Hepatocyte Nuclear Factor-1 as Marker of Epithelial Phenotype Reveals Marrow-Derived Hepatocytes, but Not Duct Cells, After Liver Injury in Mice

Abstract

The potential bone marrow origin of hepatocytes, cholangiocytes, and ductal progenitor cells in the liver was examined in female mice after transplantation of bone marrow cells from male green fluorescent protein (GFP) transgenic donors. Following stable hematopoietic engraftment, the livers of the recipients were injured with carbon tetrachloride (CCl₄, with or without local irradiation of the liver) or 3,5-diethoxycarbonyl-1,4-dihydrocollidine (DDC, with or without local irradiation of the liver). The presence of numerous marrow-derived, GFP-positive inflammatory cells had the potential to lead to erroneous interpretation of marrow-derived hepatocytes, cholangiocytes, and ductal progenitor cells. Identification of marrow-derived ductal progenitor or cholangiocyte phenotype using colocalization of GFP or Y chromosome with pancytokeratin staining also failed to distinguish epithelial cells from closely apposed inflammatory cells. To address this inadequacy, we developed a rigorous new immunofluorescence protocol to identify marrow-derived epithelial cells in the liver using Y chromosome (donor marker) and hepatocyte nuclear factor-1 (HNF1, a nuclear marker of liver epithelial, nonhematopoietic phenotype). Using the Y/HNF1 method, rare (approximately one in 20,000) hepatocytes in female mice transplanted with male bone marrow contained a donor-derived Y chromosome. On the other hand, no Y chromosomes were found in cholangiocytes or ductal progenitor cells in mice with liver injury due to DDC or CCl₄. The use of a nuclear marker of mature hepatocytes or cholangiocytes, such as HNF1, improves discrimination of marrow-derived epithelial cells in tissue sections.

Introduction

The premise that differentiation of bone marrow-derived cells is restricted to the hematopoietic lineages, established over the past 50 years [1–3], has been challenged [4, 5]. Transplantation studies demonstrate the potential conversion of bone marrow-derived cells into mature cells of the liver [6, 7], brain [8], heart [9], pancreas [10], lung [11], kidney [12], and skeletal muscle [13]. In published studies, the percentage of marrow-derived epithelial cells found in the liver varies widely [14]. Such variability likely arises from differences in the population of marrow cells transplanted; ages of donor and recipient at transplantation; immunologic differences between donor and recipient; recipient conditioning regimen; efficiency of bone marrow engraftment; method used to induce organ injury; time interval between transplantation, injury, and sacrifice; and the methods used to quantitate marrow-derived epithelial cells (for review in liver, see Thorgeirsson and Grisham [14]).

Normal adult hepatocytes are mitotically quiescent, yet the liver possesses a remarkable capacity for regeneration. After partial resection, the remaining liver cells rapidly hypertrophy and proliferate, restoring both mass and function within a few days [15, 16]. If an acute injury, such as that caused by acute overdose of acetaminophen, is not too severe, the remaining liver will regenerate over several days to weeks [17, 18]. When overwhelming injury exceeds the liver’s capacity to recover quickly enough, liver transplantation becomes necessary. Chronic hepatic injury leading to fibrosis impairs the regenerative response, although the mechanism of this failure is unclear.

Putative intrahepatic progenitor “oval” cells were originally identified in models of hepatic carcinogenesis [19] as nondescript, undifferentiated cells with an ovoid nucleus. There are no known unique phenotypic markers of oval cells; their identification was based solely on anatomic and morphologic criteria. More recently, the term “oval cell” has come to be associated with small biliary cells in the canals of Hering, defined by their location and expression of biliary cytokeratins [20]. For clarity, we will refer to these cytokeratin-positive and hepatocyte nuclear factor-1 (HNF1)-positive cells as “ductal progenitor cells” rather than “oval cells.” Ductal progenitor cells are bipotent cells located in the smallest branches of bile ducts [20] that can differentiate into either hepatocytes or cholangiocytes [21–24]. These cells have been reported to express many of the same surface markers as those of hematopoietic stem cells, such as CD34⁺, c-kit⁺, Sca-1⁺, and Thy-1, supporting a potential common origin [22, 24–26]. However, recent studies have indicated that Thy-1 is actually expressed by hepatic myofibroblasts rather than oval cells or duct progenitors [27–29]. At least some ductal progenitor cells appear to have the “side population” phenotype due to their ability to efflux Hoechst 33342 dye (Hoechst Celanese Corp., Somerville, NJ, http://www.celanese.com), another striking similarity with hematopoietic stem cells [30]. Together, these lines of evidence suggested that, under certain conditions, cells from the bone marrow might engraft the liver as epithelial progenitor cells in the small bile ducts. Upon activation by liver injury, marrow-derived cells might then proliferate and give rise to cholangiocytes and hepatocytes. However, it is not clear whether hepatic ductal progenitor cells are a necessary intermediate for the appearance of marrow-derived mature hepatocytes.

The differences in liver injury models and detection method for marrow-derived liver cells have led to substantial variability and disagreement over the interpretation of these experiments. As we investigated these issues, we learned many potential pitfalls and present a new approach to the identification of marrow-derived liver cells. We found that identification of marrow-derived epithelial cells depends critically on detection technique. In all cases, the marker of donor origin was green fluorescent protein (GFP) or Y chromosome, both of which have technical limitations. Proper identification of the phenotype of any given GFP+ or Y+ liver cell is critical because morphologic criteria are clearly not sufficient. Colocalization of GFP or Y with plasma membrane proteins, such as cytokeratin, may improve the specificity of identification of marrow-derived epithelial cells. Cytokeratins 7, 8, 18, and 19 are highly expressed in cholangiocytes and ductal progenitor cells, but at much lower levels in hepatocytes [31]. Therefore, cytokeratin may be more useful than marrow-derived hepatocytes for identifying marrow-derived cholangiocytes and ductal progenitor cells. Albumin immunostaining can be used to identify hepatocytes, although—in our experience—its cytoplasmic staining pattern does not clearly delineate cell boundaries when the majority of cells in a microscopic field are albumin-positive. Immunostaining for cytokeratin or albumin may not clearly distinguish liver epithelial cells from closely interposed lymphocytes or Kupffer cells, potentially leading to false-positive interpretation. In xenotransplant experiments using immunodeficient mice as recipients of various human progenitor cells, anti-human albumin staining identified human cells in immunodeficient mouse liver. However, these cells lacked typical hepatocyte morphology, did not express P450 3A4, and had lost human alu markers, suggesting either poor specificity of albumin staining or, perhaps, horizontal transfer of human albumin expression to mouse cells in the absence of either cell-cell fusion or donor cell reprogramming [32]. Dipeptidyl peptidase IV (DPPIV, CD26) is a transmembrane glycoprotein enzyme that cleaves dipeptides from the amino terminus of peptides in the extracellular space. DPPIV has been used as a marker of donor origin in cell transplant experiments in which the recipient is DPPIV null [33–36]. In hepatocytes, DPPIV is expressed on the canalicular membrane and can be stained enzymatically [36] or immunohistochemically [34]. However, DPPIV is expressed in several other cell types, including lymphocytes, granulocytes, and macrophages [37, 38]. Therefore, DPPIV may be a suitable marker of donor origin, but cannot be used to unequivocally demonstrate hepatocyte, cholangiocyte, or ductal progenitor cell phenotype in bone marrow transplantation (BMT) experiments.

We have established an approach to overcome these issues by using a nuclear phenotypic marker for hepatocytes, cholangiocytes, and ductal progenitor cells. The closely related transcription factors, hepatocyte nuclear factor-1 alpha (HNF1α) and beta (HNF1β) regulate expression of many hepatic genes. HNF1α is expressed strongly in hepatocytes but less strongly in cholangiocytes [39]. Conversely, HNF1β is more highly expressed in cholangiocytes than in hepatocytes [39]. In this study we found that immunofluorescent staining of liver sections using a commercially available antibody that recognizes both HNF1α and HNF1β highlights virtually all the parenchymal cells of the liver, clearly distinguishing these cells from Kupffer cells, lymphocytes, neutrophils, plasma cells, endothelial cells, fibroblasts, and stellate cells. Double-staining of tissue sections by Y chromosome fluorescence in situ hybridization (Y-FISH) and HNF1 immunofluorescence dramatically improves the resolution of marrow-derived blood cells from native or marrow-derived hepatocytes, cholangiocytes, and ductal progenitor cells. Using this rigorous method we identified rare marrow-derived hepatocytes, but no bone marrow-derived cholangiocytes or ductal progenitor cells after liver injury with carbon tetrachloride (CCl₄) or 3,5-diethoxycarbonyl-1,4-dihydrocollidine (DDC).

Materials and Methods

Animals

Most BMT experiments were conducted using 4-week-old normal female C57BL/6 recipients. Bone marrow donors were male C57BL/6-TgN(ACTbEGFP)10sb (GFP+) mice (Stock # 003291, The Jackson Laboratory, Bar Harbor, ME, http://www.jax.org), in which the GFP reporter is expressed in multiple tissues under the control of the chicken beta actin promoter, cytomegalovirus enhancer, beta actin intron, and bovine globin polyadenylation signal [40]. As a positive control for marrow-derived hepatocytes, three female fumaryl acetoacetate hydrolase (FAH)-null mice (see following text for details) received a BMT from normal male GFP+ donors.

All animals were housed either at the Wadsworth Center or at the Yale University School of Medicine, in a temperature- and humidity-controlled 12-hour light/dark cycle environment, with free access to standard rodent chow and water. All animal use protocols were approved by the Wadsworth Center or the Yale University School of Medicine Institutional Animal Care and Use Committee.

Bone Marrow Transplantation

Male GFP+ donor mice, 8–10 weeks of age, were anesthetized with isoflurane (Abbott Laboratories, North Chicago, IL, http://www.abbott.com) and sacrificed by cervical dislocation. Donor marrow was flushed with phosphate-buffered saline (PBS) from femurs and tibias using sterile technique. Cells were washed twice in Hank’s balanced salt solution (HBSS), resuspended in HBSS, and counted. Viability was >90%, based on trypan blue dye exclusion. Female recipient mice were irradiated with a ¹³⁷Cs source (1,000 rads) and injected with 1 × 10⁶ male GFP+ bone marrow cells via tail vein. Prophylactic antibiotics were added to the drinking water for 4 weeks after transplantation to limit mortality during the period of transient leukopenia prior to stable hematopoietic engaftment. High-level bone marrow engraftment (>80%) was confirmed by fluorescence-activated cell sorting for GFP and/or FISH for the Y chromosome in all mice.

3,5-Diethoxycarbonyl-1,4-Dihydrocollidine (DDC)

Thirty-six bone marrow-transplanted mice were fed a standard chow supplemented with 0.1% DDC (Bio-Serv Inc., Frenchtown, NJ, http://www.bio-serv.com) for 10 days. Some animals were sacrificed immediately at the end of DDC treatment, whereas others were allowed to recover on a standard chow diet for 1–12 weeks.

Twenty of the above-mentioned bone marrow-transplanted mice were also treated with 1,000 rads focal irradiation (Siemens Stabilipan 250-kV x-ray source, Siemens Medical Systems, Inc., Iselin, NJ, http://www.medical.siemens.com), limited by lead to the upper abdomen. The purpose of liver irradiation was to inhibit hepatocyte regeneration from native hepatocytes, favoring engraftment of marrow-derived liver cells. These mice were then treated for 10 days with a 0.1% DDC diet. Mice were sacrificed 0–12 weeks after the 10th (final) day of the DDC diet.

Carbon Tetrachloride (CCl₄)

Fifteen bone marrow-transplanted mice received one or more intra-peritoneal injections of CCl₄ (0.5 ml/kg in mineral oil, Sigma-Aldrich, St. Louis, http://www.sigmaaldrich.com), as indicated in Table 1. Seven of these mice also received 2,000 rads focal liver irradiation by the same method described earlier, but three of these mice died prior to completion of the experiment. An additional seven bone marrow-transplanted control mice received 2,000 rads liver irradiation alone, without DDC or CCl₄.

Table 1.

Experimental groups

Experimental treatment groups	Number of mice
1. Controls – no liver injury	4
2. DDC+/− liver irradiation (1000 rads)
A. DDC diet for 10 days	16
B. DDC diet for 10 days plus liver irradiation	20
3. CCl4 +/− liver irradiation (2000 rads)
A. CCl4, single dose	6
B. CCl4, single dose, plus liver irradiation	4
C. Liver irradiation alone	7
D. CCl4, 8 doses over four weeks	2

FAH-Null Mice

The FAH knockout mouse [41] was kindly provided by Dr. Marcus Grompe (Oregon Health Sciences University, Portland, OR). These mice have a defect in tyrosine catabolism that causes severe liver injury and death. These effects are largely, although incompletely, mitigated by administration of 2-(2-nitro-4-trifluoro-methylbenzyol)-1,3-cyclohexanedione (NTBC; Dr. Ronald McLard, Reed College, Portland, OR) at 10 μg/ml in the drinking water as described [42]. Female FAH-null mice underwent whole-body gamma irradiation and BMT from male GFP+ donors, as described earlier. Four weeks after BMT, NTBC was removed from and replaced into the drinking water in cycles for 6 months, as described [43]. These mice provided controls for the appearance of marrow-derived hepatocytes, although it is known that these cells arise primarily, if not exclusively, from fusion of donor-derived cells of the monocyte-macrophage lineage with diseased host hepatocytes [44].

Tissue Sampling, Fixation, and Processing

Tissues were fixed in phosphate-buffered formalin (Fisher Scientific Co. LLC, Morris Plains, NJ, http://www.fisherscientific.com) for 4–18 hours at room temperature. Formalin was replaced by 70% ethanol until tissues were embedded in paraffin and cut in 3-micron sections.

Immunofluorescence and FISH Analysis

Slides were heated to 60°C for 2 minutes, deparaffinized in Citrisolv (Fisher Scientific International, Hampton, NH, http://www.fisherscientific.com), and rehydrated through graded ethyl alcohols to PBS. Antigen retrieval was performed using Signet pH-All-2 (Signet Laboratories, Inc., Dedham, MA, http://www.signetlabs.com) or BD Retrievagen (BD Biosciences, San Jose, CA, http://www.bdbiosciences.com) for 30 minutes in steam, then cooled to room temperature and washed in PBS. We found that either heat-mediated antigen retrieval solution worked very well for Y-FISH and completely avoided tissue degradation artifacts induced by enzymatic digestion with proteinase K (data not shown). A digoxigenin-labeled mouse Y chromosome probe was applied as described [45] and detected using a rhodamine-conjugated secondary antibody to digoxigenin (Roche Diagnostics, Basel, Switzerland, http://www.roche-applied-science.com). Following detection of the Y chromosome, immunofluorescence for CD45 (rat anti-mouse pan-leukocyte marker, #550539, 1:20 dilution, BD Pharmingen, San Diego, http://www.bdbiosciences.com), F4/80 (rat anti-mouse macrophage marker, #14-4801, 1:20 dilution, eBioscience Inc., San Diego, http://www.ebioscience.com), GFP (rabbit anti-GFP, #A11122, 1:100 dilution, Invitrogen Molecular Probes, Carlsbad, CA, http://www.invitrogen.com), cytokeratin (Wide Spectrum Screening rabbit anti-cow cytokeratin, #Z0622, 1:50 dilution, DakoCytomation, Glostrup, Denmark, http://www.dakocytomation.com), or HNF1 (rabbit anti-human HNF1, #sc-8986, 1:25 dilution, Santa Cruz Bio-technology Inc., Santa Cruz, CA, http://www.scbt.com) was performed. In each case, antibodies were diluted in PBS/1% bovine serum albumin (BSA) and incubated overnight at 4°C. Following PBS washes, goat anti-rat Alexa 488 (#A11006, 1:100 dilution, Invitrogen Molecular Probes) or goat anti-rabbit fluorescein isothiocyanate (FITC) secondary antibody (#F2765, 1:100 dilution, Invitrogen Molecular Probes) in PBS/1% BSA was incubated for 1 hour at 37°C. Slides were washed in PBS and mounted in fluorescence mounting media with 4′,6-diamidino-2-phenylindole (DAPI, Vector Laboratories, Burlingame, CA, http://www.vectorlabs.com).

Tissue Analysis and Cell Counts

Counting of Y-positive, HNF1-positive hepatocytes, cholangiocytes, and ductal progenitor cells was accomplished by systematically examining the treated slides, field by field, under ×40 magnification, using an Olympus BX-51 microscope (Olympus, Tokyo, http://www.olympus-global.com). Y-FISH and HNF1 double-positive cells were identified using a dual-wavelength channel filter set designed for simultaneous visualization of FITC and rhodamine (Chroma Technology, Rockingham, VT, http://chroma.com). Photomicrographs were taken using a digital camera and IPLab imaging software (BD Biosciences, Rockville, MD, http://www.bdbiosciences.com). Images were obtained using excitation and emission filter sets for DAPI (all nuclei), rhodamine (Y chromosome), and FITC (CD45, F4/80, GFP, cytokeratin, or HNF1). Confocal microscopy was performed using a Zeiss LSM-510 microscope (Carl Zeiss, Jena, Germany, http://www.zeiss.com), with appropriate filters for DAPI, FITC, and rhodamine.

RESULTS

Liver Injury Models

Four weeks after BMT, mice were subjected to liver injury, summarized in Table 1. Representative hematoxylin and eosin micrographs are shown in Figure 1. Whole-body irradiation and BMT did not cause appreciable changes in liver histology compared with nontransplanted normal controls (Fig. 1B vs. 1A). Deliberate liver injury by focal irradiation (Fig. 1C) did not induce gross injury or inflammation, but did induce hepatocyte mitoses, indicating that hepatocytes had been lost and were being replaced by dividing hepatocytes rather than ductal proliferation. In contrast, the DDC diet induced inflammation, cholestasis, and striking ductal proliferation (Fig. 1D), but very few mitotic hepatocytes. The recipient liver after CCl₄ injury is shown in Figure 1E and 1F. Two days after administration of CCl₄, pericentral necrosis and diffuse steatosis were evident (Fig. 1E). Fourteen days after CCl₄ administration, dead pericentral hepatocytes were no longer present and very few mitotic hepatocytes were seen (Fig. 1F). Instead, proliferating bile ducts appeared to be generating new hepatocytes. Combining CCl₄ with 2,000 rads liver irradiation did not prevent ductal proliferation at 14 days, but did increase mortality (four of seven mice survived) compared with CCl₄ alone (seven of seven survived) or irradiation alone (seven of seven survived).

Figure 1.

Histology of normal and injured liver. (A): Normal liver, no BMT. (B): No liver injury, 12 weeks after BMT. (C): Liver irradiation, 7 days after injury. Arrows indicate mitotic hepatocytes. (D): DDC, 7 days. (E): CCl₄, 2 days after a single administration. Arrow indicates necrotic pericentral hepatocytes. (F): CCl₄, 14 days after a single administration. Arrow indicates ductular proliferation. Abbreviations: BMT, bone marrow transplantation; CCl₄, carbon tetrachloride; DDC, 3,5-diethoxycarbonyl-1,4-dihydrocollidine.

Markers of Donor Origin: Detection of GFP or Y Chromosome in the Bone Marrow Donor Liver

We found that tissue processing and heat-mediated antigen retrieval required for Y-FISH and all of our immunofluorescence protocols effectively quenched the intrinsic fluorescence of GFP in fixed sections. This eliminated intrinsic GFP fluorescence from the background and necessitated the use of immunofluorescent detection of GFP. Autofluorescence in the liver was also technically problematic across a broad spectrum of excitation and emission wavelengths, particularly after cholestatic liver injury. These issues required the development of very robust immunofluorescence protocols and additional negative controls.

Prior to conducting our BMT experiments, it was essential to examine the pattern of GFP expression in the livers of bone marrow donors. In all experiments, the bone marrow donors were the “1Osb” transgenic line [40], in which GFP expression under the control of the chicken beta actin promoter is widespread but not truly ubiquitous [46] (Fig. 2A). Immunofluorescent localization of GFP in the liver of bone marrow donor mice showed that only approximately half of hepatocytes expressed GFP at high levels (Fig. 2A). The remainder of hepatocytes had only background autofluorescence, equal to that seen in non-transgenic controls (Fig. 2B). Cholangiocytes in GFP donor mice also expressed GFP, also in a variegated pattern and at an absolute level lower than that seen in hepatocytes. The mechanism by which GFP is silenced in some adult cells of the 1Osb mouse is unknown, but the expression of GFP in only a fraction of liver epithelial cells in the donor suggests that bone marrow donor-derived liver cells may be underestimated if GFP is used as the sole marker of donor origin.

Figure 2.

Immunofluorescent localization of GFP. (A): GFP is expressed at high levels in approximately half of hepatocytes (arrow) in the beta actin-GFP mice used as bone marrow donors. GFP is also variably expressed in bile duct cells (arrowhead). (B): Nontransgenic control liver. Note background autofluorescence in hepatocytes. (C): Nontransgenic recipient of GFP donor BMT, with GFP detection by DAB immunohistochemistry. A potential marrow-derived hepatocyte (enlarged in inset) stains brown. (D): Nontransgenic recipient of GFP donor BMT, with GFP detection by immunofluorescence (green). Numerous GFP-positive cells are seen surrounding a bile duct and passing through sinusoids. (E): Colocalization of Y chromosome (small pink nuclear dot) and GFP. Both are markers of male GFP+ donor origin in this nontransgenic female recipient liver. A sinusoidal cell staining for GFP but not Y-FISH is indicated by the arrow. A cell staining for Y chromosome but not GFP is indicated by the arrowhead. A Y-positive potential marrow-derived hepatocyte is circled in red. The large orange cells are autofluorescent erythrocytes. Abbreviations: BMT, blood marrow transplantation; DAB, 3,3′-diaminobenzidine; GFP, green fluorescent protein; Y-FISH, Y chromosome fluorescence in situ hybridization.

Detection of the Y chromosome is not limited by transgene silencing or nuclear reprogramming events, but is limited by partial nuclear sampling in 3-micron sections, which could also lead to underestimation of marrow-derived liver cells. Y-FISH labeled 79%–91% of hepatocytes and cholangiocytes in male control mouse livers (not shown), consistent with previous reports [47]. Estimates of marrow-derived epithelial cells by detection of Y chromosome are potentially underestimated if host-donor fusion events are followed by reductive cell division, as has been suggested recently in a model of lung injury [48].

In all of the injury models examined, inflammatory cells could be detected by both Y-FISH and GFP immunofluorescence (Fig. 2D, 2E). Where inflammatory cells are densely packed, distinction of these cells from epithelial cells can be difficult. We found that Y-FISH provided a more suitable marker of donor origin than GFP and was compatible with subsequent immunofluorescent marking of epithelial phenotype (see following text).

Markers of Donor Origin After BMT

Approximately 70% of the bone marrow cells of the recipients were GFP-positive by flow cytometry, the same percentage as in the GFP+ donor, indicating close to 100% replacement (data not shown). GFP+ cells were found in all organs examined, including small intestine, spleen, liver, and pancreas. In the liver, GFP+ mononuclear cells are seen in the portal zone around the ducts and many Kupffer cells are GFP+ (Fig. 2C, 2D). Although some GFP+ cells morphologically resembled hepatocytes (Fig. 2C), it is not possible to rule out that they are macrophages, indicating the need for independent confirmation of hepatocyte phenotype by a specific marker. Combination of Y-FISH with GFP immunofluorescence in a BMT recipient liver revealed cells that were positive only for GFP, positive only for Y, or positive for both (Fig. 2E). Cells positive only for GFP likely reflect partial nuclear sampling, which eliminated the Y chromosome. Cells positive only for Y likely reflect the incomplete expression of GFP in bone marrow donors. Figure 2E also illustrates a cell that appears to be a Y+ hepatocyte (circled), but this particular cell is clearly GFP negative. As in Figure 2C, accurate confirmation of phenotype is critical, independent of the marker of donor origin (Y or GFP).

Identification of Marrow-Derived Epithelial Cells After BMT

GFP+ and Y+ cells were readily detected in the liver after BMT and increased in proportion to the degree of liver injury. The abundance of these cells in close proximity to hepatocytes, cholangiocytes, and ductal progenitor cells required strict resolution of cell boundaries and elimination of cell overlap. Attempts to label and exclude donor-derived hematopoietic cells with CD45, the pan-leukocyte marker, met with limited success in the liver because the CD45 antibody did not consistently stain macrophages (not shown). Addition of F4/80, the pan-macrophage marker antibody, to anti-CD45 increased the number of cells visualized, but still left many Y+ cells unlabeled (Fig. 3). These cells were almost certainly blood cells. Therefore, the absence of CD45 + F4/80 staining could not be reliably applied as a criterion for identification of marrow-derived liver epithelial cells. Figure 3 also illustrates another example of an apparent Y+ hepatocyte, without positive confirmation of epithelial phenotype.

Figure 3.

Costaining of recipient female liver after male BMT and liver injury with CCl₄. Y-FISH (small pink nuclear dot) and CD45 + F4/80 (combined to highlight leukocytes, green). (A): Liver section at ×40 magnification. White boxes indicate magnified views in (B) and (C). (B): Top right: magnified view: Note Y+ (pink dot) and CD45 + F4/80-positive (green) cells in the top left, and a Y+ potential hepatocyte (circled in white). (C): Bottom left: Numerous Y-positive cells near the bile duct and portal vein do not stain with CD45 + F4/80 (arrowhead in [C]), indicating that the absence of CD45 + F4/80 staining is not sufficient evidence of epithelial phenotype. Abbreviations: BMT, bone marrow transplantation; CCl₄, carbon tetrachloride; Y-FISH, Y chromosome fluorescence in situ hybridization.

Since morphologic criteria and the absence of leukocyte marker staining are not adequate to score a given marrow-derived cell as epithelial, many investigators have used cytokeratin (CK) immunostaining as a marker of epithelial cell phenotype in the liver. Following Y-FISH, we used an anti-cytokeratin antibody that recognizes CK 8, 18, and 19 (Fig. 4A–4C). This antibody gives robust staining of bile ducts and ductal progenitor cells. In most cases, Y+, CK-negative cells could be clearly distinguished from adjacent CK+ cells (Fig. 4B). However, as shown in Figure 4C, colocalization of Y chromosome and cytokeratin occasionally led to the apparent identification of marrow-derived cholangiocytes. We could not rigorously exclude the possibility that these Y+ cells represented infiltrating inflammatory cells, intercalated between adjacent cholangiocytes. In hepatocytes, CK staining was so weak it could not always be reliably distinguished from background.

Figure 4.

Colocalization of Y chromosome (pink) and cytokeratin (green). Pancytokeratin immunofluorescence stains cholangiocytes and ductal progenitor cells strongly, but stains hepatocytes weakly. (A): Male liver. (B, C): Female recipient of male BMT followed by liver injury with DDC. Note that cytokeratin-positive oval cells are surrounded by donor-derived (likely inflammatory) cells. Arrow in (C) indicates a potential marrow-derived cholangiocyte. Abbreviations: BMT, bone marrow transplantation; DDC, 3,5-diethoxycarbonyl-1,4-dihydrocollidine.

Our attempts to identify hepatocytes using immunofluorescence for albumin after Y-FISH were generally unsatisfactory due to nonspecific staining and poor single-cell resolution (not shown), possibly arising from the fact that albumin is a secreted protein, circulating in the blood. For all of the reasons stated earlier, we sought to develop a method to specifically mark the nucleus of hepatocytes and cholangiocytes, excluding cells of the blood lineage. Liver-enriched transcription factors were considered as candidates for such a marker. We found that a commercially available antibody to HNF1α/β stained hepatocytes and cholangiocytes in formalin-fixed paraffin sections with high sensitivity and specificity. Antigen retrieval with a highly alkaline unmasking buffer (Signet pH-All-2, pH 10) was critical. This antigen retrieval was optimal for Y-FISH and HNF1, but completely abolished staining for CD45 and F4/80 (data not shown). Since we had determined that the leukocyte staining was inadequate for our purposes, we used Y-FISH and HNF1 immunofluorescent costaining to identify marrow-derived liver cells. Costaining for Y-FISH (pink nuclear dot) and HNF1 (green nuclear stain) in normal control male and female livers is shown in Figure 5A and 5B, respectively. All nuclei are counterstained blue with DAPI, so HNF1-positive cells are blue-green, whereas HNF1-negative cells are only blue. Hepatocytes were distinguished by large nuclei, autofluorescent green cytoplasm, and HNF1-positive nuclei. Cholangiocytes were distinguished by ductular location, cuboidal morphology, and HNF1-positive nuclei. Ductal progenitor cells appeared as cholangiocytes not bordering duct lumina (Fig. 5E, long arrow). Periductal cells were clearly HNF1-negative, as were numerous cells in the liver parenchyma. Many of the HNF1-negative cells were located in sinusoids (Fig. 5B). Given their large size and sinusoidal location, most of them were likely Kupffer cells and could have been mistaken for hepatocytes based solely on morphological criteria, or even with staining of cytokeratin or albumin in the cells that surround them.

Figure 5.

Colocalization of Y chromosome (pink) and HNF1 (green nuclear stain). (A): Male control liver. (B): Female control liver. (C): Female FAH null recipient after male donor BMT and NTBC withdrawal to select marrow-derived hepatocytes. (D–F): Female recipient of male BMT, followed by DDC treatment. (D): Y-positive, HNF1-negative cells (arrowheads) are donor-derived blood cells, clearly distinguished from native hepatocytes (HNF1-positive, Y-negative). (E): Y-negative, HNF1+ hepatocytes (short arrow) and cholangiocytes (long arrow) can be distinguished from Y+, HNF1-negative cells (arrowheads) in DDC-treated female mice after male donor BMT. (F): Inflammatory cells infiltrating a bile duct are Y-positive, HNF1-negative. DDC-induced cholestasis causes hepatocyte cytoplasm to appear reddish-orange in (E) and (F). Abbreviations: BMT, bone marrow transplantation; DDC, 3,5-diethoxycarbonyl-1,4-dihy-drocollidine; FAH, fumaryl acetoacetate hydrolase; HNF1, hepatocyte nuclear factor-1; NTBC, 2-(2-nitro-4-trifluoro-methyl-benzyol)-1,3-cyclohexanedione.

To validate the Y-FISH and HNF1 approach, we performed costaining on liver sections from female FAH-null recipients of male BMT. After selecting for marrow-derived hepatocytes by cycling NTBC, we found clusters of Y+, HNF1+ hepatocytes (Fig. 5C), which could be clearly distinguished from Y+, HNF1-negative blood cells. Y-negative, HNF1 + hepatocytes are likely endogenous female hepatocytes, but cannot be distinguished from marrow-derived hepatocytes in which the Y chromosome has been lost due to sectioning artifact, reductive division, or chromosomal instability. In mice injured with DDC (Fig. 5D–5F), the vast majority of marrow-derived cells were clearly Y+, HNF1-negative, consistent with circulating blood cells or resident macrophages (Fig. 5D). The majority of hepatocytes were HNF1+, Y-negative, with clear resolution of their nuclei from surrounding cells. Proliferating bile duct cells were HNF1+, Y-negative (Fig. 5E) and could be distinguished from the Y+, HNF1-negative inflammatory cells surrounding them. Within larger bile ducts (Fig. 5F), Y+, HNF1-negative inflammatory cells could be clearly seen intercalating with HNF1+, Y-negative cholangiocytes. Thus, this method very clearly distinguished epithelial cells from blood cells and avoided false-positive interpretation of marrow-derived hepatocytes, cholangiocytes, and ductal progenitor cells. For each liver section examined by Y-FISH we carefully counted at least 1,000 HNF1-positive hepatocyte nuclei in approximately 20 high-power fields. In each case, a putative marrow-derived hepatocyte was viewed carefully using the DAPI filter. In some cases, it was apparent that an overlying or closely adjacent nucleus, presumably that of a blood cell, accounted for the apparent Y+ signal in a hepatocyte nucleus. These occurrences represented another type of false-positive and were excluded from further analysis. At least five portal tracts containing a total of 50–100 cholangiocytes were examined for each animal, using the same Y-FISH and HNF1 criteria. No Y-positive cholangiocytes were found in any animal, with or without liver injury. As shown in Figure 5F, Y+ inflammatory cells infiltrating bile ducts after DDC injury could be clearly distinguished from cholangiocytes by their lack of HNF1 expression.

In mice injured with CCl₄ (Fig. 6A, 6B) or DDC (Fig. 6C, 6D), we found rare Y+, HNF1+ hepatocytes, but no Y+, HNF1+ cholangiocytes or ductal progenitor cells. The data are summarized in detail in supplemental Table 1. We found no Y+, HNF1+ hepatocytes out of >4,000 nuclei examined in BMT recipients without liver injury, sacrificed 12 weeks after BMT. We found a total of four Y+, HNF1+ nuclei out of 18,345 examined (0.02%) among mice treated with DDC for 10 days. Two of these cells were found in one animal, sacrificed 1 week after completing a 10-day course of DDC. We hypothesized that addition of liver irradiation to the DDC treatment would increase the number of marrow-derived hepatocytes by inhibiting replication of native hepatocytes in response to liver injury. In fact, we found the opposite. Among mice treated with both DDC and liver irradiation, we found no Y+, HNF1+ hepatocytes among 23,816 examined, indicating that liver irradiation did not increase the appearance of marrow-derived hepatocytes. Similarly, liver irradiation alone did not result in appearance of marrow-derived hepatocytes (zero of 7,000 cells examined). Among the mice treated with CCl₄, two Y+, HNF1+ cells were found among 13,160 examined (0.16%). One appeared after a single dose of CCl₄ and another after chronic CCl₄ administration over 8 weeks. Addition of liver irradiation to CCl₄ injury did not increase the appearance of marrow-derived liver cells; we found zero Y+, HNF1+ cells out of 4,000 HNF1+ hepatocytes examined.

Figure 6.

Colocalization of Y chromosome (pink nuclear dot) and HNF1 (green nuclear stain) in female recipients of male BMT, after liver injury. (A): Y-positive, HNF1-positive hepatocyte nucleus (arrow) in a mouse treated with two doses of CCl₄ over 7 days. (B): Y-positive, HNF1-positive hepatocyte nucleus (arrow) in a mouse treated with eight doses of CCl₄ over 8 weeks. (C): Y-positive, HNF1-positive hepatocyte nucleus (arrow) in a mouse treated with DDC for 10 days and allowed to recover for 7 days. (D): Confocal image of the hepatocyte in (C) demonstrates localization of Y chromosome and HNF1 in the same plane as nuclear DNA (DAPI). Abbreviations: CCl₄, carbon tetrachloride; DAPI, 4′,6-diamidino-2-phenylindole; DDC, 3,5-diethoxycarbonyl-1,4-dihydrocollidine; HNF1, hepatocyte nuclear factor-1.

Discussion

Allogeneic hepatocyte transplantation is currently limited by poor availability, limited success, and the requirement for long-term immunosuppression. Recent exciting advances in stem cell research suggest that it may be possible to generate patient-specific pluripotent cells from both embryonic and adult sources [49, 50]. These pluripotent cells might then be induced to form hepatocyte-like cells for therapeutic applications. Whether adult stem cells from bone marrow will be a suitable source for such applications remains to be seen. Whatever the source, a better understanding of genetic “reprogramming” mechanisms is essential to the development of innovative cell therapy approaches for liver disease. Accurate identification of donor-derived liver cells both in vitro and in vivo is critical [51]. Following BMT, the overwhelming majority of marrow-derived cells in the liver are blood cells. Distinguishing the rare marrow-derived epithelial cell from the background of Y+ or GFP+ cells of hematopoietic origin requires carefully validated methods.

In our search for marrow-derived liver cells, we recognized several limitations of markers of donor origin and recipient cell phenotype. Since GFP was expressed in only approximately half of hepatocytes and 70% of nucleated blood cells in our bone marrow donors, we believed that we could miss a substantial proportion of marrow-derived liver cells using GFP as the marker of donor origin. We prefer the Y chromosome as marker of donor origin because it is a nuclear marker, not susceptible to epigenetic silencing. We found that colocalization of Y chromosome, together with HNF1 as marker of hepatocyte/cholangiocyte phenotype, reduced potential artifactual interpretation of blood cells as marrow-derived epithelial cells. This strategy would be expected to miss approximately 10%–20% of marrow-derived hepatocytes, due to missed Y chromosome in partially sectioned nuclei. Increasing section thickness reduces partial nuclear sectioning, but increases the potential for nuclear overlap, leading to false-positives. Nuclear overlap can be effectively ruled out by confocal microscopy, but we found that HNF1 immunostaining did not work well in sections greater than 10 microns (data not shown). Preservation of tissue morphology is critical for this analysis, so we chose to work exclusively with paraffin sections rather than frozen sections.

Marrow-derived inflammatory cells, which could have been misinterpreted as marrow-derived cholangiocytes using Y and cytokeratin colocalization, were unambiguously shown to be HNF1-negative. Proliferating duct cells induced by DDC expressed HNF1, allowing us to determine that these cells were not marrow-derived (Fig. 5E).

Previous investigations of the potential bone marrow origin of hepatic oval (duct progenitor) cells have yielded conflicting results. Bone marrow-derived oval cells were first reported using 2-acetylaminofluorene (2-AAF)/CCl₄ liver injury in rats [33]. In subsequent related work, marrow-derived DPPIV+, alphafetoprotein+ (AFP+) hepatic oval cells were generated in DPPIV-null BMT recipients, isolated based on Thy-1 expression, then transplanted by intrasplenic injection into DPPIV-null secondary recipients [34]. DPPIV+ hepatocytes appeared in the liver of secondary recipients, providing further support for their origin from the original bone marrow donor. The order and timing of administration of hepatotoxin and BMT were found to be critical variables. Administration of the mitotic inhibitor monocrotaline, prior to BMT, was associated with the appearance of DPPIV+, AFP+ oval cells following 2-AAF/partial hepatectomy. Mice treated with monocrotaline after BMT did not form DPPIV+, AFP+ oval cells after 2-AAF/partial hepatoectomy, indicating a possible direct mitoinhibitory effect of monocrotaline on the transplanted bone marrow cells [34]. In our studies, liver injury always followed BMT, so we cannot exclude the possibility that a toxic effect of CCl₄ or DDC on engrafted donor bone marrow cells could have limited their engraftment as cholangiocytes or ductal progenitors.

Others have found that hepatic oval cells do not arise from the bone marrow in different models of liver injury, all of which induced liver injury after BMT. Using DPPIV as a marker of donor marrow origin, Menthena et al. [35], found that oval cells did not arise from transplanted bone marrow in rats with liver injury due to D-galactosamine, retrorsine/partial hepatectomy, or 2-AAF/partial hepatectomy. Oval cells induced by DDC in mice after BMT successfully repopulated the livers of FAH-null mice but were derived entirely from the liver of the primary BMT recipient, rather than the original bone marrow donor [52]. Transplanted GFP+ bone marrow failed to generate oval cells in mice treated with α-naphthylisothiocyanate (ANIT), despite severe cholestasis and robust ductal proliferation [53]. GFP+, marrow-derived oval cells, or hepatocytes were also not found in mouse models of liver injury with ANIT, CCl₄, or DDC [54]. Because of differences in the rodent species studied and the nature of the various liver injuries it is not possible to exclude the possibility that ductal progenitor/oval cells arise from the bone marrow under some conditions but not others. In any case, the interpretation of such experiments critically depends on very fine optical resolution of cells in tissues sections, using reliable markers of donor origin and rigorous criteria of cell phenotype.

In conclusion, we have identified several important potential sources of error in BMT experiments in vivo and refined the method to identify them. By combining Y-FISH and HNF1 immunofluorescence, we improved the specificity of detection of marrow-derived hepatocytes after BMT and liver injury. In our studies, marrow-derived hepatocytes were rare, solitary cells. We did not find marrow-derived cholangiocytes or ductal progenitor cells, although we cannot conclusively rule out this mechanism of liver engraftment. Additional approaches will be required to determine whether marrow-derived hepatocytes seen in our models of liver injury arose through cell-cell fusion.