Activation of Platelet-Derived Growth Factor Receptor Alpha Contributes to Liver Fibrosis

Chronic liver injury leads to fibrosis, cirrhosis, and loss of liver function. Liver cirrhosis is the 12th leading cause of death in the United States, and it is the primary risk factor for developing liver cancer. Fibrosis and cirrhosis result from activation of hepatic stellate cells (HSCs), which are the primary collagen producing cell type in the liver. Here, we show that platelet-derived growth factor receptor α (PDGFRα) is expressed by human HSCs, and PDGFRα expression is elevated in human liver disease. Using a green fluorescent protein (GFP) reporter mouse strain, we evaluated the role of PDGFRα in liver disease in mice and found that mouse HSCs express PDGFRα and expression is upregulated during carbon tetrachloride (CCl4) induced liver injury and fibrosis injection. This fibrotic response is reduced in Pdgfrα heterozygous mice, consistent with the hypothesis that liver fibrosis requires upregulation and activation of PDGFRα. These results indicate that Pdgfrα expression is important in the fibrotic response to liver injury in humans and mice, and suggest that blocking PDGFRα–specific signaling pathways in HSCs may provide therapeutic benefit for patients with chronic liver disease.


Introduction
Chronic liver injury is a major cause of morbidity and mortality in the US and worldwide, due to complications of liver fibrosis, cirrhosis, and hepatocellular carcinoma (HCC) [1]. To date, there are no effective treatments for patients with liver fibrosis, so a better understanding of pathways that regulate fibrosis has great clinical potential [2]. Many inflammatory cytokines and growth factors are released during liver injury, including platelet derived growth factors (PDGFs), which are potent mitogens for hepatic stellate cells (HSCs) [2,3]. The PDGF family of ligands and receptors plays a central role in repair after injury, and are key regulators of the formation of connective tissue [4,5]. Elevated platelet-derived growth factor receptor (PDGFR) expression is detected in human heart disease, pulmonary fibrosis, and kidney fibrosis [6][7][8], and blocking PDGFR signaling decreases collagen deposition after myocardial infarct, in pulmonary fibrosis, and in kidney fibrosis [9][10][11]. Thus, targeting the PDGF pathway may modulate liver fibrosis.
There are five known functional ligand dimers in the PDGF family, -AA, -AB, -BB, -CC, and -DD, which bind cell surface receptor tyrosine kinases comprised of PDGFRa and PDGFRb subunits [12]. PDGFs stimulate the migration and proliferation of mesenchymal cells during development [13]. Loss of PDGFRs leads to significant abnormalities in mice [14,15]. PDGFRb is critical to vascular and hematopoietic development, and cell specific deletion or activation of PDGFRb results in failure or increased pericyte and vascular smooth muscle cell coverage of blood vessels in mice [14,16,17]. PDGFRa is required for migration and survival of neural crest cells and for skeletal development, and cell specific deletion of PDGFRa decreases bcell proliferation in the pancreas and ventricular septation of the heart [15,18,19]. Constitutive activation of PDGFRa causes fibrosis that is particularly noticeable in intestine, skin, muscle and heart, but activation has to be conditionally induced in late prenatal or adult animals, as constitutive PDGFRa activation causes lethality [5]. Deleting one allele of Pdgfra in mice does not affect development, unlike the observed phenotype in homozygous knockout mice [15,20]. PDGF signal transduction pathways play a prominent role in fibrosis [21]. It has been suggested that PDGFRa signaling is more likely to induce fibrosis than PDGFRb [22], however this notion has not been conclusively demonstrated in the liver. In summary, PDGF signaling is tightly regulated by abundance and degree of signal transduction, and perturbing either results in developmental defects and organ dysfunction.
In the present study we analyzed PDGFR in human liver disease, human liver cell lines, and a mouse model of liver injury and fibrosis. We found increased PDGFRa in human liver specimens with fibrosis and cirrhosis. PDGFRa is primarily expressed in HSCs, and Pdgfra expression increased in injured mouse livers. We investigated the role of PDGFRa in liver fibrosis using mice with only one allele of Pdgfra, and found that reducing Pdgfra copy number inhibits liver fibrosis in mice. Together our data suggest that PDGFRa inhibitors could be an effective means to reduce liver fibrosis in patients.

Animals
Mice were housed in a specific pathogen-free environment overseen by the Department of Comparative Medicine at the University of Washington with IACUC approval under protocol #4295-01. Mice that express nuclear localized green fluorescent protein (GFP) driven by the endogenous Pdgfra promoter, Pdgfra nGFP , were purchased from the Jackson Laboratory (007669) [20]. Either wild type (WT) littermates that retain both Pdgfra alleles, or control mice, i.e. male C57BL/6 mice purchased from the Jackson Laboratory (000664), were used as experimental controls. To induce fibrosis, mice were injected (i.p.) with 10 ml/g body weight CCl 4 (Sigma-Aldrich) diluted in olive oil 10% (v/v), either one time (acute injury) or twice weekly for four or six weeks (chronic injury). Olive oil-injected animals served as controls for CCl 4 -injected mice. Animals were sacrificed using CO 2 inhalation. The Institutional Animal Care and Use Committee of the University of Washington, which is certified by the Association for Assessment and Accreditation of Laboratory Animal Care International, approved all experiments.

Human Liver Samples
Human liver and HCC specimens were obtained from the University of Washington Medical Center after IRB-approval. IHC was performed on liver specimens from patients with cirrhosis who underwent liver transplantation surgery at the University of Washington Medical Center from 1989 to 2002, HSD #23602 (MMY) [23]. Immunoblot analysis was performed on resected liver specimens collected after receiving informed consent IRB #31281 (RSY). All samples were de-identified of any patient information. Specimens were either fixed in formalin or frozen at 280uC until use.

Immunohistochemistry (IHC) and Histological Staining
Formalin-fixed liver tissue was processed and embedded in paraffin using standard protocols, and IHC was performed as previously described [24], using the primary antibodies listed in Table S1. A board-certified clinical liver pathologist (MMY)  reviewed all human samples and determined the presence of cirrhosis and/or tumor and assessed for PDGFRa and PDGFRb immunoreactivity. To quantify fibrosis, formalin-fixed liver tissue was stained with picrosirius red. For morphometric analysis, picrosirius red area was imaged under polarized light [25]. Images were analyzed using NIH image J software to convert pixels to binary values and determine the relative number of positive and negative pixels.

Immunoblotting
Tissues were homogenized in a 1% Triton-x 100 lysis buffer and processed as described [26]. Membranes were incubated with primary antibodies overnight at 4uC, and then with the appropriate horseradish peroxidase (HRP)-conjugated secondary antibodies. Primary antibodies used in this study are listed in Table S1.

Immunofluorescence (IF) and ex-vivo Imaging
Livers were fixed in 4% paraformaldehyde overnight, and tissues were frozen in optimum cutting temperature compound for cryosectioning. IF was performed using standard techniques, with liver sections incubated overnight with the primary antibodies listed in Table S1. Immune complexes were detected with goat Alexa 633 conjugated anti-rat IgG (A-21094, Life Technologies) and goat Alexa 546 conjugated anti-rabbit IgG (A-11010, Life Technologies) antibodies. Sections were mounted with SlowFade Gold (S36936, Life Technologies) and imaged with a Leica SL confocal microscope (Leica Microsystems, Keck Center UW). For ex-vivo imaging, freshly harvested livers were analyzed as previously described [27]. Images were captured using a Zeiss 510 Meta confocal microscope. In Pdgfra WT/nGFP mice, GFP fluorescence was used to report PDGFRa positive cells [20].

Cell Culture and Proliferation Assay
Liver cell lines (Table S2) were grown in a 37uC incubator with 95% humidity and 5% CO2 in DMEM (Life Technologies) with 10% FBS. Confluent cells were split and allowed to attach to plates as described [28]. Cells were serum-starved for 24 hours then stimulated with 10 ng/ml PDGF-AA, -AB,-BB, or -CC (R&D systems) for 24 hours. [ 3 H]Thymidine (1 mCi/ml final concentration) was added to the media for the final 3 hours of stimulation. Unincorporated [ 3 H] thymidine was removed from the cells, and trichloroacetic acid was used to precipitate protein-bound DNA. DNA was solubilized in NaOH, quantified using a scintillation counter, and measured in triplicate.

RNA Expression Analysis
RNA was extracted from cells or liver tissue using Trizol (15596-018, Life Technologies) as described by the manufacturer. Reactions contained cDNA synthesized from 0.5 mg RNA using MMLV (28025-013, Life Technologies), and Taqman Universal Mastermix II (4440040, Life Technologies). Cycling conditions were 95uC for 10 min, and 49 cycles of 95uC for 15 sec, 60uC for 60 sec with a final extension at 72uC for 1 min. Data are represented as delta delta Ct values after normalization to Gapdh mRNA levels. Primers used in this experiment are listed in Table  S3.

Statistical Analysis
Statistical significance was analyzed using Prism software (Graphpad), either with Kruskal-Wallis non-parametric ANOVA with significance p,0.05, or Mann-Whitney U test with significance p,0.05, as indicated in the figure legends.

Expression of PDGFRa in Human Cirrhosis and HCC
Previous studies have demonstrated that over expression of PDGF ligands induces fibrosis in mice [29][30][31], and elevated expression of PDGFRb in chronic liver disease has been reported [32][33][34]. As PDGF ligands can activate both PDGFRa and PDGFRb, we sought to investigate the role of PDGFRa in chronic liver injury. 80-90% of human HCC arise in the setting of a cirrhotic liver, in which HSCs have been activated [35], so we first performed IHC analysis to determine whether PDGFRa and PDGFRb levels are elevated in human cirrhosis and HCC. Fibrotic and cirrhotic livers had focal perisinusoidal immunoreactivity for PDGFRa, which was stronger in steatotic and cirrhotic livers, while normal adult liver had relatively little PDGFRa immunoreactivity ( Figure 1 A-D). PDGFRb immunoreactivity was also increased in the fibrotic and cirrhotic areas compared to un-injured liver (data not shown). Table S4 summarizes PDGFRa and PDGFRb immunoreactivity in diseased human liver specimens, 77% of which demonstrated increased PDGFRa immunoreactivity and 56% of which demonstrated increased PDGFRb immunoreactivity. Using a separate set of specimens, we compared PDGFR protein levels in grossly dissected HCC tumors to those of adjacent non-tumor livers from the same patients by immunoblot analysis. PDGFRa protein was frequently detected in the non-tumor tissue ( Figure S1, Table S5). One specimen, patient 5, had detectable PDGFRa protein by immunoblot in the tumor (Table S5). IHC analysis of this specimen demonstrated PDGFRa immunoreactive cells within the tumor (Figure 1 E, F), but these cells did not have the histological appearance of hepatocytes, suggesting that non-parenchymal cells (NPCs) had invaded the parenchymal tumor and account for the PDGFRa immunoreactivity observed in this tumor by immunoblot analysis. Taken together, our data suggest that PDGFRa is expressed primarily in fibrotic and cirrhotic livers, predominantly in NPCs.

Expression of PDGFRa and PDGFRb in Human Liver and Stellate Cell Lines
We next analyzed mRNA transcripts from human liver cell lines, and found that both PDGFRa and PDGFRb mRNA are expressed in non-diseased human liver (Figure 2), but that the human HSC line LX-1 has a significantly higher relative expression of PDGFRa (Figure 2A). LX-2 cells, a LX-1 subclone, express PDGFRb, albeit to variable levels, and LX-1 cells express little to no PDGFRb ( Figure 2B). Transcription of both PDGFRs is reduced in human hepatoma cell lines compared to whole liver, suggesting that PDGFRs are predominantly expressed in NPCs. We next stimulated various cell lines with PDGF ligands, and found that PDGF -AA, -AB, -BB, and -CC lead to robust proliferation in stellate cells, but these ligands had little effect on the hepatocyte or hepatoma cell lines tested (Table S6).

Increased Expression of Pdgfra and Pdgfrb in Mice after CCl 4 induced Hepatocyte Injury
To investigate the role of PDGFRa in liver injury and fibrosis, we used the well-established model of CCl 4 injection, in which HSCs are activated in response to necroinflammatory injury to hepatocytes [36]. CCl 4 injury to rats has been shown to induce Pdgfrs mRNA in the liver [37]. To determine whether Pdgfr expression is induced after liver injury in mice, WT mice were injected with a single dose of CCl 4 . We found that Pdgfra expression increased during 72 hours after injury ( Figure 3A). CCl 4 injection also induced expression of Pdgfrb, although to a differing extent and with a different time course than Pdgfra ( Figure 3B). Thus, acute CCl 4 exposure induces Pdgfr expression in the liver.

Cells Expressing Pdgfra and Pdgfrb Respond to CCl 4 Injury
To determine the liver cell type that expresses PDGFRa in response to liver injury, we used transgenic Pdgfra WT/nGFP mice, in Figure 6. Compared to C57BL/6 mice, chronically CCl 4 injured Pdgfra WT/nGFP mice have reduced transcription of fibrotic genes. Livers from Pdgfra WT/nGFP and C57BL/6 mice that were either uninjured (controls) or treated for 4 weeks with CCl 4 were used to prepare total liver RNA. A) Compared to C57BL/6 mice, expression of Pdgfra is decreased in Pdgfra WT/nGFP mice in uninjured mice and after chronic CCl 4 . B) Expression of Pdgfrb does not differ between C57BL/6 and Pdgfra WT/nGFP mice. C) Expression of Acta2 is increased in uninjured Pdgfra WT/nGFP mice compared to C57BL/6 and decreased after chronic CCl 4 between Pdgfra WT/nGFP and C57BL/6 mice. D) Expression of Col1a1 is similar in uninjured Pdgfra WT/nGFP mice compared to C57BL/6 and decreased after chronic CCl 4 between Pdgfra WT/nGFP and C57BL/6 mice. E) Expression of Col4 is similar in Pdgfra WT/nGFP mice compared to C57BL/6 in both uninjured and chronic CCl 4 injected mice. F) Expression of Timp1 is increased to a similar level in both genotypes. Samples were processed as described in figure 2. Values are represented as means with SEM, and were analyzed by Mann-whitney non-parametric U test * = p,0.05, n = 3-6 mice per time point. doi:10.1371/journal.pone.0092925.g006 which the endogenous Pdgfra promoter initiates transcription of nuclear-restricted GFP reporter in place of the Pdgfra gene [20]. These mice have one copy of Pdgfra replaced by GFP and are thus heterozygous for Pdgfra. While a single injection of CCl 4 induces necrosis and injury that is repaired within seven days, repeated injection of CCl 4 induces liver fibrosis [36]. Vehicle-injected Pdgfra WT/nGFP mice have histologically normal liver ( Figure 4A). A single CCl 4 injection increases the density of small cells with a high nuclear to cytoplasmic ratio, suggestive of inflammatory cells, around central veins at 72 hours, while chronic injection increases the density of cells between portal veins ( Figure 4B, C). As shown in Figure 4D, vehicle-injected Pdgfra WT/nGFP mice have an even distribution of PDGFRa-positive cells throughout the liver lobule. 72 hours after a single injection of CCl 4 , however, PDGFRapositive cells have a higher density around central veins at areas of hepatocyte injury ( Figure 4E). We found that after six weeks of twice weekly CCl 4 injections, PDGFRa positive cells are detected around and between portal veins, where fibrotic bands form ( Figure 4F). As PDGFRb is expressed in quiescent and activated HSCs [38], we next determined whether both PDGFRs are expressed in the same cell type. PDGFRa positive cells in chronic CCl 4 injected Pdgfra WT/nGFP mouse livers ( Figure 5A, D) co-localize with PDGFRb immunoreactive cells ( Figure 5B, E), indicating that activated HSCs express both receptors ( Figure 5C, F) after chronic CCl 4 injection.
Our data from human tissue using IHC ( Figure 1) and from mice using a GFP reporter ( Figure 5) indicated that HSCs are the predominant liver cell type that expresses PDGFRa. In order to confirm these findings, we stained for specific liver cell epitopes using IF in combination with nuclear-GFP expression in Pdgfra WT/nGFP mice [20]. Images of livers from Pdgfra WT/nGFP mice indicate that PDGFRa and PDGFRb co-localize in the same cells ( Figure 5, Figure S2A), and that these cells also express desmin ( Figure S2B) and cellular retinol binding protein 1 (CRBP-1, Figure S2C), proteins expressed in HSCs. Furthermore, GFP is not detected in Kupffer cells that express F4/80 expression ( Figure  S2D), or endothelial cells as identified by CD31 expression ( Figure  S2E). Hepatocytes, identified morphologically by fluorescence as described [39], were also negative for GFP ( Figure S2F). Taken together, these data suggest that HSCs are the primary liver cell type which express PDGFRa and PDGFRb.

Reducing PDGFRa Expression Reduces Fibrosis in Mice
After confirming that Pdgfra expression increases with CCl 4induced liver injury, and that PDGFRa-positive HSCs are activated by CCl 4 exposure, we utilized Pdgfra WT/nGFP mice to evaluate the functional significance of Pdgfra expression during liver fibrosis. Pdgfra WT/nGFP mice, which are heterozygous for Pdgfra expression, are phenotypically normal [20] and have normal liver architecture ( Figure 4A). Uninjured and CCl 4 -injected Pdgfra WT/nGFP and C57BL/6 mice were analyzed for transcriptional changes in genes associated with chronic liver injury. At baseline and after 4 weeks of twice weekly injections of CCl 4 , Pdgfra expression is decreased in Pdgfra WT/nGFP mice compared to C57BL/6 mice ( Figure 6A). Pdgfrb expression was equivalently expressed in both genotypes of uninjured and chronic CCl 4 injected mice ( Figure 6B). Expression of smooth muscle a-actin (Acta2), an epitope that is upregulated when HSCs are activated [40,41], increased when C57BL/6 mice were injected with CCl 4 for 4 weeks, but Pdgfra WT/nGFP mice had greatly reduced Acta2 expression after chronic liver injury ( Figure 6C). Fibrillar collagen 1a1 (Col1a1) expression was equivalent in uninjured mice of the two genotypes, but significantly reduced in chronically injured Pdgfra WT/nGFP mice ( Figure 6D). Conversely, collagen 4 (Col4, Figure 6E) and tissue inhibitor of metalloproteinase 1 (Timp1, Figure 6F) expression was increased after chronic CCl 4 exposure, but was not significantly reduced in Pdgfra WT/nGFP mice compared to wild type mice.
Reduced mRNA expression of Col1a1 in Pdgfra WT/nGFP mice after chronic CCl 4 injection was accompanied by a reduction in liver fibrosis, as assessed by picrosirius red staining, a histochemical assay for tissue fibrosis. C57BL/6 mice injected with vehicle for 4 weeks had little to no fibrosis ( Figure 7A, D), but developed periportal fibrosis after 4 weeks of twice weekly CCl 4 injections ( Figure 7B, D). Significantly less collagen was deposited in chronically injured Pdgfra WT/nGFP mice ( Figure 7C, D). These results demonstrate that liver fibrosis in response to chronic CCl 4 injection is dependent on normal expression of PDGFRa, and are consistent with the hypothesis that liver fibrosis is regulated in part by PDGFRa ligands.

Discussion
PDGFRs stimulate proliferation, migration, and survival of mesenchymal cells, and increased activation of PDGFRs leads to organ fibrosis [21,42]. Elevated expression of PDGFRs is associated with liver fibrosis and cirrhosis, so we sought to determine whether PDGFRa regulates liver fibrogenesis using mice that have one allele of Pdgfra (Pdgfra WT/nGFP ). We found that mice with decreased Pdgfra expression have less liver fibrosis after chronic CCl 4 injury. In addition, and consistent with the notion that PDGFRa regulates the liver's response to injury, patients with liver disease have elevated expression of PDGFRa and PDGFRb. In conjunction, GFP localization in Pdgfra WT/nGFP mice indicates that PDGFRa-positive HSCs migrate to sites of injury following CCl 4 injection. These data all suggest that PDGFRa is involved in the activation of HSCs after hepatocyte injury.
PDGFRs are thought to play a central role in activating HSCs and promoting liver fibrosis and cirrhosis [33,34,43]; whether PDGFRa and PDGFRb play independent roles in fibrogenesis is not known. We and others observe that Pdgfrb expression increases in WT mice after acute liver injury by CCl 4 , implicating PDGFRb in HSC activation. Thus it is surprising that mice which systemically express a hyperactive PDGFRb allele do not develop more liver fibrosis than WT mice after 4 weeks of CCl 4 injections [44]. Our data indicate that hepatocyte injury induces Pdgfra expression above uninjured liver in both mice and humans, corroborating previously published studies [34,37]. Our results demonstrate that expression of Pdgfra and Pdgfrb are both increased after chronic CCl 4 liver injury, while reducing Pdgfra copy number reduces Pdgfra expression but not Pdgfrb expression in Pdgfra WT/nGFP mice. Reduced Pdgfra expression in Pdgfra WT/nGFP mice corrrelates with significantly reduced Col1a1 and Acta2 expression, as well as reduced picrosirius red staining, even though Pdgfrb expression remains elevated. PDGFRa and PDGFRb appear to affect HSCs differentially, despite being co-localized in the same liver cell type. Further studies will be necessary to dissect the receptor-specific contributions of PDGF signaling pathways in HSCs and in liver fibrosis.
Small perturbations in the PDGF signaling pathway, whether due to changes in expression of ligand or receptor, appear to have a large impact on specific diseases. Support for this notion is found in genetic evidence from rodents, which suggests that small changes in PDGFR activity in vivo are capable of significantly affecting a cell's function. For example in development, chimerism studies show that both Pdgfrb +/2 and Pdgfra +/2 embryonic stem cells are deficient in contributing cells to the embryo [15,45], and adult mice have a decreased number of progenitor cells in mice heterozygous Pdgfrs [46,47]. Heterozygous Pdgfra mice have been bred to mice with mutations in PDGF ligands [48] or mutations in immediate early genes directly downstream of PDGFRa [49], resulting in additive effects. However, deletion of one allele of Pdgfra and the resultant heterozygosity does not affect development [15,20]. These studies suggest that a single copy of Pdgfra is usually sufficient for development, although under certain circumstances two alleles of Pdgfra are required. In the current study, we found that Pdgfra WT/nGFP mice have reduced fibrosis and reduced expression of the profibrotic genes Acta2 and Col1a1 after chronic CCl 4 injury. Our data indicate that in chronic liver injury, PDGFRa plays a critical role in the development of fibrosis, but that other pathways also contribute to fibrogenesis. Expression of Col4 and Timp1 were reduced in Pdgfra WT/nGFP mice but not to a significant extent, suggesting that expression of these genes could be more reliant on PDGFRa independent pathways, or heterogeneity in populations of HSCs.
We also sought to better define the role of PDGFRa in liver fibrosis by utilizing both human specimens and mouse models. Using a variety of experimental approaches, increased PDGFRa was seen in cirrhotic human livers and in mice with chemicallyinduced liver fibrosis. Although no preclinical rodent model fully recapitulates human liver fibrosis, there appears to be comparable molecular pathophysiology between humans and mice. We chose to utilize a knock-in mouse model expressing nuclear-GFP driven by the Pdgfra promoter in order to discriminate between cells located in close proximity to each other, specifically different NPC populations in liver sinusoids [20]. We did not observe nuclear-GFP expression in hepatocytes, Kupffer cells, or LSECs, thus we conclude that the majority of PDGFRa is expressed in HSCs in the mouse liver, consistent with our observation that human liver specimens express PDGFRa primarily in NPCs. Our IHC data are further supported by data from the Human Protein Atlas (http://www.proteinatlas.org/ENSG00000134853/cancer), which demonstrates that NPCs are positive for PDGFRa by IHC in both normal liver and HCC [50].
We and others posit that selectively targeting PDGFRa in liver fibrosis and cirrhosis could reduce the proliferation, migration, and survival of the activated HSCs cells that contribute to collagen deposition. Therapeutic blockade of PDGFRa signaling may have a broad impact in the treatment of liver fibrosis, as four of the five PDGF ligand dimers, PDGF-AA, -AB, -BB, and -CC, bind and activate PDGFRa [12]. Targeting PDGFRb, on the other hand, would completely inhibit only signal transduction induced by PDGF-DD, and could disrupt necessary functions of PDGFRb in the liver. Targeting both PDGFRs with multi-kinase inhibitors, such as imatinib or sorafenib, leads to severe off target effects [51,52]. The breadth of multi-kinase inhibitor activity thus likely leads to inhibition of beneficial signal transduction, either via PDGFRb or other kinases. In summary, our data suggest that PDGFRa has a specific role in liver fibrosis in mice and in humans, and suggest that further mechanistic evaluation of PDGFRa function in the liver has the potential to uncover new anti-fibrotic therapies.  Table S4 Summary of PDGFRa and PDGFRb immunoreactivity in human liver specimens. Resected liver specimens with HCCs were formalin-fixed, paraffin embedded, and evaluated for the presence of cirrhosis and HCC. IHC for PDGFRa and PDGFRb was performed as described in Materials and Methods. Relative staining intensity is indicated as weak (+), moderate (++), strong (+++), or absent (0). (DOCX )   Table S5 Immunoblot detection of PDGFR expression in macroscopically dissected human tumors and surrounding liver. HCCs (Tumor) and surrounding liver (Non-Tumor) were macrodissected from patients, frozen, and processed for immunoblot analysis as described in Materials and Methods. Intensity is indicated as present (+) or absent (0). (DOCX) Table S6 PDGF stimulates proliferation 1 in stellate cell lines, but not primary hepatocytes or hepatoma cell lines. 1 Cell proliferation was measured by DNA synthesis using tritiated thymidine incorporation [28]. The data is the average of three different experiments that were each done in triplicates. Fold change represents the increase when compared to unstimulated cells for each cell line. 2 '' Positive control'' indicates that DNA synthesis was stimulated in each cell line or primary culture with a growth factor previously reported to simulate proliferation. Growth factors used for each cell and the concentrations are as follows: mouse hepatocytes, EGF (20 ng/mL); AML12 cells, EGF (20 ng/mL); NMH cells, HB-EGF (20 ng/mL); rat stellate cells (2G), 1% fetal calf serum; human stellate cells (LX-2), 1% FCS; SK-Hep (human hepatoma cells of endothelial origin), 10% fetal calf sera. (DOCX)