Pro-X-X-,X
Hy
for MMPs, most of the
cleavage sites meet the requirement of a hydrophobic
residue (Hy) at the P1Vposition [46,47]. Two of the
cleavage sites determined here, namely the Pro
106
–Tyr
107
by MMP-9 and Gln
69
–Leu
70
by MMP-7, have also been
identified earlier by others [25,26]. We found that several
MMPs generate a 24-kDa endostatin containing polypeptide
by cleaving collagen XVIII at Tyr
116
. This fragment occurs
in human plasma [19] indicating its endogenous production,
possibly by MMPs, albeit the cleavage site determined here
does not follow the consensus sequence. We also show here
its biological activity inhibiting endothelial cell migration
and proliferation (Fig. 4). Although MMP-2 and MMP-12
have been previously shown to process human recombinant
NC1 domain in vitro [25], we could not detect the cleavage
Fig. 4. Inhibition of human endothelial cell proliferation and migration by recombinant endostatin-containing polypeptides. (A) Endothelial cell proliferation.
bFGF-induced HUV-EC-C cells were incubated with different concentrations (0 – 20 Ag/ml) of recombinant human endostatin (ES, 20 kDa), the 24-kDa, the
25-kDa and the 28-kDa endostatin-containing peptides starting at residues Tyr
116
,Tyr
107
and Trp
89
, respectively, or the NC1 domain of 38 kDa. Each longer
endostatin-containing fragment was tested in parallel with recombinant endostatin. The columns represent means of triplicate cell cultures TSD except the
endostatin columns where nis 12. During the 72 h proliferation assay, the control cell number (0 Ag/ml of recombinant protein) increased from 11,500 – 12,500
to 74,000 –76,000 cells/well (6-fold) or to 110,000– 134,000 cells/well (9 – 10 -fold) depending on the HUV-EC-C cell batch used. (B and C) Endothelial cell
migration. 20,000 HUV-EC-C (B) or EA.hy926 (C) cells preincubated for 30 min with 0 (black bars), 5 (white bars) or 20 Ag/ml (gray bars) of recombinant
human endostatin (ES, 20 kDa), the 24-kDa (B) or the 28-kDa fragment (C) were plated on the Transwell filter and the VEGF-induced (2 ng/ml) cell migration
was analyzed after 20 h by scanning. The inhibition of endothelial cell proliferation and migration is presented as % of controls.
R. Heljasvaara et al. / Experimental Cell Research 307 (2005) 292– 304300
of native collagen XVIII or rhNC1 by MMP-1, -2, -8 and
-12 under the conditions prevailing here. Consistent with
and further extending the previous observations [23,24],
none of the human or rat MMPs studied was able to produce
the 20-kDa peptide with the N-terminal sequence
HSHRDFQP that corresponds exactly to the originally
identified mouse endostatin [5].
Many angiogenesis inhibitors are stored as cryptic
fragments within larger precursor molecules that are not
themselves antiangiogenic, and the regulation of the
proteolytic processing plays an important role in the
neovascularization [1,48]. It is not entirely clear, however,
which cells and mechanisms are responsible for the
release of these cryptic inhibitors. Collagen XVIII
mRNAs are expressed by epithelial and endothelial cells
as well as liver hepatocytes [38,49]. Collagen XVIII
protein occurs prominently in the vascular and epithelial
BM zones [38,49,50]. The blood circulation is also
known to contain several precursors of antiangiogenic
peptides [51,52], and our recent study suggests that
soluble full-length type XVIII collagen is also present in
human plasma [53]. Furthermore, a recent work demon-
strates a significant correlation between elevated serum
endostatin level and increased collagen XVIII expression
by tumor tissue itself in some human patients with non-
small cell lung cancer, suggesting that serum endostatin
may be partially derived from tumor cells [54].
We found that cultured hepatoma-derived HepG2 cells,
which retain the biosynthetic capabilities of normal liver
parenchymal cells [55], secreted full-length collagen XVIII
into the culture media as well as several endostatin-
containing polypeptides. HepG2 cells were also shown to
express mRNAs for MMP-3, -7, -9, -13 and -20 which are
all able to cleave endostatins from collagen XVIII in vitro,
suggesting that these MMPs may be responsible for the
endogenous cleavages occurring in our cell culture. mRNA
expression of MMP-3, -7 and -13 by cultured HepG2 cells
has already been proven previously [56]. Although MMP-
20 expression has been believed to be restricted solely to
dental tissues [57,58], we found its synthesis in HepG2 cells
using RT-PCR (Fig. 2A), DNA sequencing and Western
blotting (not shown). Using RT-PCR and zymography, we
could also evidence the expression of MMP-9 by HepG2
cells at both the mRNA and protein levels. Furthermore,
zymography analysis indicated that a portion of MMP-9 was
activated (Fig. 2B). Using a type I collagen degradation
assay, we demonstrated that HepG2-CM contains type I
collagenase activity as well, probably representing MMP-13
since MMP-1 and MMP-8 are not expressed by HepG2 cells
[56]. Previous work indicates that the serine, cysteine and
aspartic proteases are not likely to be involved in the
generation of the larger endostatin-containing polypeptides,
since their class-specific inhibitors did not prevent the
generation of endostatin fragments between 25 – 32 kDa,
while broad-spectrum MMP inhibitors and metal chelators
such as BB-3103, EDTA and 1,10-phenanthroline prevented
the formation of these fragments [23,24]. We demonstrated
that the production of endostatin fragments between 24 – 30
kDa was reduced by three different MMP inhibitors,
marimastat (Fig. 3), recombinant TIMP-1 and MMP
Inhibitor III, supporting the direct role for the MMPs in
generating endostatin fragments from collagen XVIII in the
HepG2 cell system and possibly also in tissues [19 – 21].
Nevertheless, substantial differences may exist between the
in vivo tissue and in vitro cell culture expression patterns of
MMPs [45]. In addition to MMPs, some other class of
metalloproteases might also be involved to some extent in
the processing of collagen XVIII, because marimastat does
not completely block the formation of endostatin-containing
fragments.
As mentioned before, MMPs seem to have dual or even
opposite effects on tumor angiogenesis, on one hand by
facilitating extracellular matrix degradation and neovascu-
larization [59] and on the other hand by blocking angio-
genesis by releasing cryptic inhibitors of endothelial cell
growth, such as endostatin studied in this work, angiostatin
derived from plasminogen [30 – 32] and tumstatin derived
from type IV collagen a3 chain [60]. Interestingly, endo-
statin seems to be able to regulate the activity of certain
MMPs showing how complex the regulation between
endostatin and MMPs is [61 – 63]. It is possible that also
the endostatin-containing fragments inhibit the activity of
certain MMPs in the same manner as the 20-kDa endostatin.
This would lead to impaired generation of endostatin
fragments by those MMPs. However, previously it has
been shown that, although endostatin inhibits MMP activity,
it does not completely block their activation [61 – 63].
Instead, MMPs can be half active even when endostatin or
endostatin fragments are present, especially when no APMA
or other MMP activator is present. It should also be noted
that the activity of MMPs on non-matrix substrates, such as
chemokines, growth factors, growth factor receptors, adhe-
sion molecules and apoptosis mediators, is essential for the
rapid and critical cellular responses required for tumor
growth and progression [64].
The biological activity and physiological significance of
the endostatin-related polypeptides in serum and tissues are
not fully understood. Originally, only the 20-kDa endostatin
molecule was reported to possess antiangiogenic activity as
measured using an in vitro proliferation assay of bFGF-
stimulated bovine capillary endothelial cells [5]. In other
studies, the recombinant NC1 domain of human collagen
XVIII and the endostatin and NC1 fragments with flag-tag
modified N- or C-termini also prevented VEGF-induced
migration of human umbilical vein endothelial cells in vitro
[65]. Furthermore, short peptides derived from human
endostatin have been shown to possess potent antiangio-
genic properties in vitro and in vivo [66 – 68]. Sudhakar et
al. have recently reported that endostatin is a potent inhibitor
of HUV-EC cell migration with no effect on proliferation
and that its inhibitory effect on cell migration is mediated by
binding to a5h1 integrin which leads to blocking of the
R. Heljasvaara et al. / Experimental Cell Research 307 (2005) 292– 304 301
ERK1/p38 MAPK pathway [17]. In another work, endo-
statin did not affect the phosphorylation of a series of signal
transduction components including p38 MAPK, suggesting
that it does not interfere with key intracellular signaling
cascades that regulate endothelial cell migration and
proliferation [69]. The same work also showed that endo-
statin does not affect the proliferation of bFGF-induced
human dermal microvascular endothelial cells. Furthermore,
it seems to depend on the inductive cytokine whether
endostatin has an effect on angiogenesis [70]. Finally,
cathepsin L cleaving human recombinant endostatin results
in an endostatin-like fragment that is 11 amino acids larger
than the murine endostatin [24], suggesting that functional
human and mouse endostatins may differ.
We showed here that the MMP-generated endostatin
fragments longer than 20-kDa can inhibit bFGF-stimulated
proliferation and VEGF-stimulated migration of endothelial
cells in vitro. The divergent results from our and others’
[5,17,19,65–70] experiments suggest that endostatin mole-
cules varying in length and origin may regulate endothelial
cell migration and proliferation by different mechanisms
and that the effects might be cell type specific [69]. The
effects also appear to vary depending on the growth factors
used for the in vitro cell migration and proliferation
experiments. As endostatin is suggested to act via integrin
a5h1[15–17], it might be possible that the expression of
integrins and other cell surface receptors fluctuates depend-
ing on the cell type and growth phase of endothelial cells.
In summary, using a range of MMPs in in vitro cleavage
assays, we demonstrated the production of the biologically
active endostatin-containing fragments corresponding to the
sizes of the endogenous peptides detected in tissues and in
the circulation. These MMP-generated polypeptides may
well represent intermediate cleavage products before the
formation of the final functional endostatin molecule [23].
Our findings support and further extend the concept that
they can exert biological roles of their own in tissues, acting
as local inhibitors of neovascularization, and in plasma,
participating in homeostatic control of angiogenesis [71].
Acknowledgments
We thank Jaana Peters, Aila White, Sirkka Vilmi and
Sirpa Kangas for their excellent technical assistance and
Hongmin Tu, M.Sc. for sharing her expertise in N-terminal
sequencing. Dr. Cora-Jean S. Edgell is acknowledged for
providing the EA.hy926 cells. This work was supported by
grants from the Finnish Centre of Excellence Programme
(2000–2005) of the Academy of Finland (44843), the
European Commission (QLK3-2000-00084), the Helsinki
University Research Funds, the Helsinki University Central
Hospital EVO Research Funds (TI020Y002 and TYH
4113), the Finnish Cancer Foundation, the Finnish Dental
Society Apollonia, the Maud Kuistila Foundation and the
Sigrid Juse´lius Foundation.
Appendix A. Supplementary data
Supplementary data associated with this article can be
found, in the online version, at doi:10.1016/j.yexcr.2005.
03.021.
References
[1] D. Hanahan, J. Folkman, Patterns and emerging mechanisms of the
angiogenic switch during tumorigenesis, Cell 86 (1996) 353 – 364.
[2] R. Kalluri, Basement membranes: structure, assembly and role in
tumour angiogenesis, Nat. Rev., Cancer 3 (2003) 422 – 433.
[3] M. Rehn, E. Hintikka, T. Pihlajaniemi, Primary structure of the alpha 1
chain of mouse type XVIII collagen, partial structure of the
corresponding gene, and comparison of the alpha 1(XVIII) chain with
its homologue, the alpha 1(XV) collagen chain, J. Biol. Chem. 269
(1994) 13929 – 13935.
[4] S.P. Oh, Y. Kamagata, Y. Muragaki, S. Timmons, A. Ooshima, B.R.
Olsen, Isolation and sequencing of cDNAs for proteins with multiple
domains of Gly – Xaa – Yaa repeats identify a distinct family of
collagenous proteins, Proc. Natl. Acad. Sci. U. S. A. 91 (1994)
4229 – 4233.
[5] M.S. O’Reilly, T. Boehm, Y. Shing, N. Fukai, G. Vasios, W.S. Lane, E.
Flynn, J.R. Birkhead, B.R. Olsen, J. Folkman, Endostatin: an
endogenous inhibitor of angiogenesis and tumor growth, Cell 88
(1997) 277 – 285.
[6] J. Folkman, R. Kalluri, Tumor angiogenesis, in: R.C. Bast Jr.,
D.W. Kufe, R.E. Pollock, R.R. Weichselbaum, J.F. Holland, E.I.
Frei (Eds.), Cancer Medicine-5 Review, B.C. Decker, Ontario,
2003, pp. 161 – 194.
[7] N. Ortega, Z. Werb, New functional roles for non-collagenous
domains of basement membrane collagens, J. Cell. Sci. 115 (2002)
4201 – 4214.
[8] A. Marneros, B.R. Olsen, The role of collagen-derived proteolytic
fragments in angiogenesis, Matrix Biol. 20 (1992) 337– 345.
[9] N. Fukai, L. Eklund, A.G. Marneros, S.P. Oh, D.R. Keene, L.
Tamarkin, M. Niemela¨, M. Ilves, E. Li, T. Pihlajaniemi, B.R. Olsen,
Lack of collagen XVIII/endostatin results in eye abnormalities, EMBO
J. 21 (2002) 1535 – 1544.
[10] B.R. Olsen, From the editors desk, Matrix Biol. 21 (2002) 309 – 310.
[11] A. Abdollahi, P. Hahnfeldt, C. Maercker, H.-J. Gro¨ne, J. Debus, W.
Ansorge, J. Folkman, L. Hlatky, P.E. Huber, Endostatin’s antiangio-
genic signaling network, Mol. Cell 13 (2004) 649– 663.
[12] M. Shichiri, Y. Hirata, Antiangiogenesis signals by endostatin, FASEB
J. 15 (2001) 1044 – 1053.
[13] J. Dixelius, M. Cross, T. Matsumoto, T. Sasaki, R. Timpl, L. Claesson-
Welsh, Endostatin regulates endothelial cell adhesion and cytoskeletal
organization, Cancer Res. 62 (2002) 1944– 1947.
[14] Y.M. Kim, S. Hwang, Y.M. Kim, B.J. Pyun, T.Y. Kim, S.T. Lee, Y.S.
Gho, Y.G. Kwon, Endostatin blocks vascular endothelial growth
factor-mediated signaling via direct interaction with KDR/Flk-1,
J. Biol. Chem. 277 (2002) 27872 – 27879.
[15] M. Rehn, T. Veikkola, E. Kukk-Valdre, H. Nakamura, M. Ilmonen, C.
Lombardo, T. Pihlajaniemi, K. Alitalo, K. Vuori, Interaction of
endostatin with integrins implicated in angiogenesis, Proc. Natl. Acad.
Sci. U. S. A. 98 (2001) 1024 – 1029.
[16] S.A. Wickstro¨ m, K. Alitalo, J. Keski-Oja, Endostatin associates with
integrin alpha5beta1 and caveolin-1, and activates Src via a tyrosyl
phosphatase-dependent pathway in human endothelial cells, Cancer
Res. 62 (2001) 5580 – 5589.
[17] A. Sudhakar, H. Sugimoto, C. Yang, J. Lively, M. Zeisberg, R. Kalluri,
Human tumstatin and human endostatin exhibit distinct antiangiogenic
activities mediated by alpha v beta 3 and alpha 5 beta 1 integrins, Proc.
Natl. Acad. Sci. U. S. A. 100 (2003) 4766 – 4771.
R. Heljasvaara et al. / Experimental Cell Research 307 (2005) 292– 304302
[18] G. Schuh, J.V. Heymach, M. Nomi, M. Machluf, J. Force, A. Atala,
J.P. Eder Jr., J. Folkman, S. Soker, Endostatin inhibits the vascular
endothelial growth factor-induced mobilization of endothelial progen-
itor cells, Cancer Res. 63 (2003) 8345 – 8350.
[19] L. Sta¨ndker, M. Schrader, S.M. Kanse, M. Jurgens, W.G. Forssmann,
K.T. Preissner, Isolation and characterization of the circulating form of
human endostatin, FEBS Lett. 420 (1997) 129 – 133.
[20] T. Sasaki, N. Fukai, K. Mann, W. Go¨hring, B.R. Olsen, R. Timpl,
Structure, function and tissue forms of the C-terminal globular domain
of collagen XVIII containing the angiogenesis inhibitor endostatin,
EMBO J. 17 (1998) 4249 – 4256.
[21] H. John, K.T. Preissner, W.G. Forssmann, L. Sta¨ndker, Novel
glycosylated forms of human plasma endostatin circulating endo-
statin-related fragments of collagen XV, Biochemistry 38 (1999)
10217 – 10224.
[22] N. Miosge, T. Sasaki, R. Timpl, Angiogenesis inhibitor endostatin is a
distinct component of elastic fibers in vessel walls, FASEB J. 13
(1999) 1743 – 1750.
[23] W. Wen, M.A. Moses, D. Wiederschain, J.L. Arbiser, J. Folkman, The
generation of endostatin is mediated by elastase, Cancer Res. 59
(1999) 6052 – 6056.
[24] U. Felbor, L. Dreier, R.A. Bryant, H.L. Ploegh, B.R. Olsen, W.
Mothes, Secreted cathepsin L generates endostatin from collagen
XVIII, EMBO J. 19 (2000) 1187– 1194.
[25] M. Ferreras, U. Felbor, T. Lenhard, B.R. Olsen, J.-M. Delaisse´,
Generation and degradation of human endostatin proteins by various
proteinases, FEBS Lett. 486 (2000) 247 – 251.
[26] H.-C. Lin, J.-C. Chang, S. Jain, E.E. Gabison, T. Kure, T. Kato, N.
Fukai, D.T. Azar, Matrilysin cleavage of corneal collagen type XVIII
NC1 domain and generation of a 28-kDa fragment, Invest. Ophthal-
mol. Visual Sci. 42 (2001) 2517– 2524.
[27] M.A. Moses, The regulation of neovascularization of matrix metal-
loproteinases and their inhibitors, Stem Cells 15 (1997) 180 – 189.
[28] W.G. Stetler-Stevenson, Matrix metalloproteinases in angiogenesis: a
moving target for therapeutic intervention, J. Clin. Invest. 103 (1999)
1237 – 1241.
[29] Z. Werb, T.H. Vu, J.L. Rinkenberger, L.M. Coussens, Matrix-
degrading proteases and angiogenesis during development and tumor
formation, APMIS 107 (1999) 11 – 18.
[30] Z. Dong, R. Kumar, X. Yang, I.J. Fidler, Macrophage-derived
metalloelastase is responsible for the generation of angiostatin in
Lewis lung carcinoma, Cell 88 (1997) 801 – 810.
[31] L.A. Cornelius, L.C. Nehring, E. Harding, M. Bolanowski, H.G.
Welgus, D.K. Kobayashi, R.A. Pierce, S.D. Shapiro, Matrix metal-
loproteinases generate angiostatin: effects on neovascularization,
J. Immunol. 161 (1998) 6845 – 6852.
[32] M.S. O’Reilly, D. Wiederschain, W.G. Stetler-Stevenson, J. Folkman,
M.A. Moses, Regulation of angiostatin production by matrix metal-
loproteinase-2 in a model of concomitant resistance, J. Biol. Chem.
274 (1999) 29568– 29571.
[33] T. Sorsa, T. Salo, E. Koivunen, J. Tyynela¨ , Y.T. Konttinen, U.
Bergmann, A. Tuuttila, E. Niemi, O. Teronen, P. Heikkila¨, H.
Tschesche, J. Leinonen, S. Osman, U.-H. Stenman, Activation of type
IV procollagenases by human tumor-associated trypsin-2, J. Biol.
Chem. 272 (1997) 21067 – 21074.
[34] J.M. Chen, R.T. Aimes, G.R. Ward, G.L. Youngleib, J.P. Quigley,
Isolation and characterization of a 70-kDa metalloprotease (gelatinase)
that is elevated in Rous sarcoma virus-transformed chicken embryo
fibroblasts, J. Biol. Chem. 266 (1991) 5113 – 5121.
[35] H. Turto, S. Lindy, V.-J. Uitto, O. Wegelius, J. Uitto, Human leukocyte
collagenase: characterization of enzyme kinetics by a new method,
Anal. Biochem. 83 (1977) 557 – 569.
[36] Y.T. Konttinen, O. Lindy, K. Suomalainen, C. Ritchlin, H. Saari, M.
Vauhkonen, A. Lauhio, S. Santavirta, T. Sorsa, Substrate specificity
and activation mechanisms of collagenase from human rheumatoid
synovium, Matrix 11 (1991) 395 – 403.
[37] O. Lindy, Y.T. Konttinen, T. Sorsa, Y. Ding, S. Santavirta, A. Ceponis,
C. Lo´pez-Otı´n, Matrix metalloproteinase 13 (collagenase 3) in human
rheumatoid synovium, Arthritis Rheum. 40 (1997) 1391 – 1399.
[38] J. Saarela, M. Rehn, A. Oikarinen, H. Autio-Harmainen, T.
Pihlajaniemi, The short and long forms of type XVIII collagen show
clear tissue specificities in their expression and location in basement
membrane zones in humans, Am. J. Pathol. 153 (1998) 611 – 626.
[39] J. Saarela, R. Ylika¨rppa¨, M. Rehn, S. Purmonen, T. Pihlajaniemi,
Complete primary structure of two variant forms of human type XVIII
collagen and tissue-specific differences in the expression of the
corresponding transcripts, Matrix Biol. 16 (1998) 319 – 328.
[40] R. Hanemaaijer, T. Sorsa, Y.-T. Konttinen, Y. Ding, M. Sutinen, H.
Visser, V.W. van Hinsbergh, T. Helaakoski, T. Kainulainen, H. Ro¨nka¨,
H. Tschesche, T. Salo, Matrix metalloproteinase-8 is expressed in
rheumatoid synovial fibroblasts endothelial cells. Regulation by tumor
necrosis factor-alpha and doxycycline, J. Biol. Chem. 272 (1997)
31504 – 31509.
[41] O. Teronen, P. Heikkila¨, Y.T. Konttinen, M. Laitinen, T. Salo, R.
Hanemaaijer, A. Teronen, P. Maisi, T. Sorsa, MMP inhibition and
downregulation by bisphosphonates, Ann. N. Y. Acad. Sci. 878 (1999)
453 – 465.
[42] E. Pirila¨, A. Sharabi, T. Salo, V. Quaranta, H. Tu, R. Heljasvaara, N.
Koshikawa, T. Sorsa, P. Maisi, Matrix metalloproteinases process the
laminin-5 gamma 2-chain and regulate epithelial cell migration,
Biochem. Biophys. Res. Commun. 18 (2003) 1012 – 1017.
[43] T. Sorsa, Y.-L. Ding, T. Salo, A. Lauhio, O. Teronen, T. Ingman,
H. Ohtani, N. Andoh, S. Takeha, Y.T. Konttinen, Effects of
tetracyclines on neutrophil, gingival, and salivary collagenases. A
functional and Western-blot assessment with special reference to
their cellular sources in periodontal diseases, Ann. N. Y. Acad.
Sci. 732 (1994) 112 – 131.
[44] M. Ma¨kela¨, T. Salo, V.J. Uitto, H. Larjava, Matrix metalloproteinases
(MMP-2 and MMP-9) of the oral cavity: cellular origin and relation-
ship to periodontal status, J. Dent. Res. 73 (1994) 1397 – 1406.
[45] H. Birkedal-Hansen, Proteolytic remodeling of extracellular matrix,
Curr. Opin. Cell Biol. 7 (1995) 728 – 735.
[46] H. Nagase, G.B. Fields, Human matrix metalloproteinase specificity
studies using collagen sequence-based synthetic peptides, Biopoly-
mers 40 (1996) 399 – 416.
[47] R. Visse, H. Nagase, Matrix metalloproteinases and tissue inhibitors of
metalloproteinases: structure, function, and biochemistry, Circ. Res.
92 (2003) 827– 839.
[48] J. Folkman, Seminars in medicine of the Beth Israel Hospital, Boston.
Clinical applications of research on angiogenesis, N. Engl. J. Med. 333
(1995) 1757 – 1763.
[49] O. Musso, M. Rehn, J. Saarela, N. The´ret, J. Lie´tard, E. Hintikka, D.
Lotrian, J.P. Campion, T. Pihlajaniemi, B. Cle´ment, Collagen XVIII is
localized in sinusoids and basement membrane zones and expressed
by hepatocytes and activated stellate cells in fibrotic human liver,
Hepatology 28 (1998) 98 – 107.
[50] Y. Muragaki, S. Timmons, C.M. Griffith, S.P. Oh, B. Fadel, T.
Quertermous, B.R. Olsen, Mouse Col18a1 is expressed in a tissue-
specific manner as three alternative variants and is localized in
basement membrane zones, Proc. Natl. Acad. Sci. U. S. A. 92 (1995)
8763 – 8767.
[51] G.A. Homandberg, J.E. Williams, D. Grant, B. Schumacher, R.
Eisenstein, Heparin-binding fragments of fibronectin are potent
inhibitors of endothelial cell growth, Am. J. Pathol. 120 (1985)
327 – 332.
[52] M.S. O’Reilly, L. Holmgren, Y. Shing, C. Chen, R.A. Rosenthal, M.
Moses, W.S. Lane, Y. Cao, E.H. Sage, J. Folkman, Angiostatin: a
novel angiogenesis inhibitor that mediates the suppression of
metastases by a Lewis lung carcinoma, Cell 79 (1994) 315 – 328.
[53] O. Musso, N. The´ret, R. Heljasvaara, M. Rehn, B. Turlin, J.P.
Campion, T. Pihlajaniemi, B. Cle´ ment, Tumor hepatocytes and
basement membrane-producing cells specifically express two different
forms of the endostatin precursor, collagen XVIII, in human liver
cancers, Hepatology 33 (2001) 868–876.
R. Heljasvaara et al. / Experimental Cell Research 307 (2005) 292– 304 303
[54] T. Iisaza, H. Chang, M. Suzuki, M. Otsuji, S. Yokoi, M. Chiyo, S.
Motohashi, K. Yasufuku, Y. Sekine, A. Iyoda, K. Shibuya, K.
Hiroshima, T. Fujisawa, Overexpression of collagen XVIII is
associated with poor outcome and elevated levels of circulating serum
endostatin in non-small cell lung cancer, Clin. Cancer Res. 10 (2004)
5361 – 5366.
[55] B.B. Knowles, C.C. Howe, D.P. Aden, Human hepatocellular
carcinoma cell lines secrete the major plasma proteins and hepatitis
B surface antigen, Science 209 (1980) 497 – 499.
[56] T.A. Giambernardi, G.M. Grant, G.P. Taylor, R.J. Hay, V.M. V.M.
Maher, J.J. McCormick, R.J. Klebe, Overview of matrix metallo-
proteinase expression in cultured human cells, Matrix Biol. 16 (1998)
483 – 496.
[57] G.M. Grant, T.A. Giambernardi, A.M. Grant, R.J. Klebe, Overview of
expression of matrix metalloproteinases (MMP-17, MMP-18, and
MMP-20) in cultured human cells, Matrix Biol. 18 (1999) 145 – 148.
[58] J.D. Bartlett, J.P. Simmer, J. Xue, H.C. Margolis, E.C. Moreno,
Molecular cloning and mRNA tissue distribution of a novel matrix
metalloproteinase isolated from porcine enamel organ, Gene 183
(1996) 123 – 128.
[59] N. Hiraoka, E. Allen, I.J. Apel, M.R. Gyetko, S.J. Weiss, Matrix
metalloproteinases regulate neovascularization by acting as pericel-
lular fibrinolysins, Cell 95 (1998) 365 – 377.
[60] Y. Hamano, M. Zeisberg, H. Sugimoto, J.C. Lively, Y. Maeshima, C.
Yang, R.O. Hynes, Z. Werb, A. Sudhakar, R. Kalluri, Physiological
levels of tumstatin, a fragment of collagen IV alpha3 chain, are
generated by MMP-9 proteolysis and suppress angiogenesis via
alphaV beta3 integrin, Cancer Cell 3 (2003) 589– 601.
[61] Y.M. Kim, J.W. Jang, O. Lee, J. Yeon, E.Y. Choi, K.W. Kim, S.T. Lee,
Y.G. Kwon, Endostatin inhibits endothelial and tumor cellular
invasion by blocking the activation and catalytic activity of matrix
metalloproteinase, Cancer Res. 60 (2000) 5410– 5413.
[62] S.J. Lee, J.W. Jang, O.H. Lee, J. Yeon, E.Y. Choi, K.W. Kim, S.T. Lee,
Y.G. Kwon, Endostatin binds to the catalytic domain of matrix
metalloproteinase-2, FEBS Lett. 519 (2002) 147 – 152.
[63] P. Nyberg, P. Heikkila¨ , T. Sorsa, J. Luostarinen, R. Heljasvaara, U.H.
Stenman, T. Pihlajaniemi, T. Salo, Endostatin inhibits human tongue
carcinoma cell invasion and intravasation and blocks the activation of
matrix metalloprotease-2, -9, and -13, J. Biol. Chem. 278 (2003)
22404 – 22411.
[64] L.M. Coussens, B. Fingleton, L.M. Matrisian, Matrix metalloprotei-
nase inhibitors and cancer: trials and tribulations, Science 295 (2002)
2387 – 2392.
[65] N. Yamaguchi, B. Anand-Apte, M. Lee, T. Sasaki, N. Fukai, R.
Shapiro, I. Que, C. Lowik, R. Timpl, B.R. Olsen, Endostatin
inhibits VEGF-induced endothelial cell migration and tumor
growth independently of zinc binding, EMBO J. 18 (1999)
4414 – 4423.
[66] M.G. Cattaneo, S. Pola, P. Francescato, F. Chillemi, L.M. Vicentini,
Human endostatin-derived synthetic peptides possess potent antian-
giogenic properties in vitro and in vivo, Exp. Cell Res. 283 (2003)
230 – 236.
[67] F. Chillemi, P. Francescato, E. Ragg, M.G. Cattaneo, S. Pola, L.
Vicentini, Studies on the structure – activity relationship of endostatin:
synthesis of human endostatin peptides exhibiting potent antiangio-
genic activities, J. Med. Chem. 46 (2003) 4165– 4172.
[68] S. Wickstro¨ m, K. Alitalo, J. Keski-Oja, An endostatin-derived peptide
interacts with integrins and regulates actin cytoskeleton and migration
of endothelial cells, J. Biol. Chem. 279 (2004) 20178 – 20185.
[69] K. Eriksson, P. Magnusson, J. Dixelius, L. Claesson-Welsh, M.J.
Cross, Angiostatin and endostatin inhibit endothelial cell migration in
response to FGF and VEGF without interfering with specific intra-
cellular signal transduction pathways, FEBS Lett. 536 (2003) 19– 24.
[70] T. Sasaki, H. Larsson, D. Tisi, L. Claesson-Welsh, E. Hohenester, R.
Timpl, Endostatins derived from collagens XV and XVIII differ in
structural and binding properties, tissue distribution and anti-angio-
genic activity, J. Mol. Biol. 301 (2000) 1179 – 1190.
[71] B. Cle´ment, O. Musso, J. Lie´ tard, N. The´ret, Homeostatic control of
angiogenesis: a newly identified function of the liver? Hepatology 29
(1999) 621 – 623.
R. Heljasvaara et al. / Experimental Cell Research 307 (2005) 292– 304304
