Cardiovascular Research 55 (2002) 864–869
www.elsevier.com / locate / cardiores
Matrix metalloproteinase inhibition reduces adventitial thickening and
collagen accumulation following balloon dilation
Marion J. Sierevogel a,b , *, Evelyn Velema a , Freek J. van der Meer a , Manon Oude Nijhuis a ,
Mirjam Smeets a,b , Dominique P.V. de Kleijn a,b , Cornelius Borst a , Gerard Pasterkamp a,b
a
Experimental Cardiology Laboratory, University Medical Center, Utrecht, The Netherlands
b
Interuniversity Cardiology Institute of the Netherlands, Utrecht, The Netherlands
Received 21 January 2002; accepted 15 April 2002
Abstract
Objective: Constrictive arterial remodeling following balloon angioplasty has been related to adventitial collagen accumulation and
subsequent thickening and can be prevented by matrix metalloproteinase (MMP) inhibition. Following balloon dilation, we examined the
effect of MMP inhibition on collagen turnover and the relationship between adventitial area and degree of constrictive remodeling.
Methods: In 12 non-atherosclerotic landrace pigs, balloon dilation was performed in 39 peripheral arteries with and without MMP
inhibition. Follow up with intravascular ultrasound was performed at 42 days. Collagen content was quantified using polarized light and
digital image microscopy. Procollagen expression was determined using immunochemistry and Western blotting. Results: In the MMP
inhibitor group, constrictive remodeling was inhibited at 42 days follow up. In control and MMP inhibitor groups, a positive relation was
observed between adventitial thickness and degree of constrictive remodeling (P,0.001). Adventitial thickening and adventitial collagen
content were reduced in the MMP inhibitor group (P50.002 and P50.001, respectively). Procollagen immunostaining, but not protein
analysis on Western blotting, was decreased in the MMP inhibitor group. Conclusion: MMP inhibition impaired adventitial thickening by
reduction of collagen content 42 days after balloon dilation. This might explain its inhibitory effect on constrictive remodeling.
 2002 Elsevier Science B.V. All rights reserved.
Keywords: Angioplasty; Enzyme (kinetics); Remodeling; Arteries; Extracellular matrix
1. Introduction
Constrictive arterial remodeling is the major determinant
of restenosis following balloon angioplasty. Earlier proposed explanations for constrictive remodeling are based
on adventitial changes following arterial injury [1–3].
Hypercellularity of the adventitia, myofibroblast formation
and fibrosis have been observed together with subsequent
thickening of the adventitia [2]. Lafont et al. [4] demonstrated that the ratio of adventitial area to the area of
intima1media at the lesion site correlated with chronic
constriction. Moreover, enhanced collagen breakdown and
synthesis are important features in arterial remodeling in
the first weeks following balloon angioplasty [5–8], which
*Corresponding author. Tel.: 131-30-250-7155; fax: 131-30-2522693.
E-mail address: [email protected] (M.J. Sierevogel).
results in an increase of collagen at the site of injury.
Bakker et al. [9] suggested that constrictive remodeling
follows chronic vasoconstriction by reposition and crosslinking of fresh collagen in the arterial wall.
Matrix metalloproteinases (MMPs) play important roles
in physiological as well as pathological processes including collagen turnover [6], restenosis [6,10,11], and heart
failure [12,13]. Their activities increase after balloon
angioplasty [7,14]. Previously, we demonstrated that oral
MMP inhibition significantly inhibits constrictive arterial
remodeling following balloon dilation in favor of both
neutral — and expansive remodeling [10]. With intravascular ultrasound (IVUS), we showed that only a short
duration (14–28 days) of MMP inhibition is sufficient to
block the constrictive remodeling response at 42 days
follow up [11].
Time for primary review 18 days.
0008-6363 / 02 / $ – see front matter  2002 Elsevier Science B.V. All rights reserved.
PII: S0008-6363( 02 )00467-4
M. J. Sierevogel et al. / Cardiovascular Research 55 (2002) 864 – 869
The mechanisms underlying inhibition of constrictive
remodeling by MMP inhibition are unknown. Differences
in collagen accumulation might explain the altered remodeling response. In the present study, we determined the
effect of MMP inhibition on collagen turnover by assessing collagen content, density and procollagen expression
in MMP inhibited and non-inhibited porcine vessels that
were obtained after balloon dilation. Furthermore, we
assessed the relationship between adventitial area and the
degree of constrictive remodeling at 42 days following
balloon dilation for both the control and the MMP inhibitor
group.
865
2. Methods
balloon / artery ratio of approximately 1.2, the balloon was
inflated three times for 1 min at a pressure of 6–10 atm. A
continuous infusion of nitroglycerin (20 mg / min) was
given to prevent arterial spasm. To assess arterial remodeling, intravascular ultrasound (IVUS) was performed before
intervention, immediately after intervention and at follow
up, using a 4.1 F, 30 MHz ultrasound catheter with an
axial resolution of 0.1 mm (Princeps  , DuMed / EndoSonics, Rijswijk, The Netherlands). Fluoroscopy was
performed during IVUS in order to document the location
of the images relative to an anatomic landmark for
adequate matching. After angiography and IVUS at follow
up, the animals were euthanisized by an overdose of
pentobarbital. Vessels were harvested, snapfrozen in liquid
nitrogen and stored at 280 8C until further analysis.
2.1. Animal model
2.4. IVUS-analysis
Twelve non-atherosclerotic landrace pigs with an average weight of 20 kg were studied. Balloon dilation was
performed in femoral and internal iliac arteries. Animals
were treated with a MMP inhibitor or served as controls.
Arteries (n539) were harvested at 42 days (seven control,
five MMP inhibitor pigs) follow up. The investigation was
approved by the ethical committee on Animal Experimentation of the University Medical Center Utrecht
and conforms with the Guide for the Care and Use of
Laboratory Animals published by the U.S. National Institutes of Health (NIH Publication No. 85-23, revised 1996).
The IVUS images were analyzed at regular intervals
(every 0.5 cm). In each IVUS image, lumen area (LA) and
vessel area (VA) were measured as before [10,11].
Anatomic landmarks were used to match the imaged
locations at different time-points. Untreated segments
(proximal parts of external iliac arteries, parts proximal
and distal to the dilated segments of the femoral arteries,
parts distal to the dilated segments of the internal iliac
arteries) were used for correction of growth of all areas at
42 days follow up as follows: mean pr 2 pre intervention /
mean pr 2 termination. The radius (r) was determined
angiographically. The provided value was multiplied with
the LA and VA measurements determined at termination.
Within the balloon dilated segment, the location with the
smallest LA at follow up was clearly identified and
selected for further calculations.
Definitions:
2.2. MMP-inhibition
The oral non-specific synthetic MMP inhibitor Marimastat (BB-2516) was supplied by British Biotech Pharmaceuticals Limited, Oxford, U.K. This drug inhibits in the
nanomolar range all known classes of MMPs. Marimastat
was administered for 0 (control) or 42 days, 10 mg / kg
twice a day, starting 1 day prior to the intervention. An
unpublished dosimetry study in landrace pigs in our lab
established that a dose of 10 mg / kg Marimastat twice a
day would give exposure of 100–200 ng / ml plasma, a
concentration that had previously been effective in animal
models of cancer. In a previous study [10], 1–10 ng / ml
functional Marimastat was detected in vessel extracts of
Marimastat treated animals.
2.3. Intervention
Balloon dilation was performed as described previously
[10,11]. In short, the arterial tree was accessed through a
carotid approach. For balloon dilation, a standard peripheral balloon catheter (Cordis, Opta 5, length 2–4 cm,
diameter 4–6 mm) was used. In the femoral arteries,
balloon dilation was performed approximately 1 cm distal
to the lateral femoral circumflex. In the internal iliac
arteries, the mid-segments were balloon dilated. With a
• Loss in VA, being a measure of remodeling5VA post
intervention2VA follow up.
• Remodeling Index (RI)5VA follow up / VA post intervention. A value of .1 indicates expansive remodeling.
A value of ,1 indicates constrictive remodeling.
2.5. Histomorphometry and quantification of collagen
content
Quantification of total collagen content and collagen
density was performed in intima, media and adventitia
using picro Sirius red staining and digital image microscopy with circularly polarized light as described by
Smeets et al. [15]. In short, a picro Sirius red image was
converted into a grey value image. Regions of interest
(ROI) were drawn to select the different arterial layers.
The total amount of grey values in each layer was
determined for total collagen content. Collagen density
was calculated by dividing the collagen content in each
layer by the cross-sectional area of this layer.
866
M. J. Sierevogel et al. / Cardiovascular Research 55 (2002) 864 – 869
Per dilated segment, two cross sections were selected at
fixed locations (proximally and distally) and analysed by
observers who were blinded for the different groups.
Measurements of the two cross-sections were averaged
for each artery. In addition, non-dilated femoral artery
segments, approximately 2 cm distally of the treated area,
were analysed.
Definitions:
? Injury score (IS), based on rupture of the internal (IEL)
and external (EEL) elastic lamina:
✓0: intimal hyperplasia absent; both IEL and EEL
intact
✓1: intimal hyperplasia present, both IEL and EEL
intact;
✓2: IEL ruptured, EEL intact;
✓3: both IEL and EEL ruptured.
? Adventitial area index5adventitial area histomorphometry at
follow up / vessel area IVUS at follow up.
by goat–anti-mouse (DAKO A / S Denmark). Detection
was performed using Chemiluminescence Reagent Plus
(Renaissance  , NEN) and exposure to X-Omat Blue XB-1
films (Kodak).
2.7. Statistics
Data are presented as mean6standard error of the mean.
SPSS 9.0 was used for all statistical calculations. The
independent t-test with post hoc Bonferroni correction was
used to compare normally distributed data among groups.
A Mann–Whitney test was used to compare groups when
data were not normally distributed. A Pearson’s linear
regression analysis was performed to study the relation
between adventitial area and the degree of constrictive
remodeling. P,0.05 was considered to be statistically
significant.
3. Results
2.6. Quantification of procollagen
2.6.1. Immunohistochemistry
To determine the area of procollagen producing cells, a
mouse anti-sheep procollagen I NH 2 -terminal monoclonal
antibody SP1.D8 (1:10, Developmental Studies Hybridoma
Bank) was used. Sections (8 mm) were fixed in acetone
containing 0.03% hydrogen peroxide and preincubated
with 10% normal goat serum in 1% PBSA for 30 min at
room temperature (RT). Subsequently, the sections were
incubated with SP1.D8 diluted in 0.1% PBSA for 1 h at
RT and washed three times in PBS. Incubation with the
secondary antibody goat anti-mouse peroxidase (1 / 250,
DAKO A / S Denmark) in 1% PBSA containing 5% normal
swine serum was for 1 h at RT. Visualization of the
sections was performed with diaminobenzidine (DAB).
Negative controls were carried out with IgG 1 isotype or by
omitting the primary antibody. At the site of maximal
injury, the sections were analyzed quantitatively by light
microscopy using Sis-analysis 2.1 software. Sections were
carefully studied and color thresholds were set and adjusted until the computerized detection met the visual
interpretation of positive staining. For all cross-sections,
the intima, media and adventitia were analyzed separately.
The degree of procollagen staining was expressed in
percentage of unit of intima / adventitial / medial area.
2.6.2. Western blotting
To quantify procollagen, protein was isolated using
TriPure姠 Isolation Reagent (Boehringer Mannheim).
Same amounts of proteins (5 mg) were separated on a 6%
SDS–PAGE and subsequently blotted onto Hybond-ECL.
To block the non-specific binding sites, the membrane was
put overnight at 4 8C in PBS containing 0.1% Tween 20
(Merck) and 5% protifar (Nutricia) (PBSTB). The blot was
incubated for 1 h with SP1-D8 (1:100 in PBSTB) followed
Gain in weight did not differ among groups: 11.360.6
in the control group versus 10.461.0 in the MMP inhibitor
group at 42 days follow up (P50.65). The growth correction factor was also not different among groups: 0.81 in the
control group versus 0.82 in the MMP inhibitor group.
At 42 days following balloon dilation, constrictive
remodeling was observed in control vessels but not in the
MMP inhibitor treated vessels: VA loss 3.0760.93 mm 2
versus 20.4760.74 mm 2 (P50.005).
In both the control and the MMP inhibitor group, an
inverse relation was observed between adventitial area
index and the remodeling index at follow up (Fig. 1,
p#0.005). Note the smaller averaged adventitial area index
Fig. 1. Adventitial area index versus remodeling index at 42 days follow
up. Control group (square markers): y 5 0.68X 1 1.08, r 5 0.56, P 5
0.005. MMP inhibitor group (plus markers): y 5 2 1.32X 1 1.30, r 5
0.77, P 5 0.001. Adventitial area index5adventitial area at follow up /
vessel area at follow up. Remodeling index5vessel area at follow
up / vessel area post-intervention.
M. J. Sierevogel et al. / Cardiovascular Research 55 (2002) 864 – 869
867
in the MMP inhibitor group: mean adventitial area index
0.2160.03 versus 0.3960.04 in the control group (P5
0.001).
Fig. 2 shows the relation between the degree of injury
and adventitial area for control and MMP inhibitor groups
at 42 days follow up. In both groups, adventitial area
increased with increasing injury score. A smaller adventitial area was observed in the MMP inhibitor group
irrespective of injury score.
Fig. 3 demonstrates total adventitial collagen content, as
well as its two determinants, adventitial collagen density
and adventitial area for both the control and the MMP
inhibitor pigs at follow up. In the control group, an
increase in collagen content and adventitial area was
observed in the dilated segments compared with the nondilated segments (total adventitial collagen content:
210630 grey values * 10 6 , adventitial area 2.9603 mm 2 ,
collagen density 69.464.1 grey values / mm 2 in non-dilated
segments) (total collagen P,0.0001, adventitial area P5
0.001). In the MMP inhibitor group, however, total
collagen at 42 days follow up was lower due to a smaller
adventitial area (P50.001 and P50.002, respectively).
Table 1 depicts the medial and intimal collagen content,
density and area for the different groups. At 42 days,
density did not differ among groups irrespective of the
arterial layer. However, total collagen differed between
control and MMP inhibitor group in the intimal layer
(P50.04).
In all layers of both control and MMP inhibitor groups,
the area of procollagen expressing cells was increased
following balloon dilation at 42 days (media 0.360.1
versus 6.061.3% and adventitia 0.760.2 versus 2.560.4%
in non-dilated arterial segments and control dilated arterial
segments, respectively). MMP inhibition resulted in less
procollagen stained cells in the media (3.761.3%, P5
Fig. 3. Total adventitial collagen content, adventitial area and adventitial
collagen density for both the control (C) and the MMP inhibitor pigs (M).
Fig. 2. Adventitial area in the different injury categories at 42 days
follow up for control (solid bars) and MMP inhibitor group (striped bars).
* P50.01 and 0.05 compared to control.
0.09) and the adventitia (1.960.6%, P50.048). MMP
inhibition did not influence procollagen staining in the
intimal layer (5.261.4 versus 4.660.8% in the control
group, P50.82). At protein level, no differences were
observed in procollagen expression in the MMP inhibitor
group compared to the control group (0.760.2 versus
0.960.3 ODu * mm 2 in the control group). Differences in
protein content were also not observed using Western blot
analysis in another, smaller group of pigs terminated at 2,
7, and 14 days (n52 for MMP inhibitor and control
groups) (data not shown).
M. J. Sierevogel et al. / Cardiovascular Research 55 (2002) 864 – 869
868
Table 1
Medial and intimal total collagen content (total), density and area for the different groups at 42 days follow up
Media
Control (n523)
MMPi (n516)
Intima
Total
Density
Area
Total
Density
Area
10.361.1
8.961.3
44.063.0
33.963.6
2.360.1
2.560.2
3.860.7
2.061.0
30.563.1
22.563.8
1.060.2
0.760.2
Values are mean6S.E.M., total collagen (total) in grey values * 10 7 , collagen density in grey values / mm 2 , area in mm 2 .
4. Discussion
The principal findings of the present study are (1) in
both the control and the MMP inhibitor group, a positive
relation between adventitial area and the degree of constrictive remodeling was evident at 42 days follow up; (2)
after MMP inhibition, total adventitial collagen content
and adventitial area were significantly reduced at follow
up; (3) the reduction in collagen content could not be
explained by a reduction in collagen density or altered
procollagen expression at protein level.
The adventitia is the first layer to respond to injury and
is assumed to play an essential role in constrictive arterial
remodeling following balloon angioplasty [1–3]. Three to
7 days after injury, the adventitial layer becomes hypercellular due to the proliferation of fibroblasts [2].
Between 7 and 14 days, the adventitial fibroblasts change
into myofibroblasts, followed by accumulation of collagen
containing scar tissue within the adventitia [2]. These
changes are accompanied by thickening of the adventitial
layer. In the present study, constrictive remodeling was
observed in the control group at 42 days following balloon
angioplasty. In the MMP inhibitor group, however, constrictive remodeling was significantly reduced (82%) at 42
days follow up. In both control and MMP inhibitor groups,
the adventitial area index was inversely related with the
remodeling index. However, adventitial area at 42 days
was smaller in the MMP inhibitor group, which might
explain the inhibitory effect on constrictive remodeling.
This reduction in adventitial area could not be explained
by less severe injury, since adventitial area in the MMP
inhibitor group appeared smaller compared to the control
group for all injury score categories. In addition to the
smaller adventitial area in the MMP inhibitor group at 42
days follow up, a lower total adventitial collagen content
was observed. Although the area of procollagen producing
cells was reduced in the MMP inhibitor group, on a protein
level, altered procollagen expression did not differ among
groups. Enhanced breakdown might be an alternative
explanation for the altered collagen content by MMP
inhibition. However, this is very unlikely since MMP
inhibitors are known to inhibit collagen breakdown
[16,17]. Yet, it seems that MMPs are involved in collagen
production or cross-linking apart from its role in collagen
breakdown. This hypothesis is supported by the study by
Strauss et al. [6] that showed that MMP inhibition reduces
collagen content in the injured rabbit vessel wall. Uzel et
al. [18] demonstrated in mouse embryo fibroblasts cultures,
that members of the metalloproteinase family (bone morphogenetic protein 1-related) process pro-lysyl oxidase and
control lysyl oxidase activation. Lysyl oxidase catalyzes
the final enzymatic step required for collagen and elastin
cross-linking in extracellular matrix biosynthesis. Changes
in the process of cross-linking of collagen molecules are
associated with defects in the biomechanical stability of
the extracellular matrix and might result in enhanced
breakdown of the adventitia which is the layer containing
most collagen in the vessel wall.
4.1. Limitations
One must keep in mind that MMPs play an important
role in several other processes besides collagen turnover,
like amplification of the initial signaling coagulation
pathway, and in the absorption and utilization of dietary
proteins. It is unknown how these actions affect the injury
response and other systems of the body.
The comparison between histology and IVUS is limited
by the fact that the vessels were not pressure-fixed after
sacrifice, whereas the IVUS measurements were performed
at arterial blood pressure. Post mortem shrinkage of the
pathology specimens, in part by preservation and staining
techniques, may confound these observations, but might
expected to be similar among groups.
Animals were terminated at 6 weeks after intervention.
This duration of follow up merits careful consideration. It
was assumed that this time frame reflects the 6 months
follow up in humans. We cannot exclude that the remodeling response continues after 6 weeks which could affect
the observations of this study.
In conclusion, constrictive arterial remodeling following
balloon angioplasty was positively related with adventitial
growth. MMP inhibition significantly reduced total adventitial collagen content at 42 days follow up due to a
significant reduction in adventitial area, which might
explain the inhibitory effect of MMP inhibition on constrictive arterial remodeling.
Acknowledgements
This study was supported by grants from the Sorbo
Foundation, the Netherlands Heart Foundation (99209) and
M. J. Sierevogel et al. / Cardiovascular Research 55 (2002) 864 – 869
the Netherlands Organization for Scientific Research
(NWO 902-16-239 and 902-16-222).
[10]
References
[1] Scott NA, Cipolla GD, Ross CE, Dunn B, Martin FH, Simonet L,
Wilcox IN. Identification of a potential role for the adventitia in
vascular lesion formation after balloon overstretch injury of porcine
coronary arteries. Circulation 1996;93:2178–2187.
[2] Shi Y, Pieniek M, Fard A, O’Brien J, Mannion JD, Zalewski A.
Adventitial remodeling after coronary arterial injury. Circulation
1996;93:340–348.
[3] Labinaz M, Pels K, Hoffert C, Aggarwal S, O’Brien ER. Time
course and importance of neoadventitial formation in arterial
remodeling following balloon angioplasty of porcine coronary
arteries. Cardiovasc Res 1999;41:255–266.
[4] Lafont A, Durand E, Samuel JL, Besse B, Addad F, Levy BI,
Desnos M, Guerot C, Boulanger CM. Endothelial dysfunction and
collagen accumulation. Two independent factors for restenosis and
constrictive remodeling after experimental angioplasty. Circulation
1999;100:1109–1115.
[5] Strauss BH, Chisholm RJ, Keeley FW, Gotlieb AI, Logan RA,
Armstrong PW. Extracellular matrix remodeling after balloon angioplasty injury in a rabbit model of restenosis. Circ Res
1994;75:650–658.
[6] Strauss BH, Robinson R, Batchelor WB, Chisholm RJ, Ravi G,
Natarajan MK, Logan RA, Mehta SR, Levy DE, Ezrin AM, Keeley
FW. In vivo collagen turnover following experimental balloon
angioplasty injury and the role of matrix metalloproteinases. Circ
Res 1996;79:541–550.
[7] Karim MA, Miller DD, Farrar MA, Eleftheriades E, Reddy BH,
Breland CM, Samarel AM. Histomorphometric and biochemical
correlates of arterial procollagen gene expression during vascular
repair after experimental angioplasty. Circulation 1995;91:2049–
2057.
[8] Lafont A, Guzman LA, Whitlow PL, Goormastic M, Cornhill JE,
Chisolm GM. Restenosis after experimental angioplasty. Intimal,
medial, and adventitial changes associated with constrictive remodeling. Circ Res 1995;76:996–1002.
[9] Bakker EN, van der Meulen ET, van den Berg BM, Everts V, Spaan
[11]
[12]
[13]
[14]
[15]
[16]
[17]
[18]
869
JA, VanBavel E. Inward remodeling follows chronic vasoconstriction
in isolated resistance arteries. J Vasc Res 2002;39:12–20.
Sierevogel MJ, Pasterkamp G, Velema E, de Jaegere PP, de Smet BJ,
Verheijen JH, de Kleijn DP, Borst C. Oral matrix metalloproteinase
inhibition and arterial remodeling after balloon dilation: an intravascular ultrasound study in the pig. Circulation 2001;103:302–
307.
Sierevogel MJ, Pasterkamp G, Velema E, de Kleijn DPV, de Jaegere
PPT, Borst C. Minimal duration of oral matrix metalloproteinase
inhibition to prevent constrictive arterial remodeling following
balloon dilation in the pig. Radiology 2002;222:468–473.
Thomas CV, Coker ML, Zellner JL, Handy JR, Crumbley AJ 3rd,
Spinale FG. Increased matrix metalloproteinase activity and selective upregulation in LV myocardium from patients with end-stage
dilated cardiomyopathy. Circulation 1998;97:1708–1715.
Peterson JT, Hallak H, Johnson L, Li H, O’Brian PM, Sliskovic DR,
Bocan TM, Coker ML, Etoh T, Spinale FG. Matrix metalloproteinase inhibition attenuates left ventricular remodeling and dysfunction
in a rat model of progressive heart failure. Circulation
2001;103:2303–2309.
Zempo N, Kenagy RD, Au YP, Bendeck M, Clowes MM, Reidy
MA, Clowes AW. Matrix metalloproteinases of vascular wall cells
are increased in balloon-injured rat carotid artery. J Vasc Surg
1994;20:209–217.
Smeets MB, de Kleijn DPV, Perree J, Pasterkamp G, Voorbij HAM,
de Smet BJGL, Borst C. Collagen distribution and synthesis after
balloon angioplasty: a temporal course in the atherosclerotic yucatan
micropig. Abstract. Eur Heart J 2000;1587.
Cawston T, Plumpton T, Curry V, Ellis A, Powell L. Role of TIMP
and MMP inhibition in preventing connective tissue breakdown.
Ann NY Acad Sci 1994;732:75–83.
Creemers LB, Jansen ID, Docherty AJ, Reynolds JJ, Beertsen W,
Everts V. Gelatinase A (MMP-2) and cysteine proteinases are
essential for the degradation of collagen in soft connective tissue.
Matrix Biol 1998;17:35–46.
Uzel MI, Scott IC, Babakhanlou-Chase H, Palamakumbura AH,
Pappano WN, Hong HH, Greenspan DS, Trackman PC. Multiple
bone morphogenetic protein 1-related mammalian metalloproteinases process pro-lysyl oxidase at the correct physiological site and
control lysyl oxidase activation in mouse embryo fibroblast cultures.
J Biol Chem 2001;276:22537–22543.