GASTROENTEROLOGY 2007;132:1937–1946
Gene Expression Profiles During Hepatic Stellate Cell Activation in
Culture and In Vivo
SAMUELE DE MINICIS,* EKIHIRO SEKI,* HIROSHI UCHINAMI,* JOHANNES KLUWE,* YONGHUI ZHANG,‡
DAVID A. BRENNER,* and ROBERT F. SCHWABE*
*Department of Medicine and ‡Institute for Cancer Genetics, Columbia University, College of Physicians and Surgeons, New York, New York
See editorial on page 2059.
Background & Aims: Following hepatic injury, he-
patic stellate cells (HSCs) transdifferentiate to be-
come extracellular matrix-producing myofibro-
blasts and to promote hepatic fibrogenesis. In this
study, we determine gene expression changes in 3
different models of HSC activation and investigate
whether HSC culture activation reproduces gene
expression changes of HSC in vivo activation.
Methods: HSCs were isolated by density centrifu-
gation and magnetic antibody cell sorting from
normal mice, CCl4-treated mice, and mice that un-
derwent bile duct ligation (BDL). Gene expression
was analyzed by microarray and confirmed by poly-
merase chain reaction and Western blot analysis.
Results: Two thousand seventy-three probe sets
were differentially expressed in at least 1 of 3 mod-
els of HSC activation, including novel genes that
encode proinflammatory and antiapoptotic media-
tors; transcription factors; cell surface receptors;
and cytoskeleton components such as CXCL14, sur-
vivin, septin 4, osteopontin, PRX1, LMCD1, GPR91,
leiomodin, and anillin. BDL- and CCl4-activated
HSCs showed highly correlated gene expression
patterns, whereas culture activation only partially
reproduced the gene expression changes observed
during BDL- and CCl4-induced activation. Cocul-
ture with Kupffer cells or lipopolysaccharide treat-
ment during culture activation shifted the expres-
sion of most examined genes toward the pattern
observed during in vivo activation, suggesting a
role for these factors in the microenvironment that
drives HSC activation. Conclusions: The almost
identical HSC gene expression patterns after BDL
or CCl4 treatment indicate that HSCs exert similar
functions in different types of liver injury. Because
culture activation does not properly regulate gene
expression in HSCs, in vivo activation should be
considered the gold standard for the study of HSC
biology.
H epatic stellate cells (HSCs) constitute approxi-
mately 8%–14% of cells in the normal liver and are
the primary site for retinoid storage in the body.1,2 Fol-
lowing liver injury, HSCs transdifferentiate from lipo-
cyte-like cells into contractile and highly proliferative
myofibroblast-like cells.1,2 Activation of HSCs is consid-
ered a crucial event that promotes increased extracellular
matrix (ECM) production and hepatic fibrosis.1,2 Cur-
rently, the molecular signals that are activated during
HSC activation are only incompletely understood. Trans-
forming growth factor ␤ (TGF-␤) and platelet-derived
growth factor (PDGF) are key cytokines that drive HSC
activation and proliferation following hepatocellular in-
jury, but other mediators including angiotensin II, leptin,
endothelin, insulin, insulin-like growth factor, and lipo-
polysaccharide (LPS) as well as factors released by
PANCREAS, AND
BILIARY TRACT
Kupffer cells additionally contribute to HSC activa-
BASIC–LIVER,
tion.3–13 The majority of studies investigate the HSC
activation process by culturing purified HSCs on plastic
surfaces of cell culture dishes as a surrogate for the
molecular events that occur during HSC activation in the
injured liver. Culture activation of HSC up-regulates ac-
tivation markers such as ␣-smooth muscle actin (␣-SMA)
and collagen ␣1(I) and is accompanied by a loss of reti-
noids. However, it is not known whether culture activa-
tion of HSCs can reproduce the underlying changes in
gene expression of in vivo HSC activation and whether it
represents a suitable model to study HSC activation.
Moreover, it remains elusive whether all types of liver
injury lead to the activation of a conserved and largely
identical genetic program in HSCs or whether HSC acti-
vation results in a stimulus-specific activation of HSCs
with certain gene expression patterns during one type of
hepatic injury and different gene expression patterns in
other types of hepatic injury. To answer these questions,
we performed microarray studies and compared gene
Abbreviations used in this paper: ␣-SMA, ␣-smooth muscle actin;
BDL, bile duct ligation; ECM, extracellular matrix; HSC, hepatic stellate
cells; IAP, inhibitor of apoptosis protein; LPS, lipopolysaccharide;
PDGF, platelet-derived growth factor.
© 2007 by the AGA Institute
0016-5085/07/$32.00
doi:10.1053/j.gastro.2007.02.033
 1938 DE MINICIS ET AL
expression profiles between 2 different models of HSC in
vivo activation and HSC culture activation.
Materials and Methods
Cell Isolation, Purification, and Culture
Mouse HSCs were isolated from normal and fi-
brotic liver by collagenase-pronase perfusion and subse-
quent density centrifugation on Nycodenz gradients as
described previously.3,14 For each isolation, HSCs from 3
male Balb/c mice were loaded onto 1 Nycodenz gradi-
ent.3,14 Kupffer cells were depleted by magnetic antibody
sorting (MACS; Miltenyi Biotech, Auburn, CA) using
F4/80 (eBioscience, San Diego, CA) and CD11b (Pharm-
ingen, San Diego, CA) antibodies. Purity was tested by
retinoid autofluorescence and exceeded 95% in all isola-
tions. For in vivo HSC activation, mice underwent BDL
for 2 weeks or 4 intraperitoneal injections of CCl4 (0.5 ␮L
dissolved 1:3 in olive oil, every 3 days). HSCs were cul-
tured for 20 hours (“quiescent HSCs” or “in vivo-acti-
vated HSCs”) or for 5 days (“culture-activated HSCs”) in
Dulbecco’s modified Eagle medium (DMEM) containing
10% fetal bovine serum (FBS), glutamine, HEPES buffer,
and antibiotics. For some experiments, HSCs were iso-
lated from transgenic mice that express green fluorescent
protein (GFP) under the collagen ␣1(I) promoter.15 To
judge contamination of HSC isolations with Kupffer
cells, endothelial cells, and hepatocytes, primary Kupffer
cells, endothelial cells, and hepatocytes were isolated
PANCREAS, AND
BILIARY TRACT
BASIC–LIVER,
from mouse liver and served as reference value (100%).
Kupffer cells were isolated as described above. Hepato-
cytes and endothelial cells were isolated as described
previously.14,16 All animal procedures were approved by
the Columbia University Institutional Animal Care and
Use Committee and meet guidelines of the National
Institutes of Health.
HSC-Kupffer Cell Coculture and LPS
Treatment
Kupffer cells were isolated by collagenase-pronase
perfusion as described above followed by a 15% Nycodenz
gradient and MACS-based positive selection using F4/80-
and CD11b-specific antibodies. HSCs and Kupffer cells
were cocultured in a noncontact-dependent manner by
adding 2.5 ⫻ 106 Kupffer cells to a 0.4 ␮mol/L pore cell
culture insert to 1 ⫻ 106 HSCs plated in a 6-well plate
and exchanging one third of the media daily until RNA
extraction at day 5. To investigate the effect of LPS on
gene expression, HSCs were activated for 3 days treated
with Escherichia coli LPS (100 ng/mL; Sigma-Aldrich Co.,
St. Louis, MO) for 6 hours, followed by RNA extraction.
To investigate the effects of LPS on protein expression,
HSCs were isolated, and LPS (100 ng/mL) was added
daily to the media during the 5-day culture activation,
followed by protein extraction. Each experiment was
based on 3 independent HSC and Kupffer cell isolations.
GASTROENTEROLOGY Vol. 132, No. 5
Flow Cytometric Analysis
One day after isolation, HSCs were scraped into
phosphate-buffered saline, followed by Fc-receptor block-
ade, incubation with allophycocyanin-conjugated anti-
F4/80, flurorescein isothiocyanate (FITC)-conjugated
anti-CD11b, and 2 washes. After gating viable cells, ex-
pression of F4/80 and CD11b was analyzed on channels
FL1 and FL4 using FACS Calibur (Becton Dickinson,
Franklin Lakes, NJ).
RNA Isolation, Labeling, Microarray Data
Analysis, and Quantitative Polymerase Chain
Reaction
RNA was extracted by a combination of Trizol
and RNeasy columns (Qiagen, Valencia, CA). Microarray
analysis was performed using the 1-cycle complementary
DNA (cDNA) synthesis kit, 1-cycle in vitro labeling kit,
and mouse genome 430 2.0 array gene chips (all Af-
fymetrix, Santa Clara, CA) according to the manufactur-
er’s instructions. Chips were scanned with a Genechip
Scanner 3000 7G (Affymetrix). Data normalization was
done by the GCRMA algorithm. Data analysis was per-
formed using Genespring GX 7.3 (Agilent, Palo Alto, CA).
All groups (culture-activated HSCs, n ⫽ 3; BDL-activated
HSCs, n ⫽ 3; CCl4-activated HSCs, n ⫽ 3) were compared
with quiescent HSCs (n ⫽ 5). Significant changes in gene
expression were identified by unpaired t test and subse-
quent Benjamini–Hochberg correction using a P value of
.05 as criterion of statistical significance. Correlation was
investigated by Spearman rank correlation using SPSS
13.0 software for Windows (SPSS Inc., Chicago, IL). Dif-
ferential gene expression was confirmed by quantitative
real-time polymerase chain reaction (qPCR) using com-
mercially available primer-probe sets (Applied Biosys-
tems, Foster City, CA) as described previously.17 All sam-
ples were DNAse treated before reverse transcription and
quantified by comparing threshold cycle values to a seri-
al-dilution standard curve.
Western Blot Analysis
Western blots were performed using 10% to 15%
SDS polyacrylamide gels as previously described.14,17 Af-
ter blocking, membranes were incubated with antibodies
to survivin (Abcam, Cambridge, MA) and to osteopontin
and TIMP-1 (both R&D Systems, Minneapolis, MN),
followed by incubation with horseradish peroxidase-con-
jugated secondary antibodies and enhanced chemolumi-
nescence detection.
Results
Combination of Density Centrifugation and
MACS Achieves High-Purity HSCs
Isolation of HSCs from normal mouse liver usu-
ally results in at least 95% purity.3,14 In HSCs isolated
from bile duct-ligated mice, we detected a high degree of
May 2007 GENE EXPRESSION DURING HSC ACTIVATION 1939

Figure 1. Combination of collagenase-pronase based density centrifugation and MACS achieves high purity of quiescent and activated hepatic
stellate cells. (A and B) Hepatic stellate cells were isolated from normal mice and mice that underwent BDL for 2 weeks and were either depleted of
CD11b- and F4/80-expressing cells or left untreated. Expression of CD11b and F4/80 was analyzed by flow cytometric analysis (A). Expression of
CD68 in quiescent HSCs (“Q”), culture-activated HSCs (“A”), BDL-activated HSCs (“B”), and CCl4-activated HSCs (“C”) was measured by quanti-
tative real-time PCR (B). Results were normalized to 18S and are expressed as fold induction ⫾ standard error of the mean in comparison to Kupffer
cells (“KC” ⫽ 100%). (C and D) Contamination of Kupffer cell-depleted HSCs with hepatocytes (C) and endothelial cells (D) was measured by
quantitative real-time PCR for albumin and CD31, respectively. Results were normalized to 18S and are expressed as fold induction ⫾ standard error

PANCREAS, AND
BILIARY TRACT
of the mean in comparison with primary hepatocytes (“Hep” ⫽ 100%) and primary hepatic endothelial cells (“EC” ⫽ 100%). (E) To demonstrate purity

BASIC–LIVER,
and activation status, HSCs were isolated from transgenic mice expressing GFP under the collagen ␣1(I) promoter. GFP and retinoid autofluores-
cence were checked 1 day after isolation in quiescent HSC and HSC isolated from bile duct-ligated (“BDL”) and CCl4-treated mice and after 5 days
in culture-activated HSCs.

Kupffer cell contamination similar to the levels previ-
ously reported for HSC isolation from CCl4-treated
rats.18 To achieve the high purity required for microarray
analysis, we employed MACS-based Kupffer cell deple-
tion after Nycodenz density centrifugation. F4/80- and
CD11b-based Kupffer cell depletion strongly decreased
the expression levels of macrophage markers F4/80,
CD11b, and CD68 (Figure 1A and B) and was thus
employed for all HSC isolations used in this study. To
exclude contamination with other cell types in HSCs
isolated by this method, we additionally performed real-
time PCR for the hepatocyte marker albumin and the
endothelial cell marker CD31. We found less than 0.05%
hepatocyte contamination and less than 0.2% endothelial
cell contamination in HSCs isolated from normal or
fibrotic livers (Figure 1C and D). Moreover, virtually all
cells isolated by this method displayed fluorescent retin-
oid droplets (Figure 1E). All 3 methods of HSC activation
resulted in a phenotype of activated HSCs and up-regu-
lation of collagen in virtually all HSCs as visualized by
GFP expression in HSCs isolated from mice that express
GFP under the collagen ␣1(I) promoter (Figure 1E).
The Gene Expression Pattern of Culture-
Activated HSCs Is Different From That of
BDL- and CCl4-Activated HSCs
Two thousand seventy-three probe sets were more
than 2-fold up- or down-regulated in at least 1 of the 3
groups of activated HSCs when compared with quiescent
HSCs (Supplementary Table 1; see Supplemental Table 1
online at www.gastrojournal.org). Gene expression pat-
terns in all groups showed high similarity between dif-
ferent isolations confirming consistent purity of isola-
tions and microarray quality (Figure 2A). Although BDL-
and CCl4-activated HSCs displayed an almost identical
pattern of up- and down-regulated genes, gene expression
in culture-activated HSCs overlapped only partially with
BDL- and CCl4-activated HSCs (Figure 2A). To investi-
gate further the extent to which gene expression differed
between culture-activated and in vivo-activated HSCs, we
correlated the levels of down- and up-regulation of all
45,000 probe sets between the different models of HSC
activation. Although correlation between culture-acti-
vated HSC and either BDL- or CCl4-activated HSCs was
 1940 DE MINICIS ET AL
PANCREAS, AND
BILIARY TRACT
BASIC–LIVER,
Figure 1. (Cont’d.)
significant, the level of correlation was relatively weak at
r ⫽ 0.22 and r ⫽ 0.17, respectively (Figure 2B). In con-
trast, correlation between BDL- and CCl4-activated HSCs
was much higher at r ⫽ 0.76. Next, we compared up- and
down-regulated genes among all groups of activated
HSCs. Accordingly, the Venn diagram of all 2073 differ-
entially regulated genes showed that only 56% of up-
regulated and 63% of down-regulated genes from culture-
activated HSCs were concordantly up- and down-
regulated in BDL- and/or CCl4-activated HSCs (Figure
2C). Accordingly, many genes were up- or down-regulated
in in vivo-activated HSCs but not in culture-activated
HSCs (Supplementary Table 2; see Supplemental Table 2
online at www.gastrojournal.org) and vice versa (Supple-
mentary Table 3; see Supplemental Table 3 online at
www.gastrojournal.org). In contrast, BDL- and CCl4-acti-
vated HSCs showed a strong overlap, and more than 95%
and 99% of probe sets of BDL- and CCl4-activated HSCs,
respectively, were concordantly regulated in at least 1 of
the 2 other groups of activated HSCs (Figure 2C). Even
though some genes were significantly up-regulated in
BDL-activated but not in CCl4-activated HSCs and vice
versa, only very few genes showed a strong up- or down-
GASTROENTEROLOGY Vol. 132, No. 5
regulation in one model of in vivo activation but not the
other (Supplementary Tables 4 and 5; see Supplemental
Tables 4 and 5 online at www.gastrojournal.org). Path-
way analysis revealed that differentially expressed genes
were involved in the regulation of proliferation, tran-
scription, and signal transduction, and a smaller percent-
age in inflammation, cell death, adhesion, cytoskeletal
organization, development, and metabolism, but did not
show differences between culture-activated and in vivo-
activated HSCs using these categories (Supplementary
Figure 1; see Supplemental Figure 1 online at www.
gastrojournal.org).
Identification of Novel Genes Associated With
HSC Activation
To assess the validity of the obtained data, we
analyzed whether typical HSC activation markers were
up-regulated in the microarray and additionally exam-
ined their expression by quantitative real-time PCR.
Using this method, we confirmed the strong up-regu-
lation of ␣-SMA and collagen ␣1(I) messenger RNA
(mRNA) levels (Figure 3A and Supplementary Table 1;
see Supplemental Table 1 online at www.gastrojournal.
May 2007 GENE EXPRESSION DURING HSC ACTIVATION 1941

PANCREAS, AND
BILIARY TRACT
BASIC–LIVER,
Figure 2. The gene expression pattern of culture-activated HSCs is different from that of BDL- and CCl4-activated HSCs. Hepatic stellate cells were
isolated from normal mice and mice that underwent BDL for 2 weeks or 4 CCl4 injections. HSCs were considered culture activated after 5 days of
culture in 10% fetal bovine serum on uncoated cell culture dishes. After isolation of mRNA and generation of cRNA, microarray was performed. (A)
Shown is a representative graph of 3 separate HSC isolations per group (“1,” “2,” and “3”) containing all genes that were more than 2-fold
up-regulated (shown in red) or down-regulated (shown in blue) in at least 1 of the 3 groups of activated HSCs in comparison with quiescent HSCs
(t test, P ⬍ .05 followed by Benjamini–Hochberg correction). (B) The relative expression of all 45,000 probe sets of the array was correlated among
culture-activated HSCs, BDL-activated HSCs, and CCl4-activated HSCs using Spearman rank correlation. (C) The Venn diagram shows overlapping
patterns of probe sets that were significantly (P ⬍ .05) and at least 2-fold up-regulated and down-regulated in 1 of the 3 groups of activated HSCs.
Probe sets that were 2-fold up-regulated or down-regulated in one group were considered down-regulated or up-regulated in the other groups if they
showed at least 1.5-fold up-regulation or down-regulation.

org). In addition, we discovered differential expression
of a large number of genes whose expression has not
been associated with HSC activation and confirmed
their expression by real-time PCR (Figure 3A). We
identified genes involved in proliferation (Ki 67), cell
survival (survivin, septin 4), cytoskeletal organization
and contraction (plexin C1, anillin, leiomodin), in-
flammation (CXCL14, osteopontin), ECM organiza-
tion and degradation (Col8a1, MMP10), and lipid me-
tabolism (cholesterol 25-hydroxylase) as well as
transcription factors (PRX1, LMCD1), cell surface re-
ceptors (adenosine receptor 2b, GPR91), bone morpho-
genetic protein BMP5, and neural markers (synapto-
tagmin, neurotrimin) (Figure 3A and Supplementary
Table 1; see Supplemental Table 1 online at www.
gastrojournal.org). Although the inhibitor of apo-
ptosis protein (IAP) family member surviving was up-
regulated during HSC activation, other IAP members
such as c-IAP1, c-IAP2, and XIAP were not up-regu-
lated (Figure 3A). Expression patterns of several of
these genes were confirmed by Western blot analysis.
The expression pattern of all investigated proteins was
similar to our PCR and microarray analysis data, with
survivin being up-regulated in both culture- and in
vivo-activated HSCs, TIMP-1 being up-regulated in in
vivo- but not in culture-activated HSCs, and osteopon-
 1942 DE MINICIS ET AL
PANCREAS, AND
BILIARY TRACT
BASIC–LIVER,
tin being down-regulated in culture-activated HSCs
but not in in vivo-activated HSCs (Figure 3B).
Coculture With Kupffer Cells or Exposure to
LPS Corrects the Gene Expression Pattern of
Culture-Activated HSCs
Culture activation of HSCs lacks the microenviron-
ment that is typical of HSC activation in vivo. The presence
of Kupffer cells and exposure to LPS have been demon-
strated to be critical for HSC activation and fibrogenesis in
vivo.7–13 To investigate whether the absence of these factors
contributes to the aberrant gene expression pattern during
culture activation, we cocultured HSCs with Kupffer cells or
exposed them to LPS and then analyzed genes that were
differentially expressed in culture- and in vivo-activated
GASTROENTEROLOGY Vol. 132, No. 5
Figure 3. Identification of pre-
viously known and novel differ-
entially regulated genes during
HSC activation in vivo and in
culture. (A) Expression of differ-
entially regulated genes was
confirmed in quiescent HSCs
(“Q”), culture-activated HSCs
(“A”), BDL-activated HSCs
(“B”), and CCl4-activated HSCs
(“C”) by qPCR. Results were
normalized to 18S and are ex-
pressed as fold induction ⫾
standard error of the mean in
comparison with quiescent
HSCs. Each group of HSC sam-
ples consisted of at least 3 dif-
ferent isolations. (B) Expression
of selected genes was con-
firmed by Western blot analysis.
Each Western blot is based on
HSCs isolated from 3 normal,
BDL-, or CCl4-treated mice and
is representative of at least 2 in-
dependent experiments.
HSCs. Coculture of HSCs with Kupffer cells in a noncon-
tact-dependent manner resulted in a striking up-regulation
of TIMP-1, PRX1, LMCD1, adenosine receptor 2b, CXCL14,
osteopontin, and cholesterol 25-hydroxlase that was equal
or exceeded the levels observed in in vivo-activated HSCs
(Figure 4A). Conversely, coculture with Kupffer cells down-
regulated the expression levels of GRP91 to a level that was
comparable with in vivo-activated HSCs (Figure 4A). Several
genes that were down-regulated by coculture with Kupffer
cells were also down-regulated by LPS, such as GRP91,
septin 4, and BMP5; or up-regulated, such as TIMP-1 and
cholesterol 25-hydroxylase (Figure 4B). Other genes, such as
ankyrin, were down-regulated after exposing HSCs to LPS
but not to Kupffer cell coculture. To confirm further these
May 2007 GENE EXPRESSION DURING HSC ACTIVATION 1943

Figure 4. Exposure of HSCs to

Kupffer cells or LPS during cul-
ture shifts the gene expression
toward the pattern of in vivo-ac-
tivated HSCs. (A and B) Genes
that showed a different expres-
sion pattern during culture and
in vivo activation were examined
after 5 days of culture activation
in the presence or absence of
Kupffer cells or LPS (100 ng/mL)
by quantitative real-time PCR as
described in the Materials and

PANCREAS, AND
BILIARY TRACT
BASIC–LIVER,
Methods section (“Q,” quies-
cent; “A,” culture activated with-
out additional stimuli; “K,”
Kupffer cell treated; “L,” LPS
treated). Shown is the mean of
PCRs from 3 independent HSC
isolations. (C) The influence of
Kupffer cells and LPS on the ex-
pression of selected genes was
confirmed by Western blot anal-
ysis. For this purpose, HSCs
were cultured in the presence or
absence of Kupffer cells in a
noncontact-dependent manner
for 5 days or in the presence of
LPS (100 ng/mL). This experi-
ment is representative of at least
2 independent HSC isolations.

findings, we performed Western blot analysis, which dem-
onstrated an up-regulation of TIMP-1 and osteopontin af-
ter exposure to Kupffer cells or LPS (Figure 4C). Effects of
LPS were more pronounced at the protein level with an
up-regulation of osteopontin (not observed on the mRNA
level) and a higher induction of TIMP-1, which is likely due
to the longer exposure to LPS in protein expression exper-
iments. Some differentially expressed genes, such as leiomo-
din and TGF-␤3, were not corrected by LPS treatment or
Kupffer cell coculture toward the pattern observed in in
vivo-activated HSCs (Figure 4A and B), suggesting that
Kupffer cells and LPS are important but are not the only
components of the microenvironment that regulate the
gene expression during HSC activation in vivo.
Discussion
Our microarray study detected a large number of
previously known and unknown genes that were up- or
down-regulated during HSC activation, thus confirming
 1944 DE MINICIS ET AL
the ECM-producing and -degrading, proliferative, con-
tractile, and inflammatory phenotype of activated HSCs.
Although culture activation up-regulated typical HSC
activation markers, such as ␣-SMA and collagen ␣1(I), it
did not fully reproduce the changes of gene expression
patterns seen in vivo. The finding that in vivo activation
of HSCs up-regulates a different set of genes than culture
activation is not completely surprising in view of the fact
that the microenvironment of HSCs in the fibrotic liver is
complex and exposes HSCs to a number of cellular and
humoral interactions, whereas culture activation occurs
in an artificial environment and does not include inter-
actions with other cell types unless there is a high degree
of contamination. Culture activation is by far the most
common model of HSC activation. Our data suggest that
the common practice of evaluating genetic and pharma-
cologic manipulations during HSC culture activation
may produce results that do not reflect effects on HSCs
and fibrogenesis in vivo. Our results are similar to a
recently published study by Sancho-Bru et al,19 and to
some degree to a proteome-based study by Kristensen et
al, who found a significant but only 60% overlap of 43
activation-associated proteins.18 Although the study by
Sancho-Bru et al19 compared HSC expression patterns
with whole liver, our study compared gene expression
patterns of activated HSCs to quiescent HSCs and thus
had a higher power to detect differentially regulated
HSC-specific genes. Several other groups have recently
performed microarray analysis of genes that are up-reg-
PANCREAS, AND
BILIARY TRACT
BASIC–LIVER,
ulated during HSC activation, but all these studies were
restricted to HSC activation in culture and used passaged
quiescent and activated HSCs.20,21 To understand further
the HSC gene expression in vivo, we additionally inves-
tigated gene expression patterns after different fibrogenic
stimuli. Our results show that 2 completely different
models of experimental fibrosis, BDL and CCl4, induce
very similar HSC gene expression patterns, suggesting
that HSCs undergo a conserved activation process in vivo
independently of the nature of the initial fibrogenic stim-
ulus. Moreover, the similar gene expression patterns of
BDL- and CCl4-activated HSCs imply that HSCs exert
virtually the same functions in different models of he-
patic injury.
In addition to detecting classical HSC activation mark-
ers, our study identified a large number of genes that
have not been previously described to be associated with
HSC activation. The expression of previously identified
HSC activation-associated genes such as ␣-SMA, TIMP-1,
collagen ␣1(I), CD9, cyclin B1, fibronectin 1, calponin-1,
DKK3, frizzled 2, lysyl oxidase, TGF-␤3, and ␥-SMA19 –21
was confirmed in our microarray. In comparison with
previous studies, our study showed a lower induction of
HSC activation markers such as collagen ␣1(I) and
␣-SMA. This difference is not only due to comparing
activated HSCs with quiescent HSCs instead of whole
liver in our study but also due to the fact that we
GASTROENTEROLOGY Vol. 132, No. 5
depleted Kupffer cells from the HSC isolate and culture-
activated HSCs for 5 days only. Among novel differen-
tially regulated genes identified in in vivo-activated HSCs
are genes involved in ECM formation and degradation,
such as procollagen type VIII ␣1 and MMP10 (both
up-regulated), neural crest markers including synapto-
tagmin-9 and neurotrimin (both down-regulated), genes
involved in apoptosis such as the IAP member survivin,22
genes involved in cytokinesis and proliferation such as
the actin-binding protein anillin and Ki-67 (both up-
regulated), genes involved in cell spreading and migra-
tion such as the semaphorin receptor plexin C1 (down-
regulated),23,24 chemokine CXCL14 (up-regulated), bone
morphogenetic protein BMP5 (down-regulated), and
cholesterol 25-hydroxylase (up-regulated), an enzyme
that prevents the cleavage and activation of SREBP-1 and
SREBP-2 and may thus promote an adipocyte-like phe-
notype in HSCs.25,26 In addition, we identified differential
expression of several receptors, such as the G-protein-
coupled receptors adenosine receptor 2a and GPR91, and
transcription factors and transcription factor binding
proteins, such as PRX1 and the GATA6 repressor
LMCD1.27 Several genes, such as LMCD1, osteopontin,
survivin, and PRX1,28 are expressed in vascular smooth
muscle cells and are involved in atherosclerosis22,27,29,30
and are likely to exert similar functions in HSCs. Other
genes, such as GPR91, are highly expressed in the liver,
but their function in the liver remains unknown.31 The
IAP family member survivin was up-regulated during
HSC activation, whereas c-IAP1, c-IAP2, and XIAP levels
remained unchanged, suggesting a potential role for sur-
vivin in protecting HSCs from apoptosis during hepatic
fibrogenesis.
One common problem in microarray studies of pri-
mary cells is contamination with other cell types, which
may lead to false identification of differentially regulated
genes. Our study demonstrated a low degree of contam-
ination with hepatocytes, Kupffer cells, and endothelial
cells in all HSC isolations. Moreover, we excluded cell
contamination as a reason for the different gene expres-
sion patterns during culture and in vivo activation by
additionally examining each investigated gene in Kupffer
cells, hepatocytes, and endothelial cells and setting this
expression level into relation to the degree of contami-
nation (data not shown). The only gene we found to be
falsely up-regulated in in vivo-activated HSCs was albu-
min, despite an extremely low degree of hepatocyte con-
tamination (Supplementary Table 1; see Supplemental
Table 1 online at www.gastrojournal.org). This effect was
caused by the abundant expression of albumin in hepa-
tocytes and its complete absence in HSCs. Another po-
tential problem lies within the density-based purification,
which is required to obtain a pure HSC population from
fibrotic livers. This isolation method only purifies HSCs
that contain lipid droplets and may thus potentially
contain a large number of quiescent HSCs. However, our
May 2007
data show activation of almost all HSCs in the isolate
from mice after CCl4 treatment or BDL as judged by
collagen ␣1(I)-driven GFP expression, suggesting that
gene expression changes are representative for the major-
ity of activated HSCs in the fibrotic liver. Moreover, we
achieved a similar activation status of in vivo- and cul-
ture-activated HSCs as seen by the similar up-regulation
of classical HSC activation markers and a similar retinoid
content in culture- and in vivo-activated HSCs.
The fact that culture- and in vivo-activated HSCs display
a significantly different gene expression pattern suggests
that one or several factors that drive HSC activation in vivo
are missing during culture activation. Our results indicate
that the lack of Kupffer cells and LPS are, at least in part,
responsible for the gene expression pattern of culture-acti-
vated HSCs. Following coculture with Kupffer cells or ex-
posure to LPS, a large number of genes, such as TIMP-1,
PRX1, LMCD1, adenosine receptor 2b, CXCL14, osteopon-
tin, cholesterol 25-hydroxylase, GRP91, BMP5, and ankyrin,
shifted toward the expression pattern of in vivo-activated
HSCs. These results are consistent with recent studies that
show an important role for Kupffer cells and LPS and its
receptor TLR4 in HSC in vivo activation and hepatic fibro-
sis.7–13 Because several genes were not affected by LPS treat-
ment or Kupffer cell coculture, it is likely that additional
factors, eg, cytokines and apoptotic hepatocytes, regulate
some of these genes in the fibrotic liver. Moreover, HSCs
dramatically up-regulated genes such as osteoglycin and
insulin-like growth factor-binding protein 3 during culture
activation but not in vivo activation, suggesting that culture
on plastic surfaces artificially drives the expression of some
genes and thus contributes to the aberrant gene expression
during culture activation. Our results indicate that improve-
ment of HSC culture activation may lead to more physio-
logic models of HSC activation and that culture activation
may be useful to study profibrogenic pathways activated by
Kupffer cells and potentially other hepatic cell types. None-
theless, HSC in vivo activation should be considered the
gold standard for the study of HSC biology.
Appendix
Supplementary Data
Supplementary data associated with this article
can be found, in the online version, at doi:10.1053/j.
gastro.2007.02.033.
References
1. Friedman SL. Molecular regulation of hepatic fibrosis, an inte-
grated cellular response to tissue injury. J Biol Chem 2000;275:
2247–2250.
2. Bataller R, Brenner DA. Liver fibrosis. J Clin Invest 2005;115:
209 –218.
3. Bataller R, Schwabe RF, Choi YH, Yang L, Paik YH, Lindquist J,
Qian T, Schoonhoven R, Hagedorn CH, Lemasters JJ, Brenner DA.
NADPH oxidase signal transduces angiotensin II in hepatic stel-
GENE EXPRESSION DURING HSC ACTIVATION 1945
late cells and is critical in hepatic fibrosis. J Clin Invest
2003;112:1383–1394.
4. Saxena NK, Ikeda K, Rockey DC, Friedman SL, Anania FA. Leptin
in hepatic fibrosis: evidence for increased collagen production in
stellate cells and lean littermates of ob/ob mice. Hepatology
2002;35:762–771.
5. Rockey DC, Chung JJ. Endothelin antagonism in experimental
hepatic fibrosis. Implications for endothelin in the pathogenesis
of wound healing. J Clin Invest 1996;98:1381–1388.
6. Svegliati-Baroni G, Ridolfi F, Di Sario A, Casini A, Marucci L,
Gaggiotti G, Orlandoni P, Macarri G, Perego L, Benedetti A, Folli F.
Insulin and insulin-like growth factor-1 stimulate proliferation and
type I collagen accumulation by human hepatic stellate cells:
differential effects on signal transduction pathways. Hepatology
1999;29:1743–1751.
7. Paik YH, Schwabe RF, Bataller R, Russo MP, Jobin C, Brenner DA.
Toll-like receptor 4 mediates inflammatory signaling by bacterial
lipopolysaccharide in human hepatic stellate cells. Hepatology
2003;37:1043–1055.
8. Isayama F, Hines IN, Kremer M, Milton RJ, Byrd CL, Perry AW,
McKim SE, Parsons C, Rippe RA, Wheeler MD. LPS signaling
enhances hepatic fibrogenesis caused by experimental cholesta-
sis in mice. Am J Physiol Gastrointest Liver Physiol 2006;290:
G1318 –G1328.
9. Seki E, De Minicis S, Osawa Y, Österreicher CH, Brenner DA,
Schwabe RF. TLR4 mediates hepatic fibrosis by down-regulating
TGF-␤ pseudoreceptor BAMBI and enhancing TGF-␤ signaling.
Hepatology 2006;44:225A.
10. Friedman SL, Arthur MJ. Activation of cultured rat hepatic lipo-
cytes by Kupffer cell conditioned medium. Direct enhancement of
matrix synthesis and stimulation of cell proliferation via induction
of platelet-derived growth factor receptors. J Clin Invest 1989;
84:1780 –1785.
11. Rivera CA, Bradford BU, Hunt KJ, Adachi Y, Schrum LW, Koop DR,
PANCREAS, AND
Burchardt ER, Rippe RA, Thurman RG. Attenuation of CCl(4)-
BILIARY TRACT
BASIC–LIVER,
induced hepatic fibrosis by GdCl(3) treatment or dietary glycine.
Am J Physiol Gastrointest Liver Physiol 2001;281:G200 –G207.
12. Duffield JS, Forbes SJ, Constandinou CM, Clay S, Partolina M,
Vuthoori S, Wu S, Lang R, Iredale JP. Selective depletion of
macrophages reveals distinct, opposing roles during liver injury
and repair. J Clin Invest 2005;115:56 – 65.
13. Nieto N. Oxidative-stress and IL-6 mediate the fibrogenic effects
of rodent Kupffer cells on stellate cells. Hepatology 2006;44:
1487–1501.
14. Siegmund SV, Seki E, Osawa Y, Uchinami H, Cravatt BF, Schwabe
RF. Fatty acid amide hydrolase determines anandamide-induced
cell death in the liver. J Biol Chem 2006;281:10431–10438.
15. Magness ST, Bataller R, Yang L, Brenner DA. A dual reporter gene
transgenic mouse demonstrates heterogeneity in hepatic fibro-
genic cell populations. Hepatology 2004;40:1151–1159.
16. Liu S, Premont RT, Kontos CD, Zhu S, Rockey DC. A crucial role
for GRK2 in regulation of endothelial cell nitric oxide synthase
function in portal hypertension. Nat Med 2005;11:952–958.
17. Schwabe RF, Sakurai H. IKK␤ phosphorylates p65 at S468 in
transactivation domain 2. FASEB J 2005;19:1758 –1760.
18. Kristensen DB, Kawada N, Imamura K, Miyamoto Y, Tateno C,
Seki S, Kuroki T, Yoshizato K. Proteome analysis of rat hepatic
stellate cells. Hepatology 2000;32:268 –277.
19. Sancho-Bru P, Bataller R, Gasull X, Colmenero J, Khurdayan V,
Gual A, Nicolas JM, Arroyo V, Gines P. Genomic and functional
characterization of stellate cells isolated from human cirrhotic
livers. J Hepatol 2005;43:272–282.
20. Jiang F, Parsons CJ, Stefanovic B. Gene expression profile of
quiescent and activated rat hepatic stellate cells implicates Wnt
signaling pathway in activation. J Hepatol 2006;45:401– 409.
21. Liu XJ, Yang L, Luo FM, Wu HB, Qiang Q. Association of differen-
tially expressed genes with activation of mouse hepatic stellate
 1946 DE MINICIS ET AL
cells by high-density cDNA microarray. World J Gastroenterol
2004;10:1600 –1607.
22. Blanc-Brude OP, Yu J, Simosa H, Conte MS, Sessa WC, Altieri DC.
Inhibitor of apoptosis protein survivin regulates vascular injury.
Nat Med 2002;8:987–994.
23. Aizawa H, Wakatsuki S, Ishii A, Moriyama K, Sasaki Y, Ohashi K,
Sekine-Aizawa Y, Sehara-Fujisawa A, Mizuno K, Goshima Y, Ya-
hara I. Phosphorylation of cofilin by LIM-kinase is necessary for
semaphorin 3A-induced growth cone collapse. Nat Neurosci
2001;4:367–373.
24. Walzer T, Galibert L, Comeau MR, De Smedt T. Plexin C1 engagement
on mouse dendritic cells by viral semaphorin A39R induces actin
cytoskeleton rearrangement and inhibits integrin-mediated adhesion
and chemokine-induced migration. J Immunol 2005;174:51–59.
25. Lund EG, Kerr TA, Sakai J, Li WP, Russell DW. cDNA cloning of
mouse and human cholesterol 25-hydroxylases, polytopic mem-
brane proteins that synthesize a potent oxysterol regulator of
lipid metabolism. J Biol Chem 1998;273:34316 –34327.
26. She H, Xiong S, Hazra S, Tsukamoto H. Adipogenic transcrip-
tional regulation of hepatic stellate cells. J Biol Chem 2005;280:
4959 – 4967.
27. Rath N, Wang Z, Lu MM, Morrisey EE. LMCD1/Dyxin is a novel
transcriptional cofactor that restricts GATA6 function by inhibiting
DNA binding. Mol Cell Biol 2005;25:8864 – 8873.
28. Peterson RE, Hoffman S, Kern MJ. Opposing roles of two iso-
forms of the Prx1 homeobox gene in chondrogenesis. Dev Dyn
2005;233:811– 821.
PANCREAS, AND
BILIARY TRACT
BASIC–LIVER,
GASTROENTEROLOGY Vol. 132, No. 5
29. Giachelli CM, Bae N, Almeida M, Denhardt DT, Alpers CE,
Schwartz SM. Osteopontin is elevated during neointima forma-
tion in rat arteries and is a novel component of human athero-
sclerotic plaques. J Clin Invest 1993;92:1686 –1696.
30. Yoshida T, Hoofnagle MH, Owens GK. Myocardin and Prx1 con-
tribute to angiotensin II-induced expression of smooth muscle
␣-actin. Circ Res 2004;94:1075–1082.
31. He W, Miao FJ, Lin DC, Schwandner RT, Wang Z, Gao J, Chen
JL, Tian H, Ling L. Citric acid cycle intermediates as ligands for
orphan G-protein-coupled receptors. Nature 2004;429:188 –
193.
Received May 30, 2006. Accepted January 24, 2007.
Address requests for reprints to: Robert F. Schwabe, MD, Depart-
ment of Medicine, Columbia University, College of Physicians & Sur-
geons, Russ Berrie Pavillion, Room 415, 1150 St. Nicholas Ave, New
York, New York 10032. e-mail: rfs2102@columbia.edu; fax: (212)
851-5461.
S.D.M. and E.S. contributed equally to this paper.
Supported by an Alimenti e Salute grant from the University of
Ancona, Italy (to S.D.M.), a grant from the Yamanouchi Foundation for
Research of Metabolic Disorders (to E.S.), and a Research Scholar
Award from the American Gastroenterological Association and Procter
& Gamble Co. (to R.F.S).