?The?Journal?of?Clinical?Investigation http://www.jci.org Volume 119 Number 12 December 2009
Mfge8 diminishes the severity of tissue
fibrosis in mice by binding and targeting
collagen for uptake by macrophages
Kamran Atabai,1,2 Sina Jame,1,2 Nabil Azhar,1,2 Alex Kuo,1,2 Michael Lam,1,2 William McKleroy,1,2
Greg DeHart,1,2 Salman Rahman,1,2 Dee Dee Xia,1,2 Andrew C. Melton,1,2 Paul Wolters,1,2
Claire L. Emson,3 Scott M. Turner,3 Zena Werb,4 and Dean Sheppard1,2
1Lung Biology Center, Cardiovascular Research Institute, and 2Department of Medicine, UCSF, San Francisco, California, USA.
3KineMed Inc., Emeryville, California, USA. 4Department of Anatomy, UCSF, San Francisco, California, USA.
Fibrotic diseases are characterized by replacement of normal tissue
architecture with collagen-rich matrix, disrupting organ function
(1–4). In the lung, fibrosis can occur due to abnormal remodeling
after acute lung injury, in the setting of systemic autoimmune and
inflammatory disease, or as an idiopathic process (5). The produc-
tion, deposition, and removal of collagen are dynamic processes, with
the balance between collagen production and removal determining
tissue architecture (6). When persistent collagen production outpac-
es or overwhelms mechanisms that remove collagen, excess collagen
is deposited in the extracellular matrix, leading to tissue fibrosis.
The molecular pathways responsible for collagen turnover
remain incompletely described. Collagen turnover occurs by two
pathways, extracellular proteolytic cleavage (7) and endocyto-
sis followed by lysosomal degradation (6). Extracellular collagen
cleavage facilitates subsequent intracellular uptake (8). Very little
is known about molecules that mediate collagen endocytosis.
Milk fat globule epidermal growth factor 8 (Mfge8) is a soluble
glycoprotein that negatively regulates inflammation and autoim-
munity by facilitating the clearance of apoptotic cells (9–11). Since
fibrosis can occur as a consequence of exaggerated apoptosis and
inflammation (12, 13), we hypothesized that Mfge8 may function
as a negative regulator of tissue fibrosis. In this report, we show
that mice deficient in Mfge8 (Mfge8–/– mice) develop exaggerated
pulmonary fibrosis after bleomycin treatment. Surprisingly, how-
ever, this phenotype is not a consequence of impaired apoptotic
cell clearance, exaggerated inflammation, or a more severe acute
phase of lung injury. Instead, we find that Mfge8–/– mice have a
defect in collagen turnover in vivo that is caused by a previously
unknown role for Mfge8 in binding and targeting collagen for
uptake by macrophages. Mfge8–/– macrophages have impaired
collagen uptake. We further identify the first discoidin domain
of Mfge8 as sufficient for collagen binding and internalization.
In this work, we show what we believe to be the first pathway by
which a secreted glycoprotein binds collagen and targets it for
removal from the extracellular matrix.
Mfge8 is expressed throughout the lung, and expression is increased after
injury. To determine the lung expression pattern of Mfge8, we
stained sections taken from adult Mfge8+/+ mice with the anti-
Mfge8 antibody 4F6 (9). Mfge8 was present in the alveolar inter-
stitium as well as in the pulmonary vasculature (Figure 1, A and
B). Alveolar macrophages obtained by bronchoalveolar lavage
(BAL) stained positively for Mfge8 (Figure 1C). To determine
whether lung injury induced Mfge8 expression, we challenged
mice with the chemotherapeutic agent bleomycin. Bleomycin
induces an early acute lung injury response (week 1) character-
ized by vascular leak and inflammation followed by pulmonary
fibrosis (14). While by immunohistochemical analysis, all saline-
treated alveolar macrophages expressed some Mfge8 (Figure 1C),
the intensity of expression was increased 5 days after bleomy-
cin administration (Figure 1D). We also evaluated whole-lung
expression of Mfge8. Mfge8 was induced in the first week after
injury. Interestingly, increased expression persisted at 14, 21,
and 28 days, suggesting a role for Mfge8 in the fibrotic stage of
bleomycin injury (Figure 1E). To determine whether the human
ortholog of Mfge8, lactadherin (15, 16), was induced in fibrotic
disease in humans, we evaluated expression in samples taken
from the lungs of patients with idiopathic pulmonary fibrosis
Conflict?of?interest: S.M. Turner and C.L. Emson received income and research sup-
port as employees of KineMed Inc.
Citation?for?this?article: J. Clin. Invest. 119:3713–3722 (2009). doi:10.1172/JCI40053.
3714? The?Journal?of?Clinical?Investigation http://www.jci.org Volume 119 Number 12 December 2009
(IPF) (Figure 1F). Lactadherin expression was increased in all 4
IPF samples evaluated as compared with control samples taken
from nonfibrotic lungs rejected for transplantation.
Mfge8 deficiency results in an exaggerated fibrotic response to bleomy-
cin. Since Mfge8 regulates apoptosis and inflammation, processes
closely linked with tissue fibrosis, and bleomycin induces Mfge8
expression in the lung, we evaluated the role of Mfge8 in bleo-
mycin-induced pulmonary fibrosis. Mfge8–/– mice were challenged
with intratracheal bleomycin (1.1 U/kg), and the severity of pul-
monary fibrosis was evaluated at 28 days. Histologically, Mfge8–/–
mice had more extensive deposition of collagen fibrils than con-
trol mice in tissue sections stained with picrosirius red (Figure 2,
A and B). To quantify lung fibrosis biochemically, we measured
total lung hydroxyproline content 28 days after bleomycin treat-
ment. In our initial phenotypic evaluation, Mfge8–/– and control
mice were maintained in a mixed-strain background (C57BL/6
× 129/Ola), and we used heterozygous littermates as controls. In
these studies, Mfge8–/– mice had significantly greater total lung
hydroxyproline content than Mfge8+/– littermate controls (Fig-
ure 2C). Once we had backcrossed mice 6 generations into the
C57BL/6 background, we confirmed our phenotype by repeating
the measurements of lung hydroxyproline content. Mfge8–/– mice
had significantly greater fibrosis than Mfge8+/+ controls 28 days
after bleomycin treatment (Figure 2D).
Enhanced fibrosis in Mfge8–/– mice is not associated with impaired apop-
totic cell clearance, increased inflammation, or a more severe acute phase
of lung injury. We next evaluated whether exaggerated fibrosis was
associated with impaired apoptotic cell clearance. We first evalu-
ated the efficacy of apoptotic cell clearance in vivo in adult Mfge8–/–
mice by administering an intratracheal bolus of apoptotic thy-
mocytes. We then quantified the phagocytic index (representing
the number of apoptotic cell inges-
tions per macrophage) of alveolar
macrophages obtained by BAL. We
found no differences in the phago-
cytic index between Mfge8–/– and
control alveolar macrophages (Fig-
ure 3A). To evaluate whether Mfge8
was important for in vivo apoptotic
cell clearance after lung injury, we
stained tissue sections for apoptotic
cells with TUNEL. In the early stage
(first 24 hours), the majority of apop-
totic cells were found in the alveolar
epithelium, whereas at subsequent
time points the majority were within
inflammatory cell infiltrates. At all
time points evaluated, we found no
significant difference in the number
of apoptotic cells in Mfge8–/– mice
(Figure 3, B and C). After bleomycin
treatment but prior to lung harvest,
we obtained BAL samples to quan-
tify the in vivo phagocytic index of
the alveolar macrophages. Once
again, we found no difference in the
phagocytic index of Mfge8–/– alveolar
macrophages (Figure 3, D and E). As
another measure of apoptotic cell
clearance, we quantified the propor-
tion (Figure 3F) and number (data not shown) of free apoptotic
cells recovered by BAL after injury and found no differences. All
of these studies except those represented in Figure 3C were carried
out in the mixed-strain background using heterozygous control
mice. In summary, the exaggerated pulmonary fibrosis in Mfge8–/–
mice does not appear to be associated with any abnormality in
clearance of apoptotic cells from the lung.
We next evaluated the severity of inflammation after bleomycin
treatment by quantifying BAL cell number and type. Mfge8 defi-
ciency did not lead to an exaggerated inflammatory response either
with high-dose bleomycin (5 U/kg, Supplemental Figure 1A; sup-
plemental material available online with this article; doi:10.1172/
JCI40053DS1) used for acute lung injury studies in a mixed-strain
background with Mfge8+/– controls or bleomycin at a dose used to
induce fibrosis (1.1 U/kg, Supplemental Figure 1B) in a pure-strain
background using Mfge8+/+ controls. There were no consistent dif-
ferences in interstitial inflammation in tissue sections (data not
shown) or in the number of infiltrating neutrophils in lung sections
as revealed by staining with anti-Gr1 antibody 24 hours after bleo-
mycin administration (Supplemental Figure 1C). We also quanti-
fied expression of myeloperoxidase in lung lysates as a marker of
neutrophil infiltration 3 days after bleomycin administration and
saw no difference between genotypes (Supplemental Figure 1D).
As an additional method to evaluate differences in inflammation,
we measured gene expression of markers of neutrophils and mac-
rophages 14 days after bleomycin treatment (Supplemental Figure
1E). Though expression of all markers increased in bleomycin- ver-
sus saline-treated mice, there were no differences between Mfge8–/–
and control mice. Next we evaluated the extent of the acute lung
injury response by quantifying the severity of lung vascular leak 5
days after bleomycin treatment (5 U/kg) and found no difference
Mfge8 expression is induced by lung injury. (A and B) Lung sections taken from adult wild-type mice
were stained with anti-Mfge8 antibody. Mfge8 was expressed in the alveolar interstitium (arrow in A)
and pulmonary endothelium (arrow in B). Scale bars: 10 μm. (C and D) Alveolar macrophages were
obtained by BAL after saline administration (C) or 5 days after bleomycin administration (5 U/kg) (D).
Mfge8 staining was present in macrophages from saline-treated animals (C), and the intensity of
expression increased after bleomycin treatment (D). Scale bars: 10 μm. (E) Ten micrograms of protein
from total lung homogenates taken at the indicated days after bleomycin treatment (1.1 U/kg) or 7
days after saline treatment was loaded for a Western blot using an anti-Mfge8 antibody. An antibody
against Crk was used to demonstrate equal loading of protein. (F) One microgram of protein from lung
homogenates obtained from human patients with IPF was loaded for a Western blot using an anti-lac-
tadherin antibody (Lact.). Control samples were from lungs rejected for transplantation. An antibody
against Crk was used to demonstrate loading of protein.
? The?Journal?of?Clinical?Investigation http://www.jci.org Volume 119 Number 12 December 2009
between Mfge8–/– mice (Supplemental Figure 1F) in a mixed-strain
background and Mfge8+/– controls or Mfge8–/– mice (Supplemental
Figure 1G) in a pure-strain background and Mfge8+/+ controls. All
subsequent studies were carried out in a pure-strain background
with Mfge8+/+ controls.
Mfge8–/– mice have impaired collagen degradation in vivo. An increase in
lung hydroxyproline content after injury can represent an increase
in production or a defect in turnover of collagen. To determine
whether Mfge8 deficiency led to an increase in col-
lagen production, we measured procollagen tran-
script production by gene expression array 14 days
after bleomycin injury (1.1 U/kg) (17, 18). While
bleomycin induced the expression of multiple pro-
collagen transcripts in both Mfge8–/– and control
mice, the relative induction was similar (Supple-
mental Figure 2A). We confirmed these results
for procollagen 1α transcripts on days 7 and 14
with real-time PCR (Supplemental Figure 2B). To
confirm that the lack of differences in procollagen
gene transcription indicated quantitatively simi-
lar collagen production in bleomycin-challenged
Mfge8–/– and control mice, we measured the rates
of collagen synthesis. We measured collagen syn-
thesis by quantifying incorporation of deuterium
into hydroxyproline (19). We replaced the drink-
ing water of mice at the time of bleomycin injec-
tion with deuterated water (2H2O) for 14 days and
then measured the incorporation of deuterium
(2H) into the stable C-H bonds of hydroxyproline
in newly synthesized collagen by gas chromatog-
raphy/mass spectroscopy (GC/MS) (19). Adminis-
tration of bleomycin induced a significant increase
in collagen synthesis; however, consistent with our
collagen transcript data, there were no significant
differences in the rates of synthesis between Mfge8–/–
and control mice (Supplemental Figure 2C). Total
lung hydroxyproline content at this time point was
also similar in Mfge8–/– and control mice (Supple-
mental Figure 2D).
We also evaluated whether Mfge8–/– mice had
an in vivo increase in myofibroblasts, the cell
type that produces collagen after bleomycin-induced injury (20).
We found no difference in induction of α–smooth muscle actin
expression (a marker of myofibroblasts) (Supplemental Figure
2E) or in the number of myofibroblasts in the lung as determined
by immunohistochemical quantification (Supplemental Figure
2F). These data indicate that the increase in fibrosis in Mfge8–/–
mice was due to a defect in collagen removal rather than an
increase in collagen production.
Mfge8–/– mice develop exaggerated pulmonary fibrosis after injury. (A and B) Picrosirius red staining of lung sections from Mfge8+/+ (A) and Mfge8–/–
(B) mice taken 28 days after bleomycin administration (1.1 U/kg). Scale bars: 100 μm. (C and D) Biochemical analysis of bleomycin-induced pul-
monary fibrosis. Pulmonary fibrosis measured by total lung hydroxyproline content at baseline and 28 days after bleomycin treatment in mice in a
mixed-strain background (C) or pure-strain background (D). Data are presented as mean ± SEM (n = 6–8 for saline [Sal] and 10–14 for bleomycin
[Bleo] treatment groups). *P = 0.004 (C) and *P = 0.01 (D) using Student’s t test to compare bleomycin-treated Mfge8+/+ and Mfge8–/– mice.
Mfge8–/– mice have intact apoptotic cell clearance in vivo after bleomycin treatment. (A)
Apoptotic thymocytes were instilled intratracheally and the phagocytic index of alveolar
macrophages obtained by BAL 30 minutes after instillation was similar in Mfge8–/– and
Mfge8+/– controls (n = 5–6). (B) Tissue sections taken from Mfge8–/– and Mfge8+/– mice at
the indicated hours and days after bleomycin treatment (5 U/kg) were stained by TUNEL
assay, and the number of apoptotic cells per ×200 fields (15 fields) was quantified (n = 4–10).
(C) Sections taken from Mfge8–/– and Mfge8+/+ mice 9 days after bleomycin (1.1 U/kg)
treatment were stained by TUNEL as described in B (n = 3–4). (D–F) Cytospin prepara-
tions from BAL samples after bleomycin treatment (5 U/kg) were stained with Diff-Quick
(D) or TUNEL (F), and the number of apoptotic cell ingestions (arrows in D; and E) and
percentage of free apoptotic nuclei (F) were quantified (n = 5–8). There was no differ-
ence in number of TUNEL-positive cells, alveolar macrophage phagocytic index, or free
apoptotic nuclei between Mfge8–/– and control samples. All comparisons were made
using Student’s t test, and data are expressed as mean ± SEM.
3716?The?Journal?of?Clinical?Investigation http://www.jci.org Volume 119 Number 12 December 2009
Mfge8 targets collagen for uptake by macrophages. Collagen is cleaved
in the extracellular matrix by proteases and subsequently internal-
ized and degraded in lysosomes (8). Mfge8 contains 2 discoidin
domains, homologous to those present in the collagen receptors
DDR1 and DDR2 (21). We therefore hypothesized that Mfge8
might bind and mediate uptake of collagen in areas of alveolar
scarring. To test this hypothesis, we designed an in vitro collagen
phagocytosis assay (Figure 4A). Alveolar macrophages were cul-
tured with FITC-conjugated type I collagen, and internalization
of collagen was quantified by fluorescence microscopy. In initial
studies (Figure 4B), we showed that addition of increasing doses
of type I collagen that was not conjugated with FITC inhibited the
uptake of FITC-conjugated collagen in a dose-dependent fashion.
We then compared uptake of FITC-conjugated collagen by alveo-
lar macrophages taken from Mfge8–/– and Mfge8+/+ mice. Alveolar
macrophages from Mfge8–/– mice had significantly impaired colla-
gen uptake as measured by the collagen uptake index (Figure 4C).
Addition of recombinant Mfge8 (rMfge8) restored the collagen
uptake index to Mfge8+/+ levels in a dose-dependent fashion (Fig-
ure 4C). In the absence of serum, addition of rMfge8 also increased
Mfge8+/+ macrophage collagen uptake index (Figure 4D).
We next evaluated collagen uptake by alveolar macrophages in
vivo by placing FITC-conjugated type I collagen into the airways
by intratracheal injection. Alveolar macrophages recovered by BAL
from Mfge8–/– mice had significantly impaired collagen uptake (Fig-
ure 5, A and B). In a complementary approach to evaluating in vivo
uptake, we quantified the number of retained collagen particles
in lung sections after BAL was performed with 10 ml of fluid to
remove unbound collagen. There were significantly fewer retained
collagen particles in the lung after BAL in Mfge8–/– mice (Figure
5, C and D). We next evaluated whether fibroblasts used Mfge8
for collagen uptake in vitro. Mfge8–/– lung fibroblasts did not have
impaired collagen uptake in vitro, and the addition of rMfge8 did
not increase fibroblast collagen uptake (Figure 6, A and B).
Mfge8 deficiency does not affect extracellular collagen degradation. We
next evaluated whether proteolytic collagen degradation in the
extracellular matrix was affected by Mfge8 deficiency. We mea-
sured mRNA transcript levels of proteases with known roles in
lung injury and fibrosis 14 days after bleomycin administration
by gene expression array (22–24). There were no significant differ-
ences in levels of metalloproteinase (MMP); TIMP1; cathepsins B,
D, L, and K; or urokinase (Supplemental Figure 3A). We also found
no difference in the enzymatic activity of MMP2 and MMP9, two
proteases with important roles in lung fibrosis (25–27), by gela-
tin zymography of lung homogenates taken at 7, 14, 21, and 28
days after bleomycin injury (Supplemental Figure 3B and data not
shown). We also evaluated the production of MMP8 (neutrophil
collagenase) by Western blot on days 3 and 14 after bleomycin
treatment and found no difference between Mfge8–/– and control
mice (Supplemental Figure 3, C and D).
We were next interested in whether Mfge8 deficiency led to a
functional impairment in collagen-degrading capacity in the
lung. To answer this question, we evaluated the ability of both
alveolar macrophages and bleomycin-treated lung tissue lysates
from Mfge8–/– and control mice to degrade collagen. The FITC-
conjugated collagen molecule is supersaturated with fluores-
cein such that the close proximity of the fluorescein molecules
quenches the fluorescent signal. Proteolytic degradation of col-
lagen results in separation of the fluorescent dye and an increase
in the fluorescent signal, which can then be used to quantify col-
lagen cleavage. We therefore cultured alveolar macrophages from
Mfge8–/– and control mice for 30 minutes with FITC-conjugated
type I collagen and measured the fluorescent signal to quantify
collagen degradation. We found no difference in collagen degra-
dation between experimental groups (Supplemental Figure 3E).
We next evaluated whether lung homogenates had increased
collagen-degrading activity by incubating equal concentrations
of lung lysates taken from mice 14 and 21 days after bleomycin
treatment or saline treatment with FITC-conjugated collagen and
measuring degradation by the intensity of the fluorescent signal.
In both Mfge8–/– and control mice, lung lysates had increased col-
lagen-degrading activity after bleomycin treatment as compared
with saline treatment, but there was no significant difference
between groups (Supplemental Figure 3F).
Mfge8 mediates collagen uptake in vitro. (A) Primary alveolar macrophages were cultured for 30 minutes with FITC-conjugated type I collagen,
and uptake was evaluated by fluorescence microscopy (red arrows). Scale bar: 10 μm. (B) Ingestion of collagen was quantified as the collagen
uptake index (CUI: number of macrophages with ingestions divided by the total number of macrophages counted). The addition of unlabeled type I
collagen (25, 50, 150 μg/ml) inhibited uptake of FITC-conjugated collagen in a dose-dependent fashion (*P < 0.001, 1-way ANOVA with Bonfer-
roni t test for multiple comparisons; n = 3–4; data are expressed as percent control relative to wild-type uptake). (C) Alveolar macrophages from
Mfge8–/– mice had significantly impaired collagen uptake index as compared with Mfge8+/+ alveolar macrophages. (*P = 0.001, 1-way ANOVA
with Bonferroni t test for multiple comparisons; n = 3–5). Addition of rMfge8 (μg/ml) rescued collagen uptake in Mfge8–/– alveolar macrophages
(**P = 0.007, ***P = 0.023). (D) Addition of rMfge8 (μg/ml) increased collagen uptake in Mfge8+/+ alveolar macrophages under serum-starved
conditions (*P = 0.009, Student’s t test to compare indicated columns; n = 5–6). Data are presented as mean ± SEM.
? The?Journal?of?Clinical?Investigation http://www.jci.org Volume 119 Number 12 December 2009
Mfge8-mediated collagen internalization is independent of RGD-binding
integrins. Mfge8 contains an integrin-binding RGD sequence in its
second EGF-like domain that is bound by αvβ3 or αvβ5 integrins
(10, 28). These integrins are critical for Mfge8-mediated uptake of
apoptotic cells (10, 11). We first confirmed that the αvβ3 and αvβ5
integrins mediated cell adhesion to rMfge8, whereas other RGD-
binding integrins, α5β1, αvβ6, or αvβ8, did not (data not shown). We
then evaluated the roles of αvβ3 or αvβ5 integrins in Mfge8-mediated
collagen uptake. GRGDSP had no effect on collagen uptake in vitro
at concentrations up to 10 mM (Supplemental Figure 4A). Alveolar
macrophages taken from integrin β5–/–β3+/– or β5–/–β3–/– mice had
intact collagen uptake in vitro (Supplemental Figure 4B). β5–/–β3+/–
or β5–/–β3–/– mice did not have an enhanced fibrotic response to
bleomycin (Supplemental Figure 4C). These data indicate that
Mfge8-mediated collagen endocytosis, in contrast to uptake of
apoptotic cells, is independent of RGD binding integrins.
The first discoidin domain of Mfge8 is sufficient for collagen binding and
uptake. We next determined the domain of Mfge8 that mediates
collagen binding and uptake. Mfge8 contains 2 EGF-like domains
at its N terminus and 2 discoidin domains at its C terminus. We
created a series of Mfge8 constructs fused to the human Fc (huFc)
domain (Figure 7A). We used surface plasmon resonance to evalu-
ate Mfge8-collagen interaction and found a dose-dependent bind-
ing of collagen to immobilized full-length Mfge8 construct (Figure
7, B and C). While collagen bound a construct containing only the
first discoidin domain of Mfge8 (Dd1) similar to the full-length
construct (Figure 7D), collagen did not bind a construct lacking
both discoidin domains (Ndd) (Figure 7E). To prove that the first
discoidin domain could mediate collagen uptake, we evaluated
the ability of each construct to rescue in vitro collagen uptake in
Mfge8–/– alveolar macrophages. Dd1 significantly increased the
collagen uptake index, while the Ndd construct had no signifi-
cant effect (Figure 7F). These data indicate that the first discoidin
domain of Mfge8 is sufficient for collagen binding and uptake.
Since its initial description as a milk fat protein, Mfge8 has been
shown to have a number of roles, including the regulation of
apoptotic cell clearance, autoimmunity, neoangiogenesis, and
sperm-egg binding (9, 11, 16,
29). We have found a critical
role for Mfge8 in negatively reg-
ulating tissue fibrosis mediated
by a mechanism that is inde-
pendent of any of its previously
Several lines of evidence sup-
port a model by which Mfge8
binds collagen that has accu-
mulated in the extracellular
space and targets it for uptake
and turnover by macrophages.
Mfge8 binds collagen directly.
Mfge8–/– macrophages have
impaired collagen internaliza-
tion in vitro, and this defect is
rescued with the addition of
rMfge8. Mfge8–/– macrophages
have impaired collagen uptake
in vivo. After bleomycin-induced
lung fibrosis, Mfge8–/– mice have an increase in total lung collagen
content without a relative increase in collagen production.
The content of collagen in the extracellular matrix is precisely
regulated by the balance between collagen production and col-
lagen degradation both under homeostatic conditions and in
conditions of rapid matrix remodeling such as wound healing.
Tissue fibrosis occurs when collagen production outpaces colla-
gen degradation. Several of the molecular processes that result in
increased collagen production after injury have been extensively
characterized. In lung fibrosis, a paradigm has emerged whereby
fibroblasts migrate into damaged air spaces and under the influ-
ence of activated TGF-β differentiate into myofibroblasts (30).
Myofibroblasts (and fibroblasts) then deposit collagen, produc-
ing tissue fibrosis (20). Lysophosphatidic acid (LPA) signaling
through the LPA1 receptor recruits fibroblasts into the air spaces
(31), while TGF-β promotes both transformation of fibroblasts
into myofibroblasts and production of collagen by fibroblasts.
The in vivo relevance of these pathways is apparent in that mice
deficient in the LPA1 receptor or the αvβ6 integrin (a potent acti-
vator of TGF-β in the lung) are completely protected from bleo-
mycin-induced lung fibrosis (31, 32).
Surprisingly little is known about the normal homeostatic path-
ways responsible for collagen removal. In the extracellular matrix,
collagen fibers are cleaved and denatured by matrix MMPs (MMP1,
MMP13, and MMP14) and further degraded by the gelatinases
(MMP2, MMP9) (33–35). The intracellular pathway of collagen
degradation involves phagocytosis and lysosomal degradation of
collagen by cathepsins (6, 36, 37). MMP cleavage and fragmenta-
tion of collagen prior to internalization are also important for this
pathway (34), with one group reporting that a much lower rate
of internalization can occur despite MMP inhibition (8). A few of
the steps involved in collagen uptake by cells have been defined.
The α2β1 and α3β1 bind collagen, and inhibition of these integrins
reduces phagocytosis of collagen-coated beads (38). The uPARAP/
Endo180 receptor, a member of the mannose macrophage recep-
tor family, binds collagen directly, and fibroblasts deficient in this
molecule are unable to internalize collagen in vitro (8, 39). Inter-
estingly, in our model Mfge8 does not increase in vitro fibroblast
collagen uptake, suggesting that the uPARAP/Endo180 mecha-
Mfge8 mediates collagen uptake in vivo. (A) Alveolar macrophages obtained by BAL 30 minutes after
intratracheal injection of FITC-conjugated type I collagen were examined by fluorescence microscopy
for collagen uptake (red arrow). Scale bar: 10 μm. (B) Mfge8–/– alveolar macrophages had significantly
impaired collagen uptake in vivo (*P = 0.009, Student’s t test; n = 6). (C) Frozen sections taken from lungs
after intratracheal collagen injection were counterstained with DAPI, and the number of retained colla-
gen particles divided by the total number of nuclei in each section was quantified. Scale bar: 10 μm. (D)
Mfge8–/– lungs retained significantly fewer collagen particles (*P = 0.047 using a Student’s t test, n = 4).
Data are presented as mean ± SEM.
3718?The?Journal?of?Clinical?Investigation http://www.jci.org Volume 119 Number 12 December 2009
nism of collagen uptake may be dominant in fibroblasts. The in
vivo roles of fibroblast receptors that mediate collagen uptake in
regulating tissue fibrosis after injury are unknown.
The results presented here provide the first evidence to our
knowledge that an extracellular glycoprotein binds and targets
collagen for internalization and intracellular turnover. We have
also identified a functional role for the first discoidin domain of
Mfge8. The discoidin domains of the cell-surface receptors DDR1
and DDR2 also bind collagen (40, 41). However, binding of col-
lagen by these receptors leads to receptor phosphorylation and
downstream signaling rather than collagen ingestion (42). We
show what we believe to be the first example of a discoidin domain
targeting collagen for cellular uptake.
The second discoidin domain of Mfge8 binds phosphatidyl-
serine residues on apoptotic cells and bridges them to the αvβ3
or αvβ5 integrins on phagocytes (10, 28). We did not find a role
for either of these integrins in collagen phagocytosis. Wheth-
er uptake is triggered by a specific receptor or by interaction
between Mfge8 and the phagocytic cell surface is unclear. In
sperm-egg binding, a model has been proposed by which the first
discoidin domain binds sperm and the second domain binds the
oocyte, bridging the two together (29). Alternatively, Mfge8 may
form dimers through its EGF-like domains, with the discoidin
domains serving to bridge sperm to the oocyte (29) or in our case
collagen to the macrophage.
It is important to consider whether impaired apoptotic cell
clearance in Mfge8–/– mice contributed to the increase in fibrosis.
Our initial hypothesis was that impaired clearance of apoptotic
cells may directly induce fibrosis, and we had set out to investi-
gate the relationship between apoptosis and fibrosis (12). Based
on our work in the mammary gland and the work of others (9–11),
we expected to see a role for Mfge8 in apoptotic cell clearance
after injury in the lung. We could not, however, find evidence of
impaired in vivo apoptotic cell clearance in the lungs of bleomy-
cin-treated Mfge8–/– mice.
The lack of an absolute increase in apoptotic cell numbers does
not exclude the possibility that alternate pathways of clearance
could have induced pulmonary fibrosis while still removing apop-
totic cells in a timely manner. Several recent publications have
shown that Mfge8-independent apoptotic cell clearance induces
phagocyte release of IL-1β and IL-6 (43, 44), cytokines that induce
inflammation and fibrosis (45, 46). Conversely, Mfge8-dependent
apoptotic cell clearance has been shown to induce the release of
TGF-β, a profibrotic and antiinflammatory cytokine (32, 44). We
found no increase in tissue inflammation after bleomycin admin-
istration in Mfge8–/– mice, indicating that cytokine-induced inflam-
mation is not likely to be the explanation for increased fibrosis.
Furthermore, an increase in fibrosis due to excessive inflammation
or TGF-β release and activation would occur through increases in
myofibroblast accumulation and collagen production. Our find-
ings indicate that Mfge8 deficiency does not result in increased
myofibroblast accumulation or collagen production. Furthermore,
Mfge8-independent apoptotic cell clearance would be expected to
result in less, not more, TGF-β production, so the effects of loss of
Mfge8 on this pathway would be expected to protect against the
development of fibrosis.
Another important consideration is whether partially cleaved
but uncleared collagen fragments could have induced neutrophil-
ic inflammation by acting as chemoattractants in the lung (47,
48). Our data suggest that this did not occur in bleomycin-treated
Mfge8–/– mice, since we could not find an increase in lung neutro-
phils or inflammation-induced increases in collagen production.
The fact that wild-type mice make Mfge8 in response to bleo-
mycin and do develop fibrosis (albeit to a lesser extent than their
knockout counterparts) indicates that Mfge8-dependent collagen
clearance is a pathway that can be saturated with persistent col-
lagen production. A similar pattern is apparent in lung samples
of patients with IPF, all of which have increased expression of the
human ortholog of Mfge8, lactadherin (16). Whether pathologi-
cal fibrosis occurs simply due to overwhelming collagen produc-
tion or a combination of persistent collagen production coupled
with impaired collagen degradation has yet to be determined.
Interestingly, in some forms of pharmaceutically induced fibro-
sis, the phagocytic capacity of fibroblasts is severely reduced,
suggesting that collagen turnover pathways can be inhibited (49,
50). As we learn more about the pathways responsible for remod-
eling and removal of collagen from the extracellular matrix, we
can begin to dissect out the relative contribution of increased col-
lagen production and impaired collagen degradation to fibrotic
disease. The high prevalence and lack of treatment for fibrotic
diseases underscore the importance of a better understanding of
the endogenous pathways that mediate removal of collagen from
the extracellular matrix.
Reagents. The rabbit anti-Mfge8 monoclonal antibody (4F6) (9) was used
for immunohistochemistry, and anti-Mfge8 antibody from R&D was used
for Western blotting. Lactadherin antibody was purchased from R&D;
Crk antibody was purchased from BD Biosciences; MMP8 and myelo-
peroxidase heavy chain antibodies were purchased from Santa Cruz Bio-
technology Inc.; anti-Gr1 antibody was purchased from eBioscience; and
α–smooth muscle antibody was purchased from Sigma-Aldrich. FITC–
Mfge8–/– fibroblasts do not have impaired collagen uptake. (A and
B) Primary lung fibroblasts from Mfge8–/– and Mfge8+/+ mice at pas-
sage 4 were incubated for 90 minutes with FITC-conjugated type l
collagen, and unbound/uningested collagen was removed with mul-
tiple washes. Cells were then incubated with 50 μg/ml trypsin and
50 μg/ml proteinase K to remove membrane-bound collagen. Collagen
in the membrane-bound portion (supernatant after enzymatic treat-
ment) and intracellular portion (pellet remaining after enzymatic treat-
ment) was quantified by a spectrofluorometer. There was no difference
in membrane-bound (A) or intracellular (B) collagen with or without the
addition of rMfge8 (13 μg/ml). Data are presented as mean ± SEM.
?The?Journal?of?Clinical?Investigation http://www.jci.org Volume 119 Number 12 December 2009
collagen type I was purchased from Invitrogen and rat tail type I collagen
from Sigma-Aldrich. rMfge8 was purchased from R&D. GRGDSP and
GRGESP were purchased from Anaspec.
Mfge8-deficient mice. Mice functionally deficient in Mfge8 were created
using a gene trap vector and have been previously characterized (9, 16).
Initial studies were conducted in a mixed-strain background (C57BL/6
× 129/Ola Mfge8+/– littermates used as controls), and subsequent studies
were conducted in mice backcrossed 6 or 10 generations into the C57BL/6
background (Mfge8+/+ used as controls) or 10 generations into the 129/SvEv
background. Mice in the 129/SvEv background were used only in studies
in Supplemental Figure 4. All experimental protocols were approved by the
UCSF IACUC for animal studies.
Tissue sample accrual and processing. Written informed consent was obtained
from all subjects, and the study was approved by the UCSF Committee on
Human Research. IPF lung tissues were obtained at the time of diagnostic
lung biopsy. IPF patients underwent a history, physical examination, high-res-
olution computed tomography, pulmonary function testing, and diagnostic
lung biopsy. In all cases, the pathologic diagnosis was usual interstitial pneu-
monia (UIP), and the consensus clinical diagnosis was IPF. Normal human
lung tissue was obtained from lungs not used by the Northern California
Transplant Donor Network. After harvest, lung tissue was directly snap-fro-
zen in liquid nitrogen. Samples were stored at –80°C until use for experi-
ments. For immunoblotting, frozen lung tissue was pulverized in a stainless
steel tissue pulverizer (Fisher Scientific), pre-cooled in liquid nitrogen, then
immediately lysed in RIPA buffer prior to separation by SDS-PAGE.
Immunohistochemistry. Tissues were fixed with 4% paraformaldehyde
or zinc-based formalin (Z-FIX, ANATECH) and embedded in paraffin,
and 5-μm sections were treated with 0.2% trypsin for 15 minutes for
antigen retrieval. After blocking for 1 hour with 1% BSA/5% goat serum,
sections were incubated at room temperature for 1–2 hours with pri-
mary antibody, followed by a biotinylated secondary antibody (Vector)
for 45 minutes, followed by ABC reagent (Vector) for 30 minutes and
liquid diaminobenzidine (Sigma-Aldrich). Sections were counterstained
with hematoxylin or methyl green. α–Smooth muscle actin–positive
cells taken from 25 randomly selected high-power fields (×200) were
quantified. For frozen sections, lungs were inflated with 1.5 ml of OCT
medium after BAL was performed, and 5-μm sections were air dried,
washed in PBS, and fixed in acetone. After 1 hour of blocking with 5%
rat serum, anti-Gr1 antibody was added for 60 minutes and washed
off; Alexa Fluor 594 anti-rat (Invitrogen) was added for 30 minutes
and washed off; and sections were coverslipped with anti-fade solution
containing DAPI (Vector). The number of Gr1-positive cells per 10 ran-
domly selected ×200 fields was quantified by an investigator blinded to
the genotype of tissue sections.
Bleomycin model of pulmonary fibrosis. Eight- to- 10-week-old sex-matched
mice were anesthetized, and bleomycin (1.1 U/kg Blenoxane) was instilled
directly into the trachea after cut-down. Twenty-eight (C57BL/6 or mixed-
strain) or 56 (129/SvEv strain) days after treatment, lungs were removed,
homogenized, precipitated with trichloroacetic acid, and baked overnight
at 110°C in HCl. Samples were reconstituted with 2 ml of water, and
hydroxyproline content was measured using a colorimetric chloramine T
assay. Some lungs were inflated with 4% PFA to a pressure of 25 cm H2O,
processed, and embedded in paraffin, and then 5-μm sections were stained
with picrosirius red for evaluation of fibrosis.
The first discoidin domain of Mfge8 is sufficient for collagen binding and uptake. (A) Constructs containing full-length (Fl) Mfge8, Mfge8 lacking
the terminal discoidin domain (Dd1), or both discoidin domains (Ndd) fused to a huFc domain were immobilized on a Biacore CM5 chip, and
binding to increasing doses of collagen was evaluated. P/T represents a domain present in the long isoform of Mfge8 that is rich in proline and
threonine. (B) Collagen bound full-length Mfge8 in a dose-dependent fashion with a Kd of 733 nM. (C) Flow plot demonstrating dose-dependent
binding of collagen to immobilized full-length construct. Black line, 16 nM; filled squares, 31 nM; open diamonds, 63 nM; filled circles, 125 nM;
open triangles, 250 nM; filled diamonds, 500 nM. (D) Flow plot demonstrating dose-dependent binding of collagen to immobilized Dd1 construct.
(E) Flow plot demonstrating no binding of collagen to the Ndd construct. (F) The ability of constructs (13 μg/ml) to rescue the defect in alveolar
macrophage collagen uptake was evaluated in vitro. Dd1 construct rescued Mfge8–/– alveolar macrophage collagen uptake, while the Ndd con-
struct had no significant effect (*P = 0.01, 1-way ANOVA with Bonferroni t test; n = 3–4). Data are presented as mean ± SEM.
3720? The?Journal?of?Clinical?Investigation http://www.jci.org Volume 119 Number 12 December 2009
Evaluation of collagen production with deuterated water. Eight- to-10 week-old
sex-matched mice were injected with a single bolus i.p. dose of 100% deuter-
ated water (Isotec, Sigma-Aldrich) with 0.9% NaCl, an isosmotic solution
delivered at 35 μl/g body mass, at the same time as they received bleomy-
cin. Subsequently, normal drinking water was replaced with 8% deuterated
water to maintain the animal’s total body water at 5% enrichment of deu-
terated water for a period of 14 days. Mice were then euthanized, and their
lungs removed and processed by mass spectrometry analysis.
Preparation of sample for mass spectrometric analysis. The derivatization of
hydroxyproline for GC/MS analysis has been described previously (19).
Briefly, tissue was homogenized with a bead mill in normal abundance
(non-deuterated) water, and the homogenate was subjected to 2 rounds
of acetone precipitation at –20°C in order to obtain the total tissue
protein for hydroxyproline assessment. The proteins were hydrolyzed
by incubation in 6N HCl, dried under vacuum, and then suspended in
a solution of 50% acetonitrile, 50 mM K2HPO4, and pentafluorobenzyl
bromide before incubation. Derivatives were extracted into ethyl acetate,
and the top layer was removed and dried by vacuum centrifugation. In
order to acetylate the hydroxyl moiety of hydroxyproline, we incubated
samples with a solution of acetonitrile, N-methyl-N-[tert-butyldimethyl-si
lyl]trifluoroacetamide, and methylimidazole. This material was extracted
in petroleum ether and dried with Na2SO4.
GC/MS analysis of derivatized hydroxyproline. Analysis of the derivatized
hydroxyproline was performed on a standard quadrupole GC/MS instrument
(Agilent 5973/6980) in negative chemical ionization mode (NCI-GC/MS),
with helium as carrier and methane as reagent gas. The column used was
a DB17 ZB-50 column (J&W Scientific, Agilent). Selected ion monitoring
was performed on ions with mass-to-charge ratios (m/z) of 424 and 425,
which will include all of the carbon-hydrogen bonds from hydroxyproline.
The mole fraction of the M1 mass isotopomer was calculated as the ratio
of peak areas: M1/(M0 + M1). Incorporation of 2H into hydroxyproline
was calculated as the excess mole fraction of M1 (EM1) above the M1 mole
fraction in unlabeled standards at the same abundance. Fractional turn-
over (f), the fraction of newly synthesized hydroxyproline, was calculated
as EM1/EM1*, where EM1* represents the EM1 of a newly synthesized
molecule. EM1* was calculated from measured body water 2H2O enrich-
ments as previously described (19).
Analysis of body 2H2O enrichments in body water. 2H2O enrichment in plas-
ma samples was measured as previously described (51) using a Series 3000
Cycloidal mass spectrometer (Monitor Instruments).
Evaluation of phagocytic index in vivo. Mice were treated with bleomycin,
and at indicated time points alveolar macrophages were obtained by BAL.
Cytospin preparations were stained with Diff-Quick (Fisher Scientific)
and the number of ingestions per macrophage quantified. The phagocytic
index represents the number of ingestions per macrophages counted. A
minimum of 300 macrophages were counted, and the investigator was
blinded to the genotype of each sample.
BAL. Mice were euthanized, the trachea cannulated, and serial 0.9-ml
lavage was performed with ice cold PBS with 1 mM EDTA for a total of
4.5 ml for evaluation of inflammation after bleomycin treatment. Cells
were treated with rbc lysis buffer, after which a cell count was obtained
by hemocytometer. Cytospin slides were prepared and stained with Diff-
Quick reagent and the percentage of inflammatory cell types determined
using a light microscope (Supplemental Figure 1, A and B).
TUNEL assay. Five-micrometer sections taken from mice treated with bleo-
mycin were stained with TUNEL assay (ApopTag, Chemicon), and the num-
ber of apoptotic cells taken from 15 randomly selected high-power fields
(×200) was quantified. For determination of the number of free apoptotic
cells recovered by BAL, cytospin preparations were stained with TUNEL
assay and the number and proportion of apoptotic cells quantified.
Expression arrays/real-time PCR. Seven and 14 days after treatment with
bleomycin (1.1 U/kg), mice were sacrificed, and lungs were removed and
RNA extracted using a QIAGEN RNeasy Midi kit following the manufac-
turer’s instructions. After confirming acceptable RNA quality with Agi-
lent nanotechnology, expression arrays were done using an Agilent Mouse
One-Color 4×44 K array platform by the UCSF Functional Genomics Core
Facility. Complementary DNA was generated from total RNA using a first-
strand cDNA synthesis kit (Invitrogen). Real-time PCR was performed
using SYBR Green PCR Master Mix (Invitrogen) and results analyzed on
an AB Prism 7700 analyzer (Applied Biosystems). Real-time PCR values
were normalized to β-actin RNA and expressed as fold increase in mRNA
above saline-treated controls.
Collagen uptake assay. Freshly isolated alveolar macrophages were cul-
tured for 60 minutes on glass inserts and RPMI with 0.1% BSA in either
10% mouse serum from the same genotype (Figure 4, B and C, Figure 7F,
and Supplemental Figure 4, A and B) or under serum-free conditions
(Figure 4D). FITC-conjugated type I collagen (50 μg/ml) was added for
30 minutes at 37°C. After 30 minutes, inserts were washed several times
to remove unbound/uningested collagen, counterstained with DAPI,
and mounted on slides. Slides were examined with fluorescence micros-
copy using a Leica DM 5000B camera with Spot 4.5 acquisition software;
images were obtained at ×200 magnification, and a minimum of 500 cells
were analyzed for evidence of collagen binding/ingestion. Investigators
were blinded to the experimental conditions when quantifying collagen
uptake. The collagen uptake index represented the number of ingestions
per macrophages counted and was expressed as percentage of wild-type
uptake. Absolute uptake numbers ranged from 1.5% to 4% for wild-type
macrophages. Only cells that visually had collagen surrounded by a rim of
cytoplasm were considered to have uptake. The software program Merge 2
(Venning Graphic Utilities) was used to merge images of FITC (represent-
ing collagen) and phase contrast (to see cytoplasm) to determine whether
collagen was internalized. In preliminary studies, cells that were visually
considered as having uptake were further analyzed with Z-plane stacked
imaging using a Leica CTR 6000 camera and Image-Pro 5.1 (Media Cyber-
netics) acquisition software to determine whether collagen was ingested
or bound. For collagen to be considered ingested, the maximum fluores-
cence signal of the FITC (representing collagen) and DAPI (representing
macrophage nuclei) had to be present at the same level, and there had
to be a clear rim of cytoplasm between the FITC signal and the outside
of the cell. Using these criteria, 70% of cells (of 100 counted) that were
considered to have uptake by visual analysis had uptake according to
Z-plane analysis. Therefore, the in vitro collagen uptake index represents
both ingested (70%) and bound (30%) collagen.
For all inhibitor studies and studies with recombinant protein, com-
pounds were added to macrophages 30 minutes prior to initiation of
the uptake assay.
For the in vivo uptake assay, 60 μg of collagen in 60 μl H2O was placed
intratracheally; 30 minutes later, alveolar macrophages were recov-
ered by BAL and cytospin preparations made with DAPI counterstain,
and the proportion of macrophages with ingestions was quantified by
fluorescence microscopy and expressed as percent control of wild-type
macrophage uptake. The actual percentage of wild-type macrophage
uptake in vivo was 39. Mouse lungs were further lavaged with 10 ml
PBS to remove residual collagen and then inflated with 1.5 cc of OCT
medium and prepared for frozen sectioning. Five-micrometer sections
were counterstained with DAPI, and the number of collagen particles
per total number of nuclei counted in the section was quantified from
5 randomly selected high-power fields (×200) and expressed as percent
control of wild-type. The actual proportion of retained nuclei in wild-
type sections was 15%.
? The?Journal?of?Clinical?Investigation http://www.jci.org Volume 119 Number 12 December 2009
Fibroblast collagen uptake assay. For fibroblast uptake assays, lung fibro-
blasts were isolated by digesting whole lung with Blendzyme 3 (Roche).
After 4 passages in vitro, fibroblasts were plated at a concentration of
50,000 cells per 24-well plate overnight in 10% FCS DMEM. Fibroblasts
were then washed 3 times with PBS and incubated in media containing
10% mouse serum of the same genotype as the fibroblasts. FITC-conju-
gated type I collagen (50 μg/ml) was added for 90 minutes, after which the
cells were washed vigorously with PBS. Fibroblasts were then incubated
with 50 μg/ml trypsin and 50 μg/ml proteinase K at 37°C for 5 minutes,
after which they were removed and centrifuged, and the supernatant con-
taining the membrane-bound collagen cleaved by proteolytic treatment
was separated. The remaining pellet was lysed with 0.1 M NaOH to release
intracellular collagen, and fluorescence was measured in both the super-
natant and pellet using a spectrofluorometer (Tecan).
Collagen degradation assays. For the in vitro assay, isolated alveolar macro-
phages were cultured in black 96-well black plates in triplicate. Cells were
plated at a concentration of 50,000 cells per well in PBS with 1 μg/ml of
FITC-conjugated type I collagen (Invitrogen) in a total volume of 100 μl.
The FITC-conjugated collagen from Invitrogen is designed for enzymatic
assays and is supersaturated with FITC. When the collagen is cleaved, the
fluorescent signal increases. After 30 minutes, fluorescence was quantified
using a spectrofluorometer (Tecan). The fluorescent signal from control
wells containing only collagen was subtracted from that from wells con-
taining macrophages and collagen.
For the assay using in vivo samples, 10 μg of total lung homogenates
from saline- and bleomycin-treated mice were incubated at 37°C with
FITC-conjugated type I collagen (10 μg/ml) in duplicate. After 60 minutes,
fluorescence was quantified using a spectrofluorometer (Tecan). The fluo-
rescent signal from control wells containing only collagen was subtracted
from that from wells containing tissue lysates and collagen.
Mfge8 constructs. RNA taken from the involuting mouse mammary
gland was extracted using TRIzol following the manufacturer’s instruc-
tions (Invitrogen). Full-length Mfge8 was cloned into the pMIB/V5-His
vector (Invitrogen) using the sequences 5′-GGCATGCTAAGCTTGTCTG-
GTGACTTCTGTGACTCCAGCCTGTGC-3′ (Mfge8 forward primer) and
(Mfge8 reverse primer). The huFc domain was cloned into the vector using
the sequences 5′-GGCGGCACTAGTGCACCTGAACTCCTGGGGGGACC-
GTC-3′ (Fc forward primer) and 5′-TATCTGCAGAATTCTCATTTACCC-
GGAGACAGGGAGAGGCTC-3′ (Fc reverse primer). For the Dd1 and Ndd
construct, the primers 5′-GGCATGCTAAGCTTGTCTGGTGACTTCT-
GTGACTCCAGCCTGTGC-3′ (Mfge8 forward primer), 5′-GGCGGCAC-
TAGTTCTGCCTTCGATGCCCAGGAGCTCGAAGCG-3′ (Dd1 reverse
primer), and 5′-GGCGGCACTAGTTCTGCCTTCGATGGAGGCTAG-
GTTGTTGGA-3′ (Ndd reverse primer) were used to obtain cDNA that was
then inserted into the pMIB/V5-His vector containing huFc after enzy-
matic digestion and removal of the full-length construct. High Five cells
were transfected with each vector using Cellfectin reagent (Invitrogen) and
recombinant protein isolated by binding on a protein G column and elut-
ing with a pH gradient.
Surface plasmon resonance. Proteins were immobilized on different flow cells
of a Biacore CM5 chip and analyzed on a Biacore T100. The dextran surface
on the chip was first activated by injection of a 1:1 mixture of N-hydroxy-
succinimide and N-ethyl-NP-[(3-dimethylamino)-propyl]-carbodiimide
hydrochloride (GE Healthcare). Proteins were then diluted to 20 μg/ml
in 10 mM sodium acetate at pH 4.5 and injected to equal mass density of
3,000 response units (RU) on the surface. The remaining active sites on the
surface were then blocked with 1 M ethanolamine-HCl, and washed with
HEPES-buffered saline with 0.05% P-20. Type I collagen was diluted into
running buffer (PBS with 0.05% Tween-20) and injected at different con-
centrations at a flow rate of 30 μl/min in duplicate with intermittent blank
injections of running buffer alone. The surfaces were regenerated after each
injection to remove residual bound collagen by injection of 2 M NaCl and
washing with running buffer. The response for each flow cell was represent-
ed in reference to a flow cell with EGFR1-Fc immobilized and subtracted by
the average response of the blank injections over each surface.
Evaluation of vascular permeability. Five days after treatment with intratra-
cheal bleomycin (5 U/kg), mice were administered 0.5 μCi of [125I]albumin
by i.p. injection. Four hours later, they were sacrificed and their lungs
removed and homogenized. A blood sample was used to calculate the
hematocrit. Vascular permeability was expressed as extravascular plasma
equivalents (EVPE), the ratio of radioactive counts in the lung (after sub-
traction of counts attributable to the blood content within the lung) to
counts in the plasma (52).
Statistics. Paired data columns were evaluated using Student’s t test with
Microsoft Office Excel 2007. One-way ANOVA was used for comparison
of multiple data columns using SigmaStat 3.11 (SYSTAT), and when dif-
ferences were statistically significant, this was followed with a Bonferroni
t test for subsequent pairwise analysis. Data were tested for normality and
variance, and a P value less than 0.05 was considered significant.
This work was supported by NIH grants HL64353 HL53949,
HL083950, AI024674, and NIH Program in Genomics Applica-
tions grant (BayGenomics) HL66600 (to D. Sheppard), NIH grant
AI053194 (to Z. Werb), NIH K08 mentored award HL083985
(to K. Atabai), and an American Lung Association Biomedical
Research Grant (to K. Atabai). We thank Robert Fletterick (NIH
grant S10 RR023443) and the Sandler Foundation for providing
access to and assistance with measurement of surface plasmon
resonance and Mark Looney and Michael Matthay for help with
vascular permeability studies.
Received for publication June 1, 2009, and accepted in revised form
September 9, 2009.
Address correspondence to: Dean Sheppard, Lung Biology Center,
box 2922, UCSF, San Francisco, California 94143, USA. Phone:
(415) 514-4269; Fax: (415) 514-4278; E-mail: Dean.Sheppard@
1. Raghu, G., Striker, L.J., Hudson, L.D., and Strik-
er, G.E. 1985. Extracellular matrix in normal
and fibrotic human lungs. Am. Rev. Respir. Dis.
2. Chua, F., Gauldie, J., and Laurent, G.J. 2005. Pul-
monary fibrosis: searching for model answers.
Am. J. Respir. Cell Mol. Biol. 33:9–13.
3. Hammar, S.P., Winterbauer, R.H., Bockus, D.,
Remington, F., and Friedman, S. 1985. Idiopathic
fibrosing alveolitis: a review with emphasis on ultra-
structural and immunohistochemical features.
Ultrastruct. Pathol. 9:345–372.
4. Wynn, T.A. 2008. Cellular and molecular mecha-
nisms of fibrosis. J. Pathol. 214:199–210.
5. Gross, T.J., and Hunninghake, G.W. 2001.
Idiopathic pulmonary fibrosis. N. Engl. J. Med.
6. Everts, V., van der Zee, E., Creemers, L., and Beertsen,
W. 1996. Phagocytosis and intracellular digestion
of collagen, its role in turnover and remodelling.
Histochem. J. 28:229–245.
7. Song, F., Wisithphrom, K., Zhou, J., and Windsor,
L.J. 2006. Matrix metalloproteinase dependent and
independent collagen degradation. Front. Biosci.
8. Madsen, D.H., et al. 2007. Extracellular collagenases
and the endocytic receptor, urokinase plasminogen
activator receptor-associated protein/Endo180,
cooperate in fibroblast-mediated collagen degra-
dation. J. Biol. Chem. 282:27037–27045.
9. Atabai, K., et al. 2005. Mfge8 is critical for mam-
mary gland remodeling during involution. Mol.
Biol. Cell. 16:5528–5537.
10. Hanayama, R., et al. 2002. Identification of a fac-
tor that links apoptotic cells to phagocytes. Nature.
3722? The?Journal?of?Clinical?Investigation http://www.jci.org Volume 119 Number 12 December 2009
11. Hanayama, R., et al. 2004. Autoimmune disease
and impaired uptake of apoptotic cells in MFG-
E8-deficient mice. Science. 304:1147–1150.
12. Teder, P., et al. 2002. Resolution of lung inflamma-
tion by CD44. Science. 296:155–158.
13. Kuwano, K., et al. 1999. Essential roles of the Fas-
Fas ligand pathway in the development of pulmo-
nary fibrosis. J. Clin. Invest. 104:13–19.
14. Hay, J., Shahzeidi, S., and Laurent, G. 1991. Mecha-
nisms of bleomycin-induced lung damage. Arch.
15. Veron, P., Segura, E., Sugano, G., Amigorena, S.,
and Thery, C. 2005. Accumulation of MFG-E8/lac-
tadherin on exosomes from immature dendritic
cells. Blood Cells Mol. Dis. 35:81–88.
16. Silvestre, J.S., et al. 2005. Lactadherin promotes
VEGF-dependent neovascularization. Nat. Med.
17. Kaminski, N., et al. 2000. Global analysis of gene
expression in pulmonary fibrosis reveals distinct pro-
grams regulating lung inflammation and fibrosis.
Proc. Natl. Acad. Sci. U. S. A. 97:1778–1783.
18. Shahzeidi, S., Jeffery, P.K., Laurent, G.J., and
McAnulty, R.J. 1994. Increased type I procollagen
mRNA transcripts in the lungs of mice during the
development of bleomycin-induced fibrosis. Eur.
Respir. J. 7:1938–1943.
19. Gardner, J.L., et al. 2007. Measurement of liver
collagen synthesis by heavy water labeling: effects
of profibrotic toxicants and antifibrotic inter-
ventions. Am. J. Physiol. Gastrointest. Liver Physiol.
20. Zhang, K., Rekhter, M.D., Gordon, D., and Phan,
S.H. 1994. Myofibroblasts and their role in lung
collagen gene expression during pulmonary fibro-
sis. A combined immunohistochemical and in situ
hybridization study. Am. J. Pathol. 145:114–125.
21. Vogel, W., Gish, G.D., Alves, F., and Pawson, T. 1997.
The discoidin domain receptor tyrosine kinases are
activated by collagen. Mol. Cell. 1:13–23.
22. Kimura, M., et al. 2005. The significance of cathep-
sins, thrombin and aminopeptidase in diffuse
interstitial lung diseases. J. Med. Invest. 52:93–100.
23. Sisson, T.H., and Simon, R.H. 2007. The plasmino-
gen activation system in lung disease. Curr. Drug
24. Winkler, M.K., and Fowlkes, J.L. 2002. Metallopro-
teinase and growth factor interactions: do they play
a role in pulmonary fibrosis? Am. J. Physiol. Lung Cell
Mol. Physiol. 283:L1–L11.
25. Garcia-Alvarez, J., et al. 2006. Membrane type-
matrix metalloproteinases in idiopathic pulmonary
fibrosis. Sarcoidosis Vasc. Diffuse Lung Dis. 23:13–21.
26. Cabrera, S., et al. 2007. Overexpression of MMP9
in macrophages attenuates pulmonary fibrosis
induced by bleomycin. Int. J. Biochem. Cell Biol.
27. Betsuyaku, T., Fukuda, Y., Parks, W.C., Shipley, J.M.,
and Senior, R.M. 2000. Gelatinase B is required for
alveolar bronchiolization after intratracheal bleo-
mycin. Am. J. Pathol. 157:525–535.
28. Nandrot, E.F., et al. 2007. Essential role for MFG-
E8 as ligand for alphavbeta5 integrin in diurnal
retinal phagocytosis. Proc. Natl. Acad. Sci. U. S. A.
29. Ensslin, M.A., and Shur, B.D. 2003. Identification
of mouse sperm SED1, a bimotif EGF repeat and
discoidin-domain protein involved in sperm-egg
binding. Cell. 114:405–417.
30. Fine, A., and Goldstein, R.H. 1987. The effect of
transforming growth factor-beta on cell prolifera-
tion and collagen formation by lung fibroblasts.
J. Biol. Chem. 262:3897–3902.
31. Tager, A.M., et al. 2008. The lysophosphatidic acid
receptor LPA1 links pulmonary fibrosis to lung
injury by mediating fibroblast recruitment and
vascular leak. Nat. Med. 14:45–54.
32. Munger, J.S., et al. 1999. The integrin alpha v beta 6
binds and activates latent TGF beta 1: a mechanism
for regulating pulmonary inflammation and fibrosis.
33. Egeblad, M., and Werb, Z. 2002. New functions for
the matrix metalloproteinases in cancer progression.
Nat. Rev. Cancer. 2:161–174.
34. Lee, H., Overall, C.M., McCulloch, C.A., and Sodek,
J. 2006. A critical role for the membrane-type 1
matrix metalloproteinase in collagen phagocytosis.
Mol. Biol. Cell. 17:4812–4826.
35. Visse, R., and Nagase, H. 2003. Matrix metallopro-
teinases and tissue inhibitors of metalloproteinas-
es: structure, function, and biochemistry. Circ. Res.
36. Dyer, R.F., and Peppler, R.D. 1977. Intracellular col-
lagen in the nonpregnant and IUD-containing rat
uterus. Anat. Rec. 187:241–247.
37. Melcher, A.H., and Chan, J. 1981. Phagocytosis and
digestion of collagen by gingival fibroblasts in vivo:
a study of serial sections. J. Ultrastruct. Res. 77:1–36.
38. Lee, W., Sodek, J., and McCulloch, C.A. 1996. Role
of integrins in regulation of collagen phagocytosis
by human fibroblasts. J. Cell. Physiol. 168:695–704.
39. Behrendt, N. 2004. The urokinase receptor (uPAR)
and the uPAR-associated protein (uPARAP/
Endo180): membrane proteins engaged in matrix
turnover during tissue remodeling. Biol. Chem.
40. Leitinger, B. 2003. Molecular analysis of collagen
binding by the human discoidin domain receptors,
DDR1 and DDR2. Identification of collagen bind-
ing sites in DDR2. J. Biol. Chem. 278:16761–16769.
41. Curat, C.A., Eck, M., Dervillez, X., and Vogel, W.F.
2001. Mapping of epitopes in discoidin domain
receptor 1 critical for collagen binding. J. Biol. Chem.
42. Vogel, W., et al. 2000. Discoidin domain receptor
1 is activated independently of beta(1) integrin.
J. Biol. Chem. 275:5779–5784.
43. Jinushi, M., et al. 2008. Milk fat globule EGF-8 promotes
melanoma progression through coordinated Akt
and twist signaling in the tumor microenvironment.
Cancer Res. 68:8889–8898.
44. Jinushi, M., et al. 2007. MFG-E8-mediated uptake
of apoptotic cells by APCs links the pro- and anti-
inflammatory activities of GM-CSF. J. Clin. Invest.
45. Gasse, P., et al. 2007. IL-1R1/MyD88 signaling
and the inflammasome are essential in pulmonary
inflammation and fibrosis in mice. J. Clin. Invest.
46. Saito, F., et al. 2008. Role of interleukin-6 in bleo-
mycin-induced lung inflammatory changes in
mice. Am. J. Respir. Cell Mol. Biol. 38:566–571.
47. Riley, D.J., et al. 1988. Neutrophil response follow-
ing intratracheal instillation of collagen peptides
into rat lungs. Exp. Lung Res. 14:549–563.
48. Weathington, N.M., et al. 2006. A novel peptide
CXCR ligand derived from extracellular matrix
degradation during airway inflammation. Nat.
49. Kataoka, M., Kido, J., Shinohara, Y., and Nagata,
T. 2005. Drug-induced gingival overgrowth — a
review. Biol. Pharm. Bull. 28:1817–1821.
50. McGaw, W.T., and Porter, H. 1988. Cyclosporine-
induced gingival overgrowth: an ultrastructural
stereologic study. Oral Surg. Oral Med. Oral Pathol.
51. Busch, R., et al. 2006. Measurement of protein
turnover rates by heavy water labeling of non-
essential amino acids. Biochim. Biophys. Acta.
52. Su, X., et al. 2007. Activation of the alpha7 nAChR
reduces acid-induced acute lung injury in mice and rats.
Am. J. Respir. Cell Mol. Biol. 37:186–192.