Studies in rodents suggest the presence of a hepatopancreatic stem cell in adult pancreas that may give rise to liver cells in vivo. The aim of the present study was to determine the ability of human islet-derived cells to adopt a hepatic phenotype in vivo. Cultured human islet-derived progenitor cells that did not express albumin in vitro were stained with the red fluorescent dye PKH26 and injected into the liver of severe combined immunodeficiency mice. After 3 or 12 weeks, red fluorescent cells were detected in 11 of 15 livers and were mostly single cells that were well integrated into the liver tissue. Human albumin was found in 8 of 11 animals by immunohistochemistry, and human albumin mRNA was detected in 4 of 10 host livers. The mechanism underlying this phenomenon seems to be transdifferentiation, because human and mouse albumin were found to be expressed in distinct cells in the host liver.
During embryogenesis, progenitor cells of pancreas and liver emerge from neighboring areas of the gut endoderm . There is a large body of evidence suggesting that such progenitors with the potential to generate liver cells from pancreatic cells and vice versa may still exist in adult life [2–5]. Using the model of mice with a knockout of the tyrosine catabolic enzyme fumarylacetoacetate hydrolase (FAH), which, without treatment, results in liver cirrhosis, Wang et al.  succeeded in correction of liver function by transplantation of cell suspensions from adult pancreas of wild-type animals. With this unique repopulation assay, the authors clearly demonstrated that progenitor cells with the ability not only to replace FAH knockout cells in the liver but also to correct the liver function exist in the pancreas, although the exact nature of these cells remains unknown. Interestingly, cells from cultured pancreatic ducts were not able to rescue the failing liver as did the crude pancreatic cell suspension , indicating that the presumed hepatopancreatic stem cells reside in areas outside the pancreatic ducts.
Numerous studies have shown the transdifferentiation potential of pancreatic cells of rodents into hepatocytes in vivo. Some very rare cases of human pancreatic cancer with hepatoid phenotype indicated the existence of similar cells in human pancreas . Recently, progenitor cells have been described in rodent and human islets of Langerhans that express the neural stem cell marker nestin  and the side-population phenotype marker ABCG2 . The side-population cells in bone marrow represent a particularly potent stem cell population . Interestingly, the human nestin-expressing islet-derived progenitor (NIP) cells were able to adopt a hepatic phenotype in vitro with expression of markers like alpha feto protein and the transcription factor XBP . In contrast to animal data, however, no in vivo studies have been published so far demonstrating transdifferentiation of human pancreatic cells into a hepatic phenotype. Recently, in vivo models for transplantation of human cord blood cells into severe combined immunodeficiency (SCID) mouse liver have been established [11, 12]. Using the model with direct injection of cells into the liver, we demonstrate in the present report that human cells from cultured pancreatic islets of Langerhans engraft into SCID mouse liver and form cells expressing human albumin in vivo.
Materials and Methods
Highly purified human islets are donations from the islet transplantation center in Geneva, Switzerland (Drs. José Oberholzer and Thierry Berney). Growth and expansion of nestin-positive cells were induced by RPMI-1640 medium with 10% fetal calf serum and supplemented with basic fibroblast growth factor and epidermal growth factor (20 ng/ml of each). Nestin-positive cells were characterized by reverse transcription–polymerase chain reaction (RT-PCR) and immunocytochemistry. More than 90% of cells were nestin positive. They were cultured for 4 months in the expansion medium, which was changed every 3–5 days. Trypsinization and reseeding were performed every 10–14 days. All experiments were done with cells of passage 4 through 9. Before transplantation, NIP cells were harvested by trypsinization, washed twice with phosphate-buffered saline (PBS), and stained with the red fluorescent dye PKH26 (Sigma) according to the staining protocol of the supplier (Sigma). The concentration of PKH26 during incubation with NIP cells was 4 μM. Incubation was performed for 4 minutes at 25°C with 107 NIP cells per ml. William's E medium with 1% bovine serum albumin (BSA) was used to stop the staining reaction. All centrifugation and washing steps during the staining procedure were performed at room temperature. Afterward, the NIP cells were resuspended in PBS at a concentration of 1.5 × 104 cells, 1.5 × 105, and 7.5 × 105 per 100 μl (Table 1).
Table Table 1.. Detection of donor-derived cells after transplantation of islet-derived progenitor cells into severe combined immunodeficiency mouse livers
The transplantation protocol was approved by the Animal Care and Use Committee in Rheinland-Pfalz, Germany. SCID mice (age, 6–20 weeks; weight, 18–27 g) were obtained from Charles-River (Sulzfeld, Germany). They were fed a standard diet purchased from Sniff (Soest, Germany) and acidified drinking water ad libitum. Before transplantation, mice were anaesthetized by intraperitoneal injection of 61.5 mg/kg ketamine (ketamin-ratiopharm 50, Ratiopharm, Ulm, Germany) and 2.3 mg/kg xylazine (Rompun 2%, Bayer, Leverkusen, Germany). Ketamine and xylazine were combined immediately before administration. The peritoneal cavity was opened directly below the xiphoid cartilage, and the NIP cells were slowly (100 μl in approximately 60 seconds) injected into the parenchyma of the protruding liver lobe using a 26-gauge needle (0.45 × 25, Henke-Sass, Tuttlingen, Germany). Immediately before injection, the cell suspension was warmed to 37°C. Successful injection was approved by a short-term paleness of the liver lobe. Previously, negative findings regarding fluorescence signals, RT-PCR, and immunohistochemistry (see below) have been observed after injection of human mononuclear cells as controls .
Three or 12 weeks after transplantation (Table 1), SCID mice were killed by neck dislocation, and after opening the peritoneal cavity, the protruding liver lobe that received the injection was excised. This liver lobe was divided into three parts. One part was fixed in 4% paraformaldehyde for immunohistochemical analysis, and two parts were shock frozen in liquid nitrogen for RNA isolation and fluorescence microscopy. Using a cryotome (CM 3000 cryostat, Leica Instruments GmbH, Nussloch, Germany), 5-μm-thick cryosections were produced. Cryosections were transferred onto Super-Frost Plus slides (Menzel, Braunschweig, Germany), air dried, and immediately analyzed by fluorescence microscopy using a standard filter setup for visualization of PKH26.
For immunohistochemistry, peroxidase was blocked in 7.5% H2O2 in methanol for 60 minutes at 4°C. Unspecific binding sites were blocked in 3% BSA for 2.5 hours at 37°C. Afterward, an avidin/biotin-block (Vector Laboratories, Burlingame, CA) was performed as described by the manufacturer. Slides were then incubated with a 1:50 dilution of an affinity purified human albumin antibody produced in goat (Bethyl Laboratories, Montgomery, TX; catalogue No. A80-229A) for 60 minutes at room temperature. Detection of first antibody was performed using commercial Vectastain Elite ABC Kit (Vector Laboratories) as described by the manufacturer, followed by a 5-minute incubation with 0.6 mg/mL diaminobenzidine at room temperature. Finally, the sections were counterstained with a 1:5 dilution of Mayer's hemalum (Merck, Darmstadt, Germany).
For fluorescence-immunohistochemistry (confocal microscopy, 630-fold magnification), human albumin was detected using a polyclonal anti-human albumin antibody raised in rabbit (Abcam Ltd.) in a 1:500 dilution. Detection of the primary antibody was performed using a Cy3-labeled secondary antibody against rabbit, raised in donkey (Dianova GmbH) in a 1:1000 dilution. Counterstaining was done using 4′,6′-diamidino-2-phenylindole (DAPI) (Molecular Probes) at a concentration of 2.3 μg/1,000 μL. Mouse albumin was detected using a polyclonal fluorescein isothiocyanate–labeled anti-mouse albumin antibody raised in goat (Bethyl Ltd.) in a 1:25 dilution. Counterstaining was done using DAPI (Molecular Probes) at a concentration of 2.3 μg/1,000 μL. A digital overlay was performed of Figures 2B and 2C.
For RT-PCR, homogenization of liver samples was performed by a polytron homogenizer, and total RNA was isolated using RNeasy Midi Kit (Qiagen, Hombrechtikon, Switzerland). Total RNA, 1 μg, was subjected to RT-PCR using Omniscript and Taq PCR core kit (Qiagen). Negative controls without reverse transcriptase enzyme were run in parallel to exclude possible contamination. C-DNA was amplified for 38 cycles (94°C for 30 seconds; annealing temperature 60°C for 60 seconds; 72°C for 60 seconds) using the following human-specific, intron-spanning primers: albumin forward ACTTTTATGCCCCGGAACTC and reverse AGCAGCAGCACGACAGAGTA, ABCG2 forward CACA GGTGGAGGCAAATCTT and reverse TCCAGACACAC-CACGGATAA, SCF forward GGTGGCAAATCTTCCAA AAG and reverse TCTTTCACGCACTCCACAAG, c-Kit forward GGCATCACGGTGACTTCAAT and reverse GGT TTGGGGAATGCTTCATA, Thy-1 forward GTCCTTTC TCCCCCAATCTC and reverse GGGAGACCTGCAAGAC TGTT, IPF-1 forward CCTTTCCCATGGATGAAGTC and reverse TTGTCCTCCTCCTTTTTCCA, insulin forward CTACCTAGTGTGCGGGGAAC and reverse GCTGGTA GAGGGAGCAGATG, CD45 forward CAGGCAGCAAT GCTATCTCA and reverse CTGTGATGGTGGTGTTG-GAG, and adenine phosphoribosyltransferase (APRT) forward GCGTGGTATTCAGGGACATC and reverse CAGGG CGTCTTTCTGAATCT. Identity of the amplified PCR product was confirmed by sequencing.
Normal human metaphase spreads were prepared from approximately 70% confluent NIP cells in a T25 culture flask by addition of 20 μl colcemid (10 μg/ml, Karyo Max, Gibco-BRL) and incubation for 1.5 hours. Cells were then trypsinized, pelleted, and resuspended in hypotonic solution (KCl-Na-citrate). After fixation with methanol/acetic acid (3:1), typsin/giemsa G-banding was performed according to standard laboratory procedures. Slides were examined with a ZEISS-Axiophot (Zeiss), and images were taken using a CCD camera (Cohu) and specially designed software (Karyotech 2000).
NIP cells expressed beside nestin also the side-population marker ABCG2 as well as SCF, c-Kit, and Thy-1, another potential marker for hepatic stem/progenitor cells (Fig. 1A). These cells were negative for expression of the transcription factor IPF-1 and insulin but also the specific marker for hematopoietic cells CD45 (Fig. 1B). Before transplantation, no albumin expression was found in cultured NIP cells (Fig. 1C). To exclude numerical or structural chromosomal aberration due to prolonged growth stimulation in vitro, a karyotyping was performed and revealed a normal 46, XX karyotype (Fig. 2).
Transplantation of 1.5 × 104 human NIP cells failed to result in detectable red fluorescent cells that became detectable only after transplantation of 1.5 × 105 cells in three of four animals (Table 1). We next evaluated the impact of time after transplantation on engraftment frequency and found similar results 3 and 12 weeks after transplantation. Cells expressing human albumin were found in liver sections of 8 out of 11 animals. The cells were well integrated into the liver tissue and were predominantly found adjacent to vascular structures (Fig. 3). Transplantation of NIP cells without prior tagging with PKH26 seemed to be more successful and resulted in detection of human albumin-positive cells in all four grafted animals. To analyze fusion as a possible mechanism underlying this phenomenon, immunohistochemistry studies were performed using mouse-specific and human-specific anti-albumin antibodies. In case of fusion, we expected both types of albumin to be expressed in the grafted cell. Using confocal microscopy, expression of human and mouse albumin was found in distinct cells, suggesting that NIP cells did adopt a hepatic phenotype by differentiation induced by surrounding liver tissue rather than fusion (Fig. 4). Additionally, RT-PCR was performed, demonstrating the presence of human albumin mRNA in 4 of 10 animals (Fig. 1C). In the four human albumin–positive livers, the human APRT was also amplified using RT-PCR, although the signal was less abundant than albumin (data not shown). No mononuclear cell infiltration associated with human albumin–positive cells and no neoplasm were observed 3 and 12 weeks after transplantation. In our model, the occurrence of human albumin–positive cells in general, however, was a rare event, with detection of one to three positive cells on every second slice.
Adoption of a hepatic phenotype by human NIP cells has been previously described in vitro , suggesting that these cells may represent a common hepatopancreatic precursor. In the present study, we show for the first time that cultured NIP cells from human islets of Langerhans express human albumin in vivo when transplanted into SCID mouse liver. During the expansion period, these cells were negative for the albumin transcript and thus acquired this phenotype after injection into the liver only. The immunohistochemistry results clearly show the expression of human albumin after transplantation and demonstrate that these cells are well integrated into the host liver tissue (Fig. 3), although it was in general a rare event, with one to three cells every second slice. The human albumin–positive cells were found mostly as scattered single cells, with rarely also formation of small clusters. Interestingly, we found more human albumin–positive cells after transplantation without labeling with PKH26, probably because of its known cytotoxic effect . The dose-finding experiment showed that at least 1.5 × 105 injected cells are required for detection of human albumin–positive cells, and the best results were obtained with 7.5 × 105 cells, in which human albumin–expressing cells were found in all recipient livers. Interestingly, many of the scattered cells were found adjacent to vascular structures (Fig. 3), as already described by Newsome et al.  in their study with infused human cord blood cells. In their report, the appearance of human albumin–positive cells in general was a rare event, with similar efficiency after 4, 6, or 16 weeks . Likewise, we have seen very similar results 3 and 12 weeks after transplantation, although we did not quantify the real transplantation efficiency in this proof-of-principle study. The RT-PCR studies confirmed the immunohistochemistry results in some but not all animals (Fig. 1, Table 1). This may be because of the enormous dilution of human mRNA with mouse mRNA in the RT process. The appearance of human albumin–positive cells in the transplanted livers does not necessarily indicate fully functioning hepatocytes, although albumin is the most characteristic protein synthesized by mature liver, accounting for more than 10% of total protein synthesis and the most abundant transcript in hepatocytes .
In the elegant studies by Wang et al. , a repopulation assay of FAH-deficient animals was used as a gold standard to determine the replacement of liver function by pancreatic stem/progenitor cells in the failing organ. This type of experiment, however, is not yet feasible for studies with human cells using SCID mice as recipients.
The mechanism underlying the adoption of a hepatic phenotype by grafted human NIP cells seems to be transdifferentiation rather than fusion, because expression of mouse and human albumin was found in distinct cells (Fig. 4). In a fused cell, we would expect to find expression of mouse and human albumin in the same cell. Fusion, however, cannot be excluded based on these findings alone, although it seems less likely. In particular, we cannot rule out that fusion was indeed the initial event, followed by reduction division, which restored the cell to its normal (in our case human) diploid state with expression of human albumin only. Reduction division after fusion has been described for hematopoietic cells that fused with hepatocytes . An increasing number of most recent reports demonstrates that cell fusion is a common phenomenon when hematopoietic stem cells engraft into liver tissue [15, 16] and in the central nervous system . Moreover, fusion of human hematopoietic stem cells that were injected in utero in swine were shown to yield transdifferentiation and retroviral transfer among species , indicating that fusion of stem cells may be a common phenomenon in vivo. However, transdifferentiation of hematopoietic stem cells into hepatocytes without evidence for fusion was also reported recently . Cell fusion in vivo has been described for hematopoietic stem cells but not pancreatic stem cells. The NIP cells used in our study were negative for the specific marker for hematopoietic cells CD45 (Fig. 1B).
Recently, human nestin-positive islet-derived progenitor cells have been shown to engraft into many tissues, including the liver, of immunocompetent mice  without rejection, although the investigators did not analyze the expression of tissue-specific genes of human origin. This remarkable study, however, stresses a particular characteristic of NIP cells as stem/progenitor cells that are not rejected in a xeno-transplantation setting .
In the present study, expression of human albumin in SCID mouse liver after transplantation of NIP cells is another important in vivo proof for the stem cell potential of these cells.
Human islet-derived stem cells are capable of adopting a hepatic phenotype in a SCID mouse liver in vivo, suggesting the presence of a hepatopancreatic stem/progenitor cell within or adjacent to the islets of Langerhans. The mechanism underlying this phenomenon seems to be trans-differentiation, although fusion with host hepatocyte cannot be completely ruled out. In the context of these recent findings, one could envision new therapeutic avenues for the treatment of liver cirrhosis using human pancreatic stem/ progenitor cells in which such cells could be isolated from pancreatic biopsies and expanded in vitro before trans plantation.
We thank M. Holl for expert technical assistance. This study was supported by the German Research Foundation, the German Federal Ministry of Education and Research, the Swiss National Research Foundation (404640-101232 to H.Z.) and the Juvenile Diabetes Research Foundation (5-2001-857 to H.Z.), and the Interdisciplinary Centre for Clinical Research at the University of Leipzig (01KS9504, project Z10).