Thymus cell antigen-1-expressing cells in the oval cell compartment


  • Potential conflict of interest: Nothing to report.


Thymus cell antigen-1 (Thy-1)-expressing cells proliferate in the liver during oval cell (OC)-mediated liver regeneration. We characterized these cells in normal liver, in carbon tetrachloride-injured liver, and in several models of OC activation. The gene expression analyses were performed using reverse-transcriptase polymerase chain reaction (RT-PCR), quantitative RT-PCR (Q-RT-PCR) of cells isolated by fluorescence-activated cell sorting (FACS), and by immunofluorescent microscopy of tissue sections and isolated cells. In normal liver, Thy-1+ cells are a heterogeneous population: those located in the periportal region do not coexpress desmin or alpha smooth muscle actin (α-SMA). The majority of Thy-1+ cells located at the lobular interface and in the parenchyma coexpress desmin but not α-SMA, i.e., they are not resident myofibroblasts. Although Thy-1+ cells proliferate moderately after carbon tetrachloride injury, in all models of OC-mediated liver regeneration they proliferate quickly and expand significantly and disappear from the liver when the OC response subsides. Activated Thy-1+ cells do not express OC genes but they express genes known to be expressed in mesenchymal stem cells (CD105, CD73, CD29), genes considered specific for activated stellate cells (desmin, collagen I-a2, Mmp2, Mmp14) and myofibroblasts (α-SMA, fibulin-2), as well as growth factors and cytokines (Hgf, Tweak, IL-1b, IL-6, IL-15) that can affect OC growth. Activated in vitro stellate cells do not express Thy-1. Subcloning of Thy-1+ cells from OC-activated livers yield Thy-1+ fibroblastic cells and a population of E-cadherin+ mesenchymal cells that gradually discontinue expression of Thy-1 and begin to express cytokeratins. However, upon transplantation these cells do not differentiate into hepatocytes or cholangiocytes. Activated Thy-1+ cells produce predominantly latent transforming growth factor beta. Conclusion: Thy-1+ cells in the OC niche are activated mesenchymal-epithelial cells that are distinct from resident stellate cells, myofibroblasts, and oval cells. (HEPATOLOGY 2009.)

Over the years, substantial evidence has accumulated demonstrating the existence of adult hepatic progenitor cells, also termed oval cells (OCs). When the regenerative capacity of terminally differentiated hepatocytes is exhausted or blocked, these cells are activated to proliferate and differentiate into hepatocytes and cholangiocytes. OCs sprout from the putative stem cell niche (canals of Hering), forming tortuous pseudoducts and invading the liver lobule.1–6 These pseudoducts are in close proximity to desmin-positive stellate cells.7 Recently, we found a population of activated mesenchymal cells, thymus cell antigen-1 (Thy-1)-expressing cells, surrounding OCs. These cells partially overlapped with the desmin-positive cells and produce inductive signals (growth factors and cytokines) in the OC niche.8, 9

Thy-1 is a cell surface glycophosphatidylinositol-linked glycoprotein with a molecular mass of 35 kDa and is an adhesion molecule of the immunoglobulin superfamily.10 In mice and rats, Thy-1 is expressed in the brain, on thymocytes, T lymphocytes, fibroblasts, epidermal cells, and a small population of bone marrow cells. Thy-1 is involved in T-cell activation and affects numerous nonimmunologic biological processes, such as cellular adhesion, neurite outgrowth, tumor growth, migration, and cell death.10 Because OC and Thy-1-expressing cells are in very close contact, initially Thy-1 was considered a marker of hepatic OC.11–14 However, recently it was reported that in a rat liver regeneration model, following 2-acetylaminofluorene treatment and partial hepatectomy (2-AAF/PH), CK-19+ OCs do not express Thy-1 and that Thy-1 is a marker of hepatic myofibroblasts.15 That Thy-1 is expressed in hepatic myofibroblasts/stellate cells and not in OC was also confirmed by another group, although, coexpression of Thy-1 and desmin in these cells was not found.16 Thy-1 expression was also detected in cell lines of human fetal hepatoblasts,17, 18 in a stem-like population (distinct from OCs) derived from adult human liver19 and in hepatocellular carcinoma (HCC) cell lines.20 All these findings pose the question: What is the nature of Thy-1+ cells?

In the present work we studied the activation of Thy-1-expressing cells in OC-mediated liver regeneration and in carbon tetrachloride (CCl4)-injured liver. Our results show that Thy-1+ cells are a desmin+ mesenchymal-epithelial population in the liver, distinct from the resident myofibroblasts and portal fibroblasts. They are rarely detected in the normal rat liver. Thy-1+ cells are activated and expand in parallel with OCs; they express alpha smooth muscle actin (α-SMA), produce growth factors, cytokines, and extracellular matrix components, and disappear when liver injury is resolved.


2-AAF, 2-acetylaminofluorene; α-SMA, alpha smooth muscle actin; CDE, choline deficient, ethionine supplemented; D-gal, D-galactosamine; EpCAM, epithelial cell adhesion molecule; FACS, fluorescence-activated cell sorting; Fsp1, fibroblast-specific protein 1; GAPDH, glyceraldehyde-3 phosphate dehydrogenase; HCC, hepatocellular carcinoma; IF, immunofluorescent; OC, oval cell; PDGF, platelet-derived growth factor; PH, partial hepatectomy; Rs/PH, retrorsine/partial hepatectomy; RT-PCR, reverse-transcriptase polymerase chain reaction; Thy-1, thymus cell antigen-1; TGF-β, transforming growth factor beta.

Materials and Methods

Animals and Animal Treatment.

Male Fisher 344 rats (180 g and 300–350 g) were purchased from Taconic Farms (Germantown, NY). All studies with animals were conducted under a protocol approved by the Animal Care Institute of the Albert Einstein College of Medicine in accordance with National Institutes of Health (NIH) guidelines. 2-AAF pellets (35 mg/tablet; 14 day release) were purchased from Innovative Research of America (Sarasota, FL). Retrorsine, D-galactosamine (D-gal), and CCl4 were purchased from Sigma (St. Louis, MO). A choline-deficient diet (catalog no. 960210) was obtained from MP Biomedicals (Solon, OH).

The liver injury models, cell isolation and culture, fluorescence-activated cell sorting (FACS) analysis and sorting, Thy-1+ cell subcloning and determination of active and total TGF-β in activated Thy-1+ clones are described in the Supporting Material.

Antibodies and Immunofluorescent (IF) Analysis.

Primary antibodies are listed in Supporting Table 1. All secondary antibodies were purchased from Jackson Immunoresearch Laboratories (West Grove, PA) and used at a dilution of 1:100. Five-micrometer frozen sections, cytospins, or cells on chamber slides were fixed for 10 minutes in ice-cold methanol, washed twice with phosphate-buffered saline (PBS), and blocked with 5% normal serum from the animal in which the secondary antibody was raised and 1% bovine serum albumin. Incubation with the antibodies and IF was performed as described.9, 23

Reverse-Transcriptase Polymerase Chain Reaction (RT-PCR), Quantitative PCR.

All reverse-transcriptase reactions were carried out with SuperScript II Reverse Transcriptase (Invitrogen, Carlsbad, CA), according to the manufacturer's protocol. Rat specific primers for different genes were chosen with the Primer3 program. The expression level of glyceraldehyde-3 phosphate dehydrogenase (GAPDH) was used as an internal control.

Quantitative real-time RT-PCR was performed in three repeats using the ABI-Prism 7900 Sequence Detection System (Applied Biosystems, Foster City, CA). Rat-specific sequences for PCR primers (presented in Supporting Table 2) were designed to generate amplicons of 50 to 150 base pairs, required for quantitative real-time detection, using SYBR Green Master Mix (Applied Biosystems). Messenger RNA (mRNA) abundance was determined by normalization of the data to the expression levels of GAPDH mRNA. The mRNA expression level in normal liver (NL) was taken as a baseline and considered equal to 1.


Expression of Thy-1 in Normal Liver.

In normal livers, few Thy-1-expressing cells could be detected in the periportal region. Most of these cells do not express desmin (Fig. 1A-C). However, at the lobular interface there are Thy-1+ cells, which coexpress desmin (Fig. 1A-C, small arrows). Also, some Thy-1+ cells coexpressing desmin could be identified in the liver parenchyma (Fig. 1D-F). Coexpression of Thy-1 and fibroblast-specific protein 1 (Fsp1)21 (Fig. 1G-I) and Thy-1 and α-SMA (Fig. 1J-L) was not detected.

Figure 1.

Expression of Thy-1 in normal liver. Double IF labeling for Thy-1 (green) and desmin (red): (A-C) Periportal zone and (D-F) liver lobule. The majority of desmin+ cells surrounding blood vessels are Thy-1-negative. Double-labeled cells are visible in the periportal zone interfacing the liver lobule and in the lobule itself (arrows). Double IF labeling of Thy-1 (green) and Fsp1 (red): (G-I) no coexpression of the two antigens in periductular cells. Double IF labeling for Thy-1 (green) and α-SMA (red): (J-L) periductular expression of the two markers but no colocalization. Original magnification: ×200. Coexpression of the three markers is clearly visible with individual cells collected on cytospins (M): Triple labeling of two cells: 1: the right one is Thy-1+ (green), the left one is Thy-1; 2: both cells are desmin+ (red); 3: the right one is positive for Thy-1 and desmin (yellow); 4: the same two cells labeled with α-SMA Ab, only the left one is α-SMA+ (pink) and also desmin+; whereas in 5 the right one is Thy-1+ (green) but α-SMA. Original magnification: ×400.

Because Thy-1 is a surface antigen and desmin and α-SMA are intracellular filaments, coexpression of these three molecules is not always obvious on liver sections. This explains why other researchers were not able to demonstrate it.15, 16 We analyzed their expression on cytospins and found that many Thy-1+ cells coexpressed desmin but none of them coexpressed α-SMA. These findings were confirmed by triple IF microscopy showing coexpression of Thy-1 and desmin and the lack of coexpression of Thy-1 and α-SMA (Fig. 1M). The number of double-labeled Thy-1/desmin and Thy-1/α-SMA cells was quantitated. Approximately 85% of the Thy-1+ cells expressed desmin, but none of them were positive for α-SMA. On the other hand, only 4% of all desmin-positive cells coexpressed Thy-1. Our data reveal that there are two populations of Thy-1+ cells in the quiescent liver: one population located in the periportal region that does not express desmin and a second population located at the lobular interface and parenchyma that coexpresses desmin. In addition, in normal liver Thy-1+ cells do not coexpress α-SMA; thus, they are not resident portal myofibroblasts.

Expression of Thy-1 After Acute CCl4 Liver Injury.

To understand the role of Thy-1+ cells in acute liver injury, we studied the appearance of these cells after CCl4 treatment. CCl4 injury causes severe pericentral hepatocyte necrosis with very little activation of OC.22 This is clearly demonstrated in Fig. 2A-C, in which there is a modest increase in Thy-1-expressing cells in the periportal region, but no concomitant claudin-7 expression, an OC marker23 (Fig. 2B). Figure 2C depicts the same section stained with hematoxylin and eosin (H&E), where the necrotic pericentral region is marked. In contrast to the modest increase in the number of Thy-1-expressing cells, there is a massive increase in desmin+ and α-SMA+ activated stellate cells/myofibroblasts (Fig. 2D-I). These results were confirmed by Q-RT-PCR analysis (Fig. 3A), which demonstrates a marked increase in desmin+ and α-SMA+ cells. The modest proliferation of Thy-1+ cells and the high proliferation of desmin+ and α-SMA+ cells in the CCl4 injury model indicates that Thy-1+ cells are distinct from activated stellate cells and myofibroblasts.

Figure 2.

Expression of Thy-1 in the liver of CCl4-treated rats. Frozen liver sections from normal and CCl4-treated rats on Day 4. Double IF labeling of Thy-1+ cells (red) and claudin-7+ OCs (green): normal liver (A); CCl4-injured liver (B); liver histology of the same section given in B (C); marked are the necrotic pericentral regions. Double IF labeling for Thy-1 (green) and α-SMA (red) (D-F): (F) merged picture of Thy-1 and α-SMA labels; double IF labeling for Thy-1 (green) and desmin (red) (G-I); (I) merged picture of Thy-1 and desmin labels. Note the low number of Thy-1+ cells in (D) and (G), the high number of α-SMA-positive cells in (E), and the massive increase of desmin-positive cells in H in the same regions as those presented in (D) and (G), respectively. Original magnification: ×200.

Figure 3.

RT-PCR analysis of Thy-1, α-SMA, desmin, and claudin-7 mRNA expression. (A) Q-RT-PCR analysis of mRNA expression in the livers of normal, CCl4, 2AAF-PH, and Rs/PH-treated rats. Note the lower activation of desmin+ and α-SMA+ cells in OC models (2AAF-PH and Rs/PH) compared to the activation of Thy-1+ cells. (B) Kinetics of increase and decrease of Thy-1 and claudin-7 mRNA expression in the livers of 2AAF-PH-treated rats. (C) FACS analysis of the major cellular components of the OC compartment. The cells from the 13/15% Nycodenz interface fraction were labeled with antibodies detecting OCs (EpCAM), Thy-1 cells (CD90), Kupffer cells (CD68), and endothelial cells (CD31).

Activation of Thy-1+ Cells in OC-Mediated Liver Regeneration.

Subsequent studies were performed to confirm that Thy-1+ cells proliferate predominantly in models of OC expansion. In all models of OC activation: D-galactosamine (D-gal) liver injury: Fig. 4D-F; choline-deficient, ethionine-supplemented (CDE) diet: Fig. 4G-I; 2-acetylfluorene/partial hepatectomy (2-AAF/PH): Fig. 4J-L; and retrorsine/partial hepatectomy (Rs/PH): Fig. 4M-O, a huge expansion of Thy-1+ cells intermingled with claudin-7+ OCs was observed. In the OC models, Q-RT-PCR analyses confirmed that activation of Thy-1-expressing cells was much higher than the activation of desmin and α-SMA-expressing cells (Fig. 3A).

Figure 4.

Proliferation of Thy-1+cells in OC-mediated liver regeneration. Double IF labeling of Thy-1+ cells (red) and claudin-7+ OCs (green). (A-C) Normal liver, 2 days after partial hepatectomy; (D-F) D-Gal model, 5 days after injection; (G-I) 3 weeks of CDE diet; (J-L) 2-AAF/PH model, 10 days post-PH; (M-O) Rs/PH model, 1 month post-PH. Merged pictures are presented in (C,F,I). Note the proliferation of Thy-1 cells in these livers, surrounding the OC pseudoducts. Highest accumulation of Thy-1 cells is observed in the Rs/PH model. Original magnification: ×200.

To determine the expansion of Thy-1+ cells in the 2-AAF/PH model of OC activation, we FACS analyzed the major cellular components located in the OC compartment.9 Nonparenchymal cells from normal and 2AAF/PH livers were isolated, fractionated, and the OC-enriched fraction (13%-15% OptiPrep) analyzed (see Supporting Materials and Methods). For FACS analysis, sinusoidal cells were labeled with antibody against CD31, Kupffer cells with antibody against CD68, Thy-1+ cells with Thy-1 antibody, and OC and cholangiocytes with epithelial cell adhesion molecule (EpCAM) antibody. EpCAM is expressed on the surface of small cholangiocytes and OC.23 As shown in Fig. 3C, the percentage of different cell populations in the OC compartment is ≈20% sinusoidal, 18% Kupffer, 21% EpCAM+, and 23% Thy-1+. In contrast, the percentage of EpCAM+ and Thy-1+ cells in normal liver was below 5% (data not shown).

Thy-1+ Cells Appear and Disappear in Parallel with OCs.

If Thy-1+ cells function to support OC in the liver, their number should increase and decrease simultaneously with OC. To test this notion we followed the expression of Thy-1+ cells in the 2-AAF/PH liver. In this model the highest accumulation of OC and Thy-1+ cells is observed around Day 10 after PH, after which their number decreases gradually (Fig. 5A-F). Thirty days after PH, the histology of the liver returns to normal and Thy-1+ cells disappear (Fig. 5F). The kinetics of proliferation of Thy-1+ and claudin-7+ cells, determined by counting double-labeled Thy-1/Ki-67 and claudin-7/Ki-67 cells, are presented in Fig. 5G. Thy-1+ cells proliferate early after PH (the peak is on Day 2) and become quiescent after 30 days. That Thy-1+ cells and OC disappear from the liver after 30 days was also confirmed by RT/PCR analysis for Thy-1 and claudin-7 mRNAs (shown in Fig. 3B).

Figure 5.

Appearance and disappearance of Thy-1+ cells and OCs in the livers of 2-AAF/PH-treated rats. (A) 2 days, (B) 4 days, (C) 6 days, (D) 10 days, (E) 20 days, and (F) 30 days post-PH. Thy-1 (red); OC (claudin-7, green). The highest increase in number of both cell types is observed on Day 10. Thirty days after PH their number returned to the normal. Original magnification: ×200. (G) Proliferation of Thy-1+ cells and OCs (claudin-7+) in 2-AAF/PH liver at different timepoints after PH, determined by counting the double-labeled Thy-1+/Ki-67+ and claudin-7+/Ki-67+ cells on cytospins of nonparenchymal cells.

These data demonstrate that Thy-1+cells are a dynamic cell population. They proliferate early after 2-AAF/PH and disappear when liver injury is resolved. Furthermore, proliferation of OC occurs in conjunction with Thy-1+ cell activation.

Activated Thy-1+ Cells Express α-SMA.

It is generally accepted that upon activation stellate cells transdifferentiate into myofibroblasts and express desmin and α-SMA.24–26 Recent research has also shown that hepatic myofibroblasts may have different sources and are represented by several subpopulations.27–29 A population of hepatic myofibroblasts distinct from stellate cells has been reported to be present in the stellate cell-enriched fraction.30 We found that in normal liver most Thy-1+ cells are desmin+ but α-SMA, i.e., they are not resident myofibroblasts. To determine the changes in the phenotype of these Thy-1+/desmin+ cells in OC-mediated liver regeneration, frozen sections from three categories of livers (control, and 10 days and 30 days after 2-AAF/PH) were subjected to triple labeling: Thy-1 (green), α-SMA (blue), and desmin (red), (Fig. 6A-C). In normal livers we did not detect triple-labeled cells. Ten days post-PH, 60% of all Thy-1+ cells were triple-labeled. Thirty days after PH, the triple-labeled cells were gone. In addition, we isolated the Thy-1+-enriched cell fraction by Nycodenz gradient centrifugation (see Supporting Materials and Methods) from the livers of 2-AAF-treated rats on Day 10 after PH. Representative photomicrographs of triple-labeled cells 10 days post-PH are presented in Fig. 6D-I. The double-labeled cells Thy-1/desmin and Thy-1/α-SMA were quantified; ≈95% of the Thy-1+ cells coexpressed desmin (which is close to what we observed in normal liver) and 75% of them were α-SMA+. On the other hand, 60% of all desmin+ cells were Thy-1-negative and 30% of all α-SMA+ cells did not express Thy-1.

Figure 6.

Expression of Thy-1, desmin, and α-SMA in OC-mediated liver regeneration. Triple labeling for Thy-1 (green), α-SMA (blue), and desmin (red) of tissue sections from normal and 2-AAF/PH livers: (A) normal liver; (B) 2-AAF/PH, Day 10; (C) 2-AAF/PH, Day 30. There are no triple-positive cells in normal liver or 30 days post-PH. In contrast, 10 day after PH, 60% of the cells are triple-positive for Thy-1, desmin, and α-SMA. Original magnification: ×200. Triple labeling of Thy-1 enriched cell fraction on cytospins. The cells are labeled for: (D) Thy-1 (green); (E) desmin (red); and (F) α-SMA (pink); (G) Thy-1 (green) and desmin (red); (H) Thy-1 (green) and α-SMA (pink); (I) Thy-1 (green), desmin (red), and α-SMA (pink). Arrowheads point to double-labeled cells and arrows to triple-labeled cells; Three of the presented cells coexpress Thy-1, desmin, and α-SMA. Original magnification: ×400.

These data demonstrate that, whereas in normal liver Thy-1+ cells express desmin, proliferating Thy-1 cells in 2-AAF/PH liver also express α-SMA; however, only a subpopulation of activated desmin+ and α-SMA+ cells coexpress Thy-1. This finding indicates that Thy-1+ cells are a distinct subpopulation of activated mesenchymal cells in OC-mediated liver regeneration.

Activation of Thy-1+ Cells In Vitro.

The above data showed that Thy-1+ cells actively proliferate in several OC models of liver regeneration and acquire expression of α-SMA. We studied further the activation of Thy-1+ cells in vitro. For this purpose we isolated a stellate cell-enriched fraction (see Supporting Materials and Methods), with a purity ranging from 85% to 95%, as judged by vitamin A fluorescence. The cells were plated on uncoated dishes with 20% fetal bovine serum (FBS). The autofluorescent cells on the dish were labeled. Thy-1, desmin, and α-SMA gene expression by these cells was followed over the course of 8-12 days. After 1 week in culture the labeled quiescent stellate cells lost their autofluorescence and expressed desmin and α-SMA, but not Thy-1 (Fig. 7A). On the other hand, Q-RT-PCR analysis of all cells plated on the dish showed an increase in all three studied markers, suggesting that other mesenchymal cells present in this fraction were activated (Fig. 7B). The increase in Thy-1 expression was also confirmed by FACS analysis (Fig. 7C). In a parallel experiment, the cells were labeled with Thy-1/desmin and Thy-1/α-SMA antibodies (Fig. 7D) and also with Thy-1/Ki-67 and desmin/Ki-67 antibodies (Fig. 7E) to assess their proliferative activity. In normal livers, 4% of all desmin-positive cells expressed Thy-1 and 86% of all Thy-1+ cells coexpressed desmin; none of them coexpressed α-SMA (Fig. 7D). However, after 5 days in vitro, more than 90% of Thy-1+ cells began to express α-SMA and became triple-positive (Fig. 7D). The proliferation of Thy-1+ and desmin+ cells has similar kinetics and 12 days after plating they became confluent (Fig. 7E). It should be noted that only 28% of the desmin+ and α-SMA+ cells coexpressed Thy-1 (Fig. 7F), whereas all α-SMA+ cells expressed desmin (data not shown). These data are in accord with the in vivo results for Thy-1+ cells (Fig. 6B,F); they demonstrate again that Thy-1+ cells are a distinct desmin+ mesenchymal subpopulation, and when activated they express α-SMA. It is unlikely that these cells originate through transdifferentiation of quiescent stellate cells as judged by our in vitro studies.

Figure 7.

Activation of Thy-1+ cells in vitro. Activation of quiescent stellate cells in vitro. (A) Purified autofluorescent stellate cells were plated on uncoated dishes and activated with 20% serum. After 1 week the cells were labeled with Thy-1, α-SMA, and desmin antibodies and the positive cells were counted. (a,b) Labeling for desmin (red) and Thy-1 (green); (c,d) labeling for α-SMA (red) and Thy-1 (green). No double-labeled cells were detected; only 3% of all plated cells expressed Thy-1. In vitro activation of hepatic mesenchymal cells from normal liver (11%/13% Nycodenz gradient fraction). (B) Increase in Thy-1, desmin, and α-SMA mRNA expression analyzed by Q-RT-PCR; (C) Increase in the number of Thy-1 expressing cells determined by FACS analysis. (D) Double labeling of Thy-1/desmin and Thy-1/α-SMA-positive cells. In normal liver, Thy-1+ cells do not express α-SMA but 86% coexpress desmin; after 5 days of activation more than 90% of the Thy-1+ cells coexpress desmin and α-SMA. (E) Double labeling of Thy-1/Ki-67 and desmin/Ki-67-positive cells. Thy-1+ and desmin+ cells display the same pattern of proliferation. (F) After 12 days in culture 28%–30% of desmin+ and α-SMA+ cells coexpress Thy-1+.

Gene Expression Analysis of Thy-1+ Cells in the OC Niche.

To determine the phenotypic differences of Thy-1-expressing cells, we used FACS to purify Thy-1+ cells from 2-AAF-treated rats on Day 10 after PH (Fig. 8A). Gene expression analysis of purified Thy-1+ cells (Fig. 8B) confirm our previous data that activated Thy-1+ cells isolated from OC-activated livers express genes known to be expressed in mesenchymal stem cells (CD105, CD73, CD29), genes considered specific for activated stellate cells (desmin, collagen I-a2, Mmp2, Mmp14), genes expressed by myofibroblasts (α-SMA, fibulin-2), and growth factors and cytokines (Hgf, Tweak, IL-1b, IL-6, IL-15) (Fig. 8B).9

Figure 8.

Gene expression analysis of purified Thy-1-expressing cells. (A) Analysis of Thy-1+ cells isolated from 2-AAF/PH-treated rats on Day 10 after PH. Cells from the 11%/13% Nycodenz gradient fraction were plated overnight and the adherent cells separated by FACS. Post-sorted cells were analyzed with MoFlo cell sorter, using FlowJo software. (B) RT-PCR analysis of the isolated Thy-1+ cells after in vivo activation (OC-mediated liver regeneration, using the 2-AAF/PH model). Note that these cells do not express any markers for OCs, sinusoidal cells, or Kupffer cells. (C-E) Analysis of Thy-1+ cells propagated and subcloned in vitro. Phase contrast microscopy for the two Thy-1+ clones: (C) 2–6 and (D) 13–4, showing different morphology. Original magnification ×200. (E) Gene expression analysis of four Thy-1+ clones studied by RT-PCR. Thy-1+ clones established from isolated Thy-1 cells were additionally subcloned and propagated. Their gene expression was studied again at passage 10, using equal amounts of cDNA, specific gene primers, and 32 cycles.

It has to be noted that two markers of portal fibroblasts, CD39L (ectonucleoside triphosphate diphosphohydrolase-like 1) and elastin, are expressed at very low levels in Thy-1+ cells. These two genes are considered markers for portal fibroblasts.31, 32 This result shows that proliferating Thy-1+ cells in the rat OC-mediated liver regeneration model represent a Thy-1+/desmin+ mesenchymal subpopulation, which is distinct from portal fibroblasts.

Nature of In Vitro Propagated Thy-1 Cells Isolated from OC-Activated Livers.

Further purification and propagation of the Thy-1+ clones revealed changes in their gene expression and morphology. Although all cells continued to express Thy-1, they separated into two major groups that differed in morphology and gene expression. Clone 2 and its subclones (e.g., 2–6, presented in Fig. 8C) manifested as spindle-shaped fibroblasts; clone 13 and its subclones (e.g., 13–4, depicted in Fig. 8D) had a myofibroblastic appearance. A second noteworthy observation was that these subclones gradually discontinued expression of mRNAs for IL-1b and Hgf. Clones 13-1 and 13-4 expressed gp38, Vcam, CD73, Bmp6, and E-cadherin mRNA, whereas 2–6 and 2–8 did not (Fig. 8E). The third and most important observation was that several clones originating from clone 13 (13–11, 13–10), all of which initially expressed Thy-1, changed morphology after several cycles of freezing, thawing, and propagation in normal Dulbecco's Modified Eagle Medium (DMEM)/10% FBS (Supporting Fig. 1A,B), discontinued expression of Thy-1, decreased expression of α-SMA and desmin, and began to express epithelial cell markers: CK14 and CK18 but not any other OC-specific marker: AFP, claudin-7, EpCAM. Albumin, CD133, etc. (Supporting Fig. 1C). However, when these cells were activated to proliferate with platelet-derived growth factor (PDGF), and 20% serum, they again changed gene expression: discontinued expression of cytokeratins but continued to express E-cadherin, and began to express Thy-1 (Supporting Figs. 1C, 2A-D). These results show that the Thy-1+ cells behave like mesenchymal-epithelial cells: when activated, they turn down expression of epithelial cell markers, turn on expression of Thy-1, and appear morphologically and phenotypically as mesenchymal cells.

To study whether cells with dual mesenchymal-epithelial phenotype are present in normal liver, we isolated the nonparenchymal cell fraction and determined the presence of cells coexpressing CK18 and desmin. As anticipated from the above results, such cells were identified (Supporting Fig. 3). To study further whether the cells with dual mesenchymal-epithelial phenotype (clone 13-11-3-2) can differentiate in vivo into hepatocytes and cholangiocytes, we transplanted 5–106 cells into Rs/PH preconditioned DPP4-deficient rats (see Supporting Materials and Methods). One month later, the livers were examined for the presence of DPP4+ hepatocytes, cholangiocytes, or OCs. Such cells were not identified, demonstrating that the described hepatic mesenchymal-epithelial cells are not hepatic epithelial progenitor cells.

Thy-1 Expression Modulates the Synthesis of Latent TGF-β.

At present, the function of Thy-1 antigen is not known. It was reported that lung fibroblasts expressing Thy-1 modulate TGF-β activation.34 To determine whether Thy-1 antigen may play a similar role in activation of the OC compartment, we treated the mesenchymal clones that have discontinued expression of Thy-1 with 20% FBS and PDGF B polypeptides (BB). After 8 days of treatment the cells were further cultured for 4 days and 8 days, respectively, in DMEM/20% serum. Active and total TGF-β was determined in the conditioned medium using the plasminogen activator inhibitor-1 luciferase assay, as described in Supporting Materials and Methods. The results presented in Supporting Fig. 1D clearly show that “quiescent” Thy-1 mesenchymal 13-11-3-2 cells express 15 times more latent than active TGF-β. Furthermore, activated Thy-1+ cells express 30 times more latent TGF-β. These data demonstrate that Thy-1 mesenchymal cells produce predominantly latent TGF-β. They correlate very well with our gene expression analysis of isolated Thy-1+ cells using the rat Affymetrix 230-2 arrays chips: that in the OC-activated rat liver they produce 3 times more TGF-β binding proteins 1,2,3, and 4 than all other nonparenchymal cells (data not shown), which implies that when activated, Thy-1 cells synthesize predominantly latent TGF-β bound to TGF-β binding proteins.


Thy-1 (CD90) has been identified as a marker on almost all mesenchymal stem cells and those that can form fibroblast colonies in vitro (CFU-F).34, 35 Thy-1+ mesenchymal cells were also identified in rat fetal liver36 and it was reported that they support hepatic progenitor cell maturation in the mouse fetal liver.37, 38

Our studies show that Thy-1-expressing cells in rat liver are a heterogeneous population. In adult quiescent rat liver, Thy-1+, desmin cells were previously localized in the portal area around small bile ducts.15, 16 However, in our studies we found that these cells do not express α-SMA, i.e., they are not endogenous myofibroblasts. In a more thorough analysis of Thy-1-expressing cells in normal liver, we identified a second population of Thy-1-expressing cells located at the lobular interface and in the parenchyma that are desmin+. This population of Thy-1-expressing cells is rare, as only 4% of all desmin-positive cells express Thy-1. After in vivo activations, Thy-1-expressing cells from 2-AAF/PH-treated livers were isolated, subcloned, and propagated for an extended period in culture. They segregated into two major populations: one fibroblastic (Thy-1, α-SMA, and desmin-positive) and a second one of mesenchymal-epithelial phenotype (expressing in addition gp38, Vcam, CD73 Bmp6, and E-cadherin). After longer propagation under normal standard conditions, these cells turned off expression of Thy-1 but continued to express lower levels of α-SMA and desmin, and gained expression of cytokeratins. The mesenchymal-epithelial nature of these cells prompted us to transplant them into Rs/PH-preconditioned liver. However, they did not differentiate into hepatocytes or cholangiocytes, proving that these mesenchymal-epithelial cells are not hepatic epithelial progenitor cells. “Quiescent” Thy-1 cells possess unusual plasticity, as they can be reactivated to reexpress Thy-1 and lose expression of cytokeratins, although, E-cadherin continued to be expressed. The plasticity of these cells explains the recently reported data that pluripotent fetal and adult human mesenchymal, stem-like cells (but not OCs) express Thy-1.18, 19 It seems also that activated HCC cells, expressing Thy-1, are of a much more aggressive phenotype, capable of developing metastasis.20

Activated hepatic mesenchymal cells synthesize high levels of TGF-β. TGF-β is a multifunctional growth factor with antiproliferative and apoptotic effects on hepatocytes and OC.39 In the context of our work, we studied the relationship between Thy-1 expression and production of latent TGF-β by activated mesenchymal-epithelial cells. Our data show that, although highly activated, Thy-1+ cells produce predominantly latent TGF-β, which cannot inhibit OC proliferation, whereas the other growth factors and cytokines produced by Thy-1 cells efficiently activate OC growth (see Fig. 8). This notion is confirmed by our findings that feeder layers of Thy-1 cells maintain OC growth in vitro (data not shown).

It is unlikely that activated Thy-1-expressing cells are transdifferentiated stellate cells. Our in vitro studies showed that upon activation quiescent stellate cells acquire desmin and α-SMA, but not Thy-1 expression. It should be noted that both our in vivo and in vitro analyses demonstrate that Thy-1+-expressing cells are a subpopulation of all desmin+ and α-SMA+ cells in the liver, indicating that they are a distinct subpopulation of mesenchymal cells.

It is also not likely that Thy-1+ proliferating cells in the OC activation models are activated portal fibroblasts. The periductular Thy-1+ cells are described as desmin-negative,15, 16 whereas Thy-1+ cells in quiescent livers and in the OC compartment express desmin. Thy-1+ cells do not express CD39L1 and elastin, both of which are markers for portal fibroblasts.31, 32 In addition, Thy-1+ cells propagated in vitro continue to express desmin after more than 10-15 passages, whereas activated portal fibroblasts discontinue expression of desmin in vitro.37 It is also unlikely that Thy-1+ cells are fibroblasts from bone marrow. It was reported that the latter are CD45+,40 whereas hepatic Thy-1+ cells are CD45.

In this study we identified a population of hepatic mesenchymal-epithelial cells in the liver that is activated during OC-mediated liver regeneration. These cells express Thy-1 and genes specific for mesenchymal stem cells and for activated myofibroblasts/stellate cells. Thy-1 cells proliferate, expand in the OC niche, and produce growth factors, cytokines, and extracellular matrix components, all of which can support OC growth. Thy-1+ cells are not resident myofibroblasts, activated fat-storing stellate cells, portal fibroblasts, or fibroblasts originating from bone marrow and they are not hepatic epithelial progenitors. These cells exhibit high plasticity: when propagated in vitro under conditions conferring a quiescent state, Thy-1+ cells discontinue expression of Thy-1 and begin to express epithelial markers (CK14 and CK18) along with E-cadherin, switching to partially epithelial phenotype. These cells can be reactivated with serum and PDGF to resume Thy-1 expression. We propose that activated hepatic mesenchymal-epithelial cells support OC growth.


The authors are grateful to Dr. D.B. Rifkin (New York Medical Center, NY) for the generous gift of PAI-1 promoter luciferase reporter system, Dr. E.G. Neilson (Vanderbilt University School of Medicine, Nashville, TN) for Fsp1 antibody, Dr. R. Morris (King's College, London) for Thy-1 antibody, Drs. D.A. Shafritz and J. Locker (AECOM) for helpful discussion, and Ethel Hurston for technical assistance.