Efficient human fetal liver cell isolation protocol based on vascular perfusion for liver cell–based therapy and case report on cell transplantation


  • Jörg C. Gerlach and Bruno Gridelli designed and coordinated the program, developed the methods for cell isolation and characterization, and participated in writing this article. Giovanni Vizzini and Bruno Gridelli designed and coordinated the clinical work and performed the clinical application. Giada Pietrosi coordinated the study, managed the tissue procurement, and performed the medical care and analytics for the clinical application. Angelo Luca oversaw the cell application and patient monitoring. Marco Spada, Salvatore Gruttadauria, Davide Cintorino, Giandomenico Amico, and Cinzia Chinnici performed the cell isolations. Giandomenico Amico performed the cell characterization and data analysis. Cinzia Chinnici performed the cell characterization. Toshio Miki provided training for the tissue cannulation and cell isolation and performed the literature analysis. Eva Schmelzer performed the literature analysis and the conventional isolations. Pier Giulio Conaldi coordinated the donor screening and the quality control of the cell preparation. Fabio Triolo designed and coordinated the experimental work and cell production and participated in writing this article. All the authors analyzed the data and provided useful suggestions and discussions.

  • This work was sponsored by a grant from the University of Pittsburgh Medical Center. The development of the good manufacturing practice cell procurement clean room area and the associated quality management system was supported by a grant awarded to the Mediterranean Institute for Transplantation and Advanced Specialized Therapies by the Italian Ministry for Innovation and Technologies and the Region of Sicily.


Although hepatic cell transplantation (CT) holds the promise of bridging patients with end-stage chronic liver failure to whole liver transplantation, suitable cell populations are under debate. In addition to hepatic cells, mesenchymal stem cells (MSCs) and hematopoietic stem cells (HSCs) are being considered as alternative cell sources for initial clinical cell work. Fetal liver (FL) tissue contains potential progenitors for all these cell lineages. Based on the collagenase incubation of tissue fragments, traditional isolation techniques yield only a fraction of the number of available cells. We report a 5-step method in which a portal vein in situ perfusion technique is used for tissue from the late second trimester. This method results in the high viabilities known for adult liver vascular perfusion, addresses the low cell yields of conventional digestion methods, and reduces the exposure of the tissue to collagenase 4-fold. We used donated tissue from gestational weeks 18 to 22, which yielded 1.8 ± 0.7 × 109 cells with an average viability of 78%. Because HSC transplantation and MSC transplantation are of interest for the treatment of hepatic failure, we phenotypically confirmed that in addition to hepatic progenitors, the resulting cell preparation contained cells expressing typical MSC and HSC markers. The percentage of FL cells expressing proliferation markers was 45 times greater than the percentage of adult hepatocytes expressing these markers and was comparable to the percentage of immortalized HepG2 liver hepatocellular carcinoma cells; this indicated the strong proliferative capacity of fetal cells. We report a case of human FL CT with the described liver cell population for clinical end-stage chronic liver failure. The patient's Model for End-Stage Liver Disease (MELD) score improved from 15 to 10 within the first 18 months of observation. In conclusion, this human FL cell isolation protocol may be of interest for further clinical translation work on the development of liver cell–based therapies. Liver Transpl 18:226–237, 2012. © 2011 AASLD.

Orthotopic liver transplantation is considered the only definitive treatment for more than the 25,000 patients at risk of death from chronic liver disease each year in the United States alone. Still, orthotopic liver transplantation is associated with long waiting times because of the limited availability of donor organs.1 Liver cell transplantation (CT) has been suggested to partially increase the remaining metabolic activity of a patient on the waiting list and may thus provide a bridge to gain time for the patient.2

Animal studies on portal vein CT of isolated adult hepatocytes have shown improved hepatic function, and this suggests that support for patients with chronic liver failure is possible.3 Initial clinical studies of adult hepatocyte CT have demonstrated that this procedure can be performed as a minimally invasive cell injection, but studies on the use of adult liver cells have shown various outcomes. Some major problems are associated with the use of adult human liver cells for transplantation. A primary problem is cell availability, which is limited because cells are typically derived from discarded organ transplants.4-8 Furthermore, the transplantable cell mass from 1 transplantation session is limited.9 Transplantation via the portal vein into an anatomically and pathophysiologically altered cirrhotic liver does not provide an optimal environment for cell immobilization and survival.10-13 However, successful engraftment at an extrahepatic site (eg, via the splenic artery14-16) could reduce the risks associated with the hepatic route and might provide a suitable environment. Another consideration is the fact that primary adult hepatocytes may not proliferate after CT and yield a sufficient cell mass; the use of actively dividing hepatic cells may address this issue.

An interesting alternative to the use of adult liver cells is the use of liver progenitor cells17, 18 derived from human fetal liver (FL) tissue19 donated after medically indicated abortions. Human fetal tissue has received less consideration with respect to the potential development of cell-based therapy. The standard cell isolation method for this fetal tissue consists of mechanical disruption of the tissue into small fragments and then static incubation with collagenase for the digestion of the connective tissue. This method, however, results in a yield that is less than 109 cells per tissue donation, which is approximately 10 times less per gram of tissue than the yield from the collagenase perfusion techniques used for tissues from adult organs.20 Our own experiments have yielded an average of 3.8 × 107 cells per gram of tissue and a maximum of 13 × 107 cells per gram of tissue (data not published) from 25 human fetal tissue isolations by nonperfused tissue digestion. With improved isolation techniques, however, human FL cells may become a more attractive source for studies of the development of liver cell–based therapies.

Although human FL tissue obtained during the first trimester mainly exhibits cells expressing hematopoietic21-24 and endothelial cell markers,25 the majority of the cells expressing hepatic markers appear in the early second trimester.26 The presence of hematopoietic cells could be interesting because initial clinical studies of the transplantation of bone marrow–derived stem cell marker–positive cells27 and bone marrow–derived mesenchymal stem cell (MSC) marker–expressing cells28 have indicated that both hematopoietic and mesenchymal cell populations in the human FL may also be of interest for the development of liver CT therapy.

In the late second trimester, the number of cells that can be isolated from the FL reaches a level of interest for clinical translation work. In this article, we describe the cannulation of the fetal portal vein with microsurgery techniques and the subsequent in situ vascular perfusion of human FLs at 18 weeks of gestation and later. Enzymatic liver cell isolation was performed by dynamic collagenase perfusion according to a 5-step method previously developed for the adult human liver.29 Applying this method, we were able to obtain the expected higher yields. The cell isolation results that are presented here are based on the use of liver tissues with gestational ages of 18 to 22 weeks; the experiments were performed with donated tissue from 15 medically indicated abortions. Because hematopoietic, mesenchymal, and hepatic marker–positive cells may be of interest for CT development, we evaluated the immunophenotype of the isolated cell preparations with flow cytometry (FC). Adult liver cell preparations and HepG2 liver hepatocellular carcinoma cells, an established in vitro model system for the study of human hepatocytes, were used for comparison. After we obtained institutional research review board approval and achieved reproducible results with the isolation method, we applied a cell preparation to liver progenitor CT. We report the first clinical application of the cell preparation via splenic artery transplantation in a patient with end-stage liver disease.


αFP, alpha-fetoprotein; APC, allophycocyanin; cGMP, current good manufacturing practice; CK, cytokeratin; CO2, carbon dioxide; CT, cell transplantation; CYP1B1, cytochrome P450 1B1; CYP2B6, cytochrome P450 2B6; D-PBS, Dulbecco's phosphate-buffered saline; EGTA, ethylenediamine tetraacetic acid; EpCAM, epithelial cell adhesion molecule; FBS, fetal bovine serum; FC, flow cytometry; FITC, fluorescein isothiocyanate; FL, fetal liver; GGT, gamma-glutamyltransferase; GMP, good manufacturing practice; HNF4, hepatocyte nuclear factor 4; HSC, hematopoietic stem cell; hTERT, human telomerase reverse transcriptase; IgG, immunoglobulin G; MELD, Model for End-Stage Liver Disease; MSC, mesenchymal stem cell; PE, phycoerythrin; PerCP, peridinin chlorophyll protein; RT-PCR, reverse-transcription polymerase chain reaction; SD, standard deviation; Thy-1, thymus cell antigen 1.


Human FL Tissue and Adult Hepatocytes

FLs (n = 15) were obtained from tissue donations after selective, therapeutically induced abortions at a gestational age of 18 to 22 weeks. The gestational age was calculated from the first day of the donor's last menstrual period. Tissues were collected after informed consent was obtained from each mother according to a protocol approved by the institutional research review board and the ethics committee. The donors agreed to the donations only after they agreed to the abortion procedure. Fetuses were collected and transferred to the current good manufacturing practice (cGMP) facility for human cell processing. All procedures were compliant with local and national legislation, regulations, and guidelines.30 The abortions associated with our protocol were performed by routine medical induction; the labor was induced by local prostaglandin administration. This procedure was medically indicated and was planned independently of our protocol. For this reason, our protocol required no alteration of the routine treatment. All abortions were due to medical indications. The only adaptation for our protocol was our acquisition of the fetuses for liver cell isolation from the gynecologists before they underwent the routine pathological examination; the fetuses went to the pathologists for routine analysis after liver removal. The specimens were placed into sterile bags containing University of Wisconsin liver storage solution, and each specimen was transported on ice immediately after the abortion to minimize the transfer time until cell isolation. Because we obtained the tissue from intact abdomens and removed the livers surgically under cGMP conditions, the tissue could be obtained in a sterile manner. The logistics of the transfer of the fetus to the cell isolation facility required no more than 1 hour, and our protocol excluded the use of cells that were isolated more than 6 hours prior to transplant. Cell fixation for the characterization of the cell populations was performed at time points comparable to those for clinical CT.

To compare the results of the characterization, we used adult liver cell preparations (Clonetics normal human primary hepatocytes) that were isolated from nontransplantable donor tissue [purchased in collagen-coated 6-well plates (CC-2691A) from Lonza, Ltd., Basel, Switzerland] and a HepG2 human hepatocellular carcinoma cell line (BS TCL 79, Lombardy and Emilia Romagna Experimental Zootechnic Institute, Brescia, Italy).

Fetal Tissue Procurement for Perfusion via the Portal Vein

Upon its arrival at the cGMP facility, each fetus was weighed, rinsed with an iodine solution (B. Braun, Melsungen, Germany), and placed onto a sterile surgical tray. The entire procurement procedure was performed in a sterile environment with a laminar air flow unit providing European Union GMP class A (US class 100) air quality. The abdominal and thoracic cavities were accessed through a sternotomy and a midline laparotomy with bilateral subcostal lateral extensions. Peritoneal swab cultures were initiated. The umbilical vein was identified within the umbilical cord and was cannulated with a 16- to 24-gauge cannula (Becton Dickinson Infusion Therapy AB, Helsingborg, Sweden) according to the size of the vessel, and the cannula was ligated31 with 4-0 silk (Tyco Healthcare, Gosport, United Kingdom). The liver's falciform and left coronary ligaments were sectioned, and the Arantius duct was approached by gentle lifting and rotation of the right and left lateral segments. Once it was identified and encircled with 4-0 silk (Tyco Healthcare), the Arantius duct was clamped with microclips (Teleflex Medical, Research Triangle Park, NC). Similarly, the hepatic hilus was encircled and clamped. The umbilical vein cannula was then connected to the tissue perfusion tubing (Midial SpA, Trapani, Italy), the intrathoracic and infrahepatic inferior vena cava were sectioned, and the liver perfusion was started via gravity open-loop perfusion [30-cm height for 95% air and 5% carbon dioxide (CO2) gassed solutions]. Immediately after the initiation of collagenase perfusion, the liver was completely mobilized by the sectioning of the right coronary ligament, the right diaphragm, and the hepatic hilus. A second swab culture was initiated, and the liver was then placed into a 100-mm Petri dish (Corning, Palo Alto, CA) so that the perfusion and digestion process could be continued by mechanical disruption with forceps and a cell scraper.

Five-Step Perfusion Method for Cell Isolation

The tissue dissociation protocol that was used to isolate cells was based on previously published protocols for the isolation of adult human liver cells by vascular perfusion29, 32 so that we could achieve improved tissue dissociation and thus obtain higher cell yields and improved cell viability. Perfusion solutions A to D were prepared as described in Table 1, which also details the temperature and gas values. Solution E can be used in non clinical studies. All solutions used in this study were preformulated and manufactured in compliance with GMP guidelines (Biochrom AG, Berlin, Germany). We carried out an open-loop perfusion. Step D included cGMP-grade collagenase (NB 1 GMP grade, Serva Electrophoresis GmbH, Heidelberg, Germany), which was used at a concentration of 0.08% (wt/vol). The steps of the protocol and solutions A to D are described in Tables 1 and 2. Initially, each organ was perfused in situ for 10 minutes with 150 to 200 mL of solution A [which contained 2 mmol/L ethylenediamine tetraacetic acid (EGTA)] to flush out the remaining blood and loosen the desmosomal cell-cell junctions (which depended on bivalent ions such as Ca2+ and Mg2+) by chelation. Perfusion was continued for 10 minutes with solution B, which had a composition similar to that of solution A but contained no EGTA; this was required to prepare the tissue for digestion by collagenase. Because collagenase is Ca2+- and Mg2+-dependent, these electrolytes were supplied by perfusion with solution C for 2 to 5 minutes. Collagenase perfusion with solution D was performed for 7 to 10 minutes until the organ became soft and the tissue started to macroscopically disintegrate beneath the liver capsule. The liver was excised and transferred into a 100-mm Petri dish in such a way that ex situ perfusion could be easily continued without vessel kinking. The capsule was ruptured with forceps and the mechanical disruption of the liver was completed by scraping tissue away from the biliary tree with a sterile, disposable cell scraper. The capsule material and the undigested tissue were removed by filtration of the cell suspension through a sterile 500-μm nylon mesh (Miami Aqua Culture, Boyton Beach, FL). The remaining undigested tissue on the filter was rinsed with 10 to 20 mL of cold solution D (4°C).

Table 1. Solutions for 5-Step In Situ Perfusion Isolation and Preparation for Experimental Purposes
SolutionSolution Working Temperature (°C)Solution CompositionAdd BicarbonateAdd Phenol RedAdd EGTAAdd Glucose/ Amino AcidsAdd Antibiotics
  • NOTE: Solutions intended for clinical use can be prepared and used only in the framework of regulatory body–authorized and controlled clinical studies. These solutions need to be evaluated and approved by regulatory bodies for each study. All solutions, additives, vessels, and environments used in the preparation of these solutions need to be sterile and clinical-grade and must be prepared according to GMP requirements. The osmolarity should be 300 to 310 mOsm for all these solutions. The pH value needs to be kept within the range of 7.35 to 7.45 when CO2 perfusion is being used. These solutions should be stored at 4°C for no longer than 1 week if prepared fresh. Preformulated clinical grade reagents should not be stored beyond expiration date.

  • *

    Solution A consists of 200 mL of EGTA in D-PBS. A liquid EGTA solution is preferred; if this is not available, use 1 N hydrochloric acid and adjust the pH of a 100-mL bottle of a 0.5 M EGTA solution (pH 8.0) until it reaches 7.3 to 7.5 (sterile filter). Add 2.0 mL of 0.5 M EGTA to 500 mL of D-PBS with no calcium or magnesium with a sterile serological pipette. Freshly prepare the solution immediately before each scheduled cell isolation procedure (do not use amphotericin B if the cells are for clinical applications).

  • Solution B consists of 200 mL of D-PBS with no calcium or magnesium (without EGTA), 15 mM glucose, 25 mM (2.2 g/L) sodium bicarbonate (which needs to be gassed with 5% CO2), and 11 mg/mL phenol red (do not use phenol red if the cells are intended for clinical applications).

  • Solution C consists of 100 mL of D-PBS with calcium and magnesium.

  • §

    Solution D consists of 100 mL of D-PBS with calcium and magnesium, 25 mM (2.2 g/L) sodium bicarbonate, 15 mM glucose, and sterile-filtered collagenase. The solution needs to be gassed with 5% CO2. If clinical use is not planned, 11 mg/mL phenol red and antibiotics may be added.

  • Collagenase from Clostridium histolyticum. Collagenase concentrations ranging from 0.025% to 0.1% (wt/vol) were used according to the activity; the batches varied in their activity. Collagenase perfusion for 5 to 20 minutes should suffice (gentle pressure with an atraumatic instrument will rupture the organ capsule, or fissures in the capsule will form spontaneously without any mechanical irritation); a shorter period of action is generally better for cells. For clinical use, prepare stock aliquots of collagenase (NB 1 GMP grade, Serva Electrophoresis GmbH) by resuspending 500 mg of collagenase powder in 30 mL of solution D and then freezing six 5-mL aliquots at −20°C. Freshly prepare solution D immediately before each scheduled cell isolation procedure by adding a 5-mL stock aliquot of collagenase to 100 mL of the solution and then sterile-filtering the solution.

  • We used a 1 mg/mL stock solution of Pulmozyme (Roche, Basel, Switzerland). Freshly prepare solution D by adding Pulmozyme to the solution in a 1/1000 vol/vol ratio.

  • #

    Solution E includes a 4-(2-hydroxyethyl)-1-piperazine ethanesulfonic acid–buffered culture medium (with no FBS) for longer transfer and a bicarbonate-buffered culture medium (with FBS) for immediate in vitro experiments [eg, Williams' E medium–based Heparmed Vito 143 (Biochrom) supplemented with 0.8 mg/L insulin, 5 mg/L transferrin, and 0.003 mg/L glucagon]. For clinical use, use a clinical-grade storage solution instead.

A*37–38D-PBS without calcium and magnesiumNoNo2 mMNoGentamicin (50 μg/mL); amphotericin B if desired (not for clinical use)
B37–38 (typically gassed with 5% CO2)D-PBS without calcium and magnesium25 mM (2.2 g/L) solution of sodium bicarbonate (5% CO2 gassing is required)If desired (not for clinical use)NoIf desiredIf desired (not for clinical use)
C37D-PBS with calcium and magnesiumNoNoNoNoIf desired (not for clinical use)
D§37–38 (perfused with 5% CO2)D-PBS with calcium and magnesium as well as collagenase and deoxyribonuclease to prevent cell clumping25 mM (2.2 g/L) solution of sodium bicarbonate (5% CO2 gassing is required)If desired (not for clinical use)NoIf desiredIf desired (not for clinical use)
E#15 (wash solution for in vitro experiments)D-PBS with calcium and magnesiumNoNoNoAmino acids if desired; also glucose (15 mM) and pyruvate (1 g/L)Amphotericin B and penicillin/ streptomycin if desired (not for clinical use)
Table 2. Antibodies and Dilutions Used in the FC Investigations
IgG1G18-145FITCMouse monoclonal IgG1/kUndilutedBecton Dickinson Biosciences (San Jose, CA)
IgG2PC10PEMouse monoclonal IgG2a/kUndilutedBecton Dickinson Biosciences
CD45+2D1PerCPMouse monoclonal IgG1UndilutedBecton Dickinson Biosciences
CD34+581FITCMouse monoclonal IgG1/kUndilutedBecton Dickinson Biosciences
CD117+ (c-Kit)YB5.B8PEMouse monoclonal IgG1/kUndilutedBecton Dickinson Biosciences
CD133+293/C3PEMouse monoclonal IgG2b1/10Miltenyi Biotech Bisley (Surrey, United Kingdom)
CD90+ (Thy-1)5E10FITCMouse monoclonal IgG1/k1/20Becton Dickinson Biosciences
αFP+189506UnconjugatedMouse monoclonal IgG11/40R&D Systems (Minneapolis, MN)
CK18+C-04FITCMouse monoclonal IgG1/20Abcam (Cambridge, MA)
CK19+RCK108PEMouse monoclonal IgG11/20Santa Cruz Biotechnology (Santa Cruz, CA)
Albumin188835UnconjugatedMouse monoclonal IgG2a1/20R&D Systems
CD68Y1/82APEMouse monoclonal IgG2kUndilutedBecton Dickinson Biosciences
EpCAMEBA-1FITCMouse monoclonal IgG1UndilutedBecton Dickinson Biosciences
E-cadherin36FITCMouse monoclonal IgG21/100Becton Dickinson Biosciences
CD49fGoH3FITCRat monoclonal IgG2a/kUndilutedBecton Dickinson Biosciences
Asialoglycoprotein8D7FITCMouse monoclonal IgG11/100Hycult Biotechnology (Uden, the Netherlands)
Ki-67B56FITCMouse monoclonal IgG1/kUndilutedBecton Dickinson Biosciences
Annexin V FITCMouse monoclonal IgG1UndilutedBecton Dickinson Biosciences
VimentinVI-RE/1PEMouse monoclonal IgG1/k1/100Abcam (Cambridge, United Kingdom)
hTERTY182UnconjugatedRabbit monoclonal IgG1/100Abcam (Cambridge, MA)
Oval cellOV-6UnconjugatedMouse monoclonal IgG1/k1/100Santa Cruz Biotechnology
CD29MAR4APCMouse monoclonal IgG1/kUndilutedBecton Dickinson Biosciences
CD31WM59FITCMouse monoclonal IgG1/kUndilutedBecton Dickinson Biosciences

The cell pellet was washed 2 to 3 times (until the supernatant appeared clear), and the cells were resuspended in a medium based on Williams' E medium (Table 1) and were maintained at 15°C in a 5% CO2 and 95% relative humidity environment until transplantation. Cell viability was determined with the trypan blue exclusion method. Aliquots were taken for quality-control testing (sterility, endotoxin, and Mycoplasma content) and characterization. When quality-control assays were negative, a second cell count and viability calculation was performed to ensure that the viability was still above the threshold defined in the clinical protocol (70%). As soon as the patient was ready for transplantation, the cells were collected, washed with Ringer's lactate (which was supplemented with 2% human serum albumin), resuspended in the same solution, and transferred to a 50-mL sterile syringe for infusion.

Cell Culture

Adult hepatocytes were cultured in Williams' E medium–based Heparmed Vito 143 (Biochrom AG, Berlin, Germany), which was supplemented with 5% fetal bovine serum (FBS), 2 mM glutamine, 0.8 mg/L insulin, 5 mg/L transferrin, 0.003 mg/L glucagon, 7.4 mg/L dexamethasone, 100 U/mL penicillin, and 100 μg/mL streptomycin. HepG2 cells were cultured in Dulbecco's Modified Eagle's Medium (DMEM) supplemented with 10% FBS, 100 U/mL penicillin and 100 μg/mL streptomycin. The cells were cultured at 37°C in a humidified incubator in a mixture of 95% air and 5% CO2. The microscopic evaluation of the cells was performed with phase contrast light microscopy.

Reverse-Transcription Polymerase Chain Reaction (RT-PCR)

Total RNA was isolated from the samples with TRIzol (Invitrogen) according to the manufacturer's suggested protocol. First-strand complementary DNA was synthesized from approximately 1 μg of RNA with SuperScript II reverse transcriptase (Invitrogen) according to the manufacturer's instructions. The complementary DNA was used as a template for amplification by semiquantitative RT-PCR with the housekeeping gene β-actin as an internal control. The analyzed genes, primer sequences, annealing temperatures, and sizes of the amplified fragments are listed in Table 3. The polymerase chain reaction products were separated by 1% agarose gel electrophoresis and were visualized via staining with ethidium bromide.

Table 3. RT-PCR Primer Sequences and Annealing Temperatures Used in the Investigations
Gene NameForward (5′-3′)Reverse (5′-3′)Product Size (bp)Annealing Temperature (°C)
  • *

    Housekeeping gene and control.



We used flow cytometry (FC) for the phenotypic marker analysis of the resulting cell population immediately after our isolation method was used, and a comparison was performed with HepG2 cells and primary adult hepatocyte cultures. One million cells were used for antibody staining. The fluorescein isothiocyanate (FITC)–conjugated, phycoerythrin (PE)-conjugated, or allophycocyanin (APC)-conjugated antibodies and isotype controls are summarized in Table 2. Isotype control antibodies were used as negative controls for the measurement of the nonspecific binding of the specific antibodies. Forward and side scatter gates were set to include all viable cells. Routinely, debris and doublets were excluded from the cell population data by the application of forward and side scatter selection. The coexpression of 2 markers was analyzed by the gating of the population that was positive for the first marker and the subsequent analysis of the percentages of the second marker. FC data were acquired with a BD FACS Aria II instrument and were analyzed with Diva 6.1.2 (BD Biosciences, Franklin Lakes, NJ).


Isolation Technique and Outcome

Portal vein cannulation was surgically possible with fetuses no younger than a gestational age of 18 weeks. The portal vein of an earlier tissue donation could not be cannulated without the rupture of the vessel during the onset of perfusion. With portal vein in situ perfusion, the tissue's exposure time to collagenase could be reduced to 7 to 10 minutes (minimum to maximum) instead of 30 to 40 minutes (unpublished data) with conventional tissue fragment incubation.33, 34 To demonstrate differences in remaining tissue fragments of biliary tree–associated cells after isolation with conventional collagenase incubation and the vascular perfusion method, we investigated the remaining tissue microscopically. According to a microscopic and macroscopic comparison of our method with the conventional method, the remaining vascular/biliary structures were distinctly less suspended with the conventional method of tissue fragment collagenase incubation versus collagenase vascular perfusion. Vascular perfusion resulted in almost complete digestion of the connective tissue. As a result of vascular perfusion with steps A and B, the blood cells could be removed from the tissue before collagenase digestion, and we see this as an advantage over conventional nonperfused collagenase tissue digestion.

Cell Yield and Viability

We used 15 FLs; a mean yield of 1.8 ± 0.7 × 109 cells with a hemocytometer was achieved. Occasionally occurring enucleated erythrocytes were not counted. The mean viability was 79.2% ± 5% according to the trypan blue exclusion method and 77.5% ± 6% according to FC with 7-amino-actinomycin D. In all cases, a postisolation phase contrast microscopy examination of cells that were isolated with the in situ perfusion method showed mostly single cells (>90%) along with a few small aggregates (2-12 cells) and a negligible number of larger aggregates. The typical medium-size aggregates and larger aggregates known from the conventional method were not observed. The overall apoptosis rate after isolation was 0.86% ± 0.9% (annexin V and FC), the proliferation rate was 61% ± 12% (Ki-67 and FC), and the proportion of cells positive for human telomerase reverse transcriptase (hTERT) was 20.4% ± 15% (FC).

The clinical safety screening panels for the donors and cells that we used to determine whether a cell batch from a cell isolation qualified for potential clinical studies are described in Supporting Tables 1 and 2.

FC Cell Characterization

A marker expression analysis was performed for the cells immediately after cell isolation without culturing. For comparison, we used adult hepatocytes (n = 4) and cells from the HepG2 cell line (n = 4). The results are summarized in Fig. 1.

Figure 1.

FC data for (A) hepatic markers, (B) hematopoietic markers, and (C) mesenchymal cell–associated markers in the isolated cell population (n = 15), HepG2 cells (n = 4), and primary human hepatocytes (n = 4). (D) Summary of the coexpression studies.

FC data for the isolated cells showed a heterogeneous population, which included hepatocyte markers [cytokeratin 18–positive (CK18+) cells (15.9% ± 6%), albumin-positive cells (17.4% ± 7%), alpha-fetoprotein–positive (αFP+) cells (23.6% ± 14%), and asialoglycoprotein receptor–positive cells (17.5% ± 9%)] as well as an epithelial marker [E-cadherin–positive cells (11.8% ± 10%)]. The population also included CK19+ cells (a bile duct marker; 2.8% ± 3%), CD68+ cells (Kupffer cells; 4.9% ± 1%), and cells expressing the endothelial CD31 marker (12.4% ± 12%).

Some cells (13.3% ± 8%) coexpressed CK18 and CK19, and this suggested the presence of bipotential precursors of the hepatic and biliary lineages; 5.7% ± 3%, 5.3% ± 8%, and 4.9% ± 8% of the cells expressed the hepatic progenitor markers OV-6, epithelial cell adhesion molecule (EpCAM), and CD49f, respectively. Further coexpression analysis showed that 3.6% ± 8% of the cells were positive for both αFP and CK18, whereas the proportion of αFP+/CK19+ double positives was less than 1%.

Interestingly, the cell preparation contained a relevant percentage of cells expressing markers associated with mesenchymal cells: 16.4% ± 7%, 1.5% ± 2%, and 7.6% ± 14% of the cells were positive for vimentin, CD90 [thymus cell antigen 1 (Thy-1)], and CD29, respectively. The coexpression analysis showed that 1% ± 1.3% of the cells were albumin-positive/CD90+, whereas the proportions of CD34+/CD90+, CD133+/CD90+, CD117+/CD90+, and αFP+/CD90+ were less than 1%.

FC also indicated the presence of cells expressing hematopoietic progenitor markers: CD133 (2.2% ± 1% of the cells), CD117 (c-Kit) as an early marker (1.2% ± 1%), CD34 (4.1% ± 5%), and CD45 as an overall white blood cell marker (6% ± 4%). The coexpression analysis showed that 0.6% ± 0.9% of the cells were albumin-positive/CD117+, whereas the percentage of αFP+/CD117+ was less than 1%. Finally, 1.8% ± 4% of the cells were positive for both CD34 and CD45.

Enzyme activity markers were analyzed with RT-PCR; the expression of alpha-1-antitrypsin and gamma-glutamyltransferase (GGT) was determined in all 3 cell sources (FL cells, adult hepatocytes, and HepG2 cells). The expression of cytochrome P450 1B1 (CYP1B1) was weak in fetal cells, strongly positive in adult cells, and negative in HepG2 cells, whereas the expression of cytochrome P450 2B6 (CYP2B6) was negative in fetal cells, strongly positive in adult cells, and negative in HepG2 cells; all cell types were positive for hepatocyte nuclear factor 4 (HNF4) and CK7 (mature biliary markers), CK8 (a mature hepatocyte and biliary marker), and albumin.

When the results of cell isolations for developmental ages between 18 to 19 weeks and 20 to 22 weeks were compared, statistically significant differences were seen in the expression of the Ki-67 proliferation marker (P = 0.04); that is, the percentage of Ki-67+ cells decreased as the gestational age increased (ie, 69% at 18-19 weeks versus 45% at 20-22 weeks). Moreover, the expression of Ki-67 in cells that coexpressed the hepatic and biliary epithelial markers CK18 and CK19 (P = 0.04) also strongly decreased with age (ie, 22% at 18-19 weeks versus 1.8% at 20-22 weeks). Furthermore, as the gestational age increased, the percentage of CK18+/CK19+ immature bipotential progenitors significantly decreased, and the percentage of mature hepatocytes increased, as expected. Finally, the percentage of fetal cells expressing proliferation markers was on average 45 times greater than the percentage of adult hepatocytes expressing these markers and was comparable to the percentage of immortalized HepG2 cells; this indicated the proliferative capacity of the fetal cells.

Clinical Case Report

To qualify the cell preparation for the initial clinical application, we developed a quality management system, which included risk management and standard operation procedures for cell procurement and quality control in a dedicated clean room area, in full compliance with the GMP guidelines of the European Union and the Food and Drug Administration (United States). Risk management methods for the production of clinical-grade cells were previously described in detail (CITE REF35). After we obtained approval by the ethics committee and the institutional research review board, we started a clinical trial to determine the safety and efficacy of human FL CT for the treatment of end-stage liver disease in candidates for liver transplantation. Here we report the first patient to be included in the trial after informed consent was obtained: a 57-year-old male with hepatitis C virus–related cirrhosis who had a Child Pugh score of B-9, a Model for End-Stage Liver Disease (MELD) score of 15, and recurrent episodes of portosystemic encephalopathy. The patient received 2 intrasplenic infusions of freshly isolated ABO-compatible FL cells (5 × 108) through the splenic artery by a percutaneous angiographic femoral approach on days 0 and 80 (for cell safety/quality control testing, see Supporting Tables 1 and 2). The choice of the spleen was supported by data from animal models in which liver cells were able to proliferate in the splenic pulp and maintain their metabolic functions,15, 36 and intrasplenic infusions into the splenic artery were chosen to prevent potential bleeding complications10; this provided an opportunity to use a large reservoir without potential obstacles to engraftment expected to be encountered in the cirrhotic liver microenvironment.

Both injections were well tolerated, and there were no side effects related to the radiological procedure. The patient received immunosuppression (tacrolimus) with a target blood level of 3 to 6 mg/L, but the treatment was stopped after 3 months because of severe neurotoxicity. The MELD score, which was the main clinical endpoint in the study protocol, decreased from 15 to 11 at 3 months. At the 18-month follow-up, the MELD score was 10, and the patient continued to show no signs of encephalopathy.


We describe the adaptation of a 5-step portal vein in situ perfusion method (originally developed for the adult liver29) to human FL cell isolation in order to obtain cell preparations from the human FL in the late second trimester (starting at gestational week 18). In contrast to the static collagenase incubation of tissue fragments without vascular perfusion,37 high viability rates in conjunction with the high yields known from cell isolations from adult livers can be obtained with human FLs. For the reduction of the collagenase digestion time and thus associated cell injuries,38, 39 the exposure to collagenase can be reduced from 1 to 2 periods of up to 40 minutes each (needed to incubate FL fragments of comparable ages33, 34, 37) to 1 period of 7 to 10 minutes under in situ perfusion. A further advantage is that collagenase perfusion releases most cells from the biliary tract region; these cells along with potentially interesting progenitors are often discarded during the mechanical manipulation of conventional methods. With the introduction of vascular perfusion steps A and B, blood cell contamination can also be drastically reduced.

The resulting FL-derived hepatic cell preparation was found to contain hematopoietic, hepatic, and mesenchymal progenitor marker–expressing cells. This is interesting because, in addition to hepatic cells, bone marrow–derived hematopoietic and mesenchymal cells have already been examined in studies of the development of clinical CT protocols for liver regeneration.27, 28

The expression of hepatic markers in the characterized cell preparation overlapped with the expression of markers of adult human hepatic progenitor cells to some degree, as previously reported by other groups.33, 40-42 Because of their availability, most studies of the isolation and in vitro propagation of adult liver progenitors have been performed with animal-derived cells,40 and the majority of the isolation strategies for liver progenitors are based on the in vivo activation of the progenitor cell population43 followed by the enzymatic digestion of tissue fragments and the selection of progenitors with a variety of techniques, such as centrifugation techniques,44, 45 fluorescence-activated cell sorting,46 and selective cell aggregate formation.47 Approaches to the isolation of adult progenitors from normal human livers are rarely described; cell harvesting after the collagenase digestion of tissue fragments is performed with various methods, including differential centrifugation,48 magnetic or fluorescence-activated cell sorting,33, 49 and selective conditions for liver cell cultures.50 Typically, these methods result in low cell numbers.34, 50 Because a therapeutic application for CT may be limited by the yields from isolation protocols using adult liver tissue, progenitor-rich FL tissue appears to be an interesting alternative source for CT development.

Little is known about the marker expression of hepatic precursors in the FL, and the isolation methods are sometimes based on culture behavior; for example, Suzuki et al.51 identified albumin-negative hepatic stem cells with a potential for multilineage differentiation and a self-renewing capability in the mouse FL. Our own work on the cell preparation derived with the aforementioned in situ isolation technique from human tissue obtained between the 18th and 22nd gestational weeks characterized the raw cell population. The marker expression results indicate that the method yields an interestingly high number of hematopoietic, hepatic, and putative mesenchymal progenitor cell marker–expressing cells for future potential studies on stem cell behavior (eg, the use of specifically selected cells after sorting).

The transplantation of FL cells that express hematopoietic,52 hepatic, and mesenchymal markers (eg, the preparation that we characterized) could be interesting from various points of view. Although the hematopoietic and hepatic lineages in the FL have been thought to be expressed separately, the shared expression of hematopoietic and hepatic markers has been reported, and a potential lineage relationship between nonhepatic hematopoietic cells and hepatic cells has been discussed.53, 54 In contrast to our work focusing on cells after week 18 in the late second trimester, Nava et al.26 investigated the emergence of progenitors during the earlier development of the human FL and studied the coexpression of known hematopoietic and hepatic cell markers. They reported that hepatic cells emerged after the early second trimester during human FL development and pointed out that early hepatic progenitors coexpressed the hematopoietic markers CD117 and CD34 in the first and early second trimesters. These cells could develop in vitro into cells expressing hepatic markers: isolated CD117+/CD34+/CD90 cells differentiated into cholangiocytes and hepatocytes, which expressed both the genes and the proteins for hepatic markers such as albumin, αFP, alpha-1-antitrypsin, and CK19. Our own coexpression studies confirmed the presence of CD117+/CD34+ cells (approximately 0.2%) in the late trimester FL. The presence of CD117+/CD34+ cells in the early mouse FL was also reported by Sánchez et al.55 They reported that within 11 days post coitum, the mouse liver already contained long-term repopulating hematopoietic stem cells (HSCs); all these cells expressed the CD117/c-Kit cell surface markers, and most were also CD34+. Crosby et al.49 used CD117+ selection for the enrichment of adult human liver progenitor cells, which were able to differentiate in a culture into cholangiocyte marker–positive cells; they also used CD34+ cells with similar results. Likewise, Theise et al.56 selected HSCs that differentiated into hepatocytes after their transplantation and engraftment into the livers of lethally irradiated mice. Work on human cells has described hepatic stem cells as being CD34 but CD326+ and CD44+, for example34; early mesenchymal/endodermal progenitors have been described as CD34+ and coexpressing (among other markers) CD326+, CD44+, and αFP cells49 as well as CD34+ and CD117+ cells with positivity for no hepatic marker such as albumin.57 These most likely represent hematopoietic or early mesendodermal progenitors. Because the coexpression of potential stem cell markers on liver progenitors is still not clearly defined, the human fetal isolation method presented here might be interesting for further studies of hepatic stem cell biology.

The interpretation of the investigated expression of MSC markers is hampered by the fact that liver-derived MSC markers are still under discussion. Clinical protocols for the use of MSCs in liver therapy development are mostly based on cells isolated by their culture attachment behavior and morphology instead of their marker expression.58, 59 The FL cell population that we characterized with FC contained cells that were vimentin-positive, CD90+, and CD29+. CD29 is often associated with MSCs,60 but it may also represent a surface maker of adult fibroblasts. This is comparable to vimentin, which is also a marker for adult fibroblasts from other organs. Dan et al.61 isolated vimentin-positive, CD34+, CK18+, CK19+, αFP, and albumin-negative cells from human FLs that were able to differentiate toward hepatic and mesenchymal lineages. Haruna et al.62 studied the expression of intermediate filaments during human liver development and proposed a temporary coexpression of vimentin with CK19 and CK14 during hepatoblast differentiation toward the biliary cell lineage. Little is known about whether CD90/Thy-1, which has been described for adult hepatic progenitors in rodents,63-65 characterizes human hepatic progenitors (as described in 1 case50), although it is often not found in the human liver.34, 48, 66 Nava et al.26 could not find CD90 in human fetal hepatic progenitors from the first trimester, but they did find it in cells from the second trimester that coexpressed CD90 and hepatic markers (as we found in low percentages for the late second trimester). In contrast, Fiegel et al.64 found CD90 expression during all stages of rat liver development but at higher levels only in the early developmental stages. CD90, however, has also been used by many authors to identify human MSCs in nonhepatic tissue; in a literature overview, Kolf et al.60 reported that CD90 was often described as negative in murine MSCs but clearly positive in human MSC populations. Our CD90 coexpression analysis of hepatic and hematopoietic markers showed that CD90 was coexpressed in both lineages at percentages comparable to the overall CD90 percentage. A more systematic study of the potential role of CD90 as an MSC marker on human FL cells would be interesting, and such a study might also consider the inclusion of other MSC markers such as CD73 and CD105.60

In addition to addressing biological questions about cells that can be found in the developing human liver, their isolation and characterization have implications for the advancement of therapeutic approaches to liver disorders. A problematic issue is the availability of fetal hepatic cells in numbers greater than 108 (the cell number usually used for studies of adult hepatocyte transplantation).4 Implementing these conventional methods would require tissue pooling. With the isolation method described here, the average cell number was greater than 109, and the mean viability was greater than 77% of the raw cell population. Therefore, we believe that a technique involving portal vein cannulation and this 5-step in situ perfusion method is promising for the isolation of human FL cells that might be candidates for in vitro expansion67 and could enable larger CT studies. With a focus on donations from induced abortions, the use of an intact fetus with an intact abdomen is enabled, and this addresses issues of sterility, which is problematic with tissues obtained by vacuum aspiration.68

Our clinical case report on splenic artery CT with the human FL preparation in a patient with end-stage chronic liver disease can provide only very limited information. Because the available fetal tissue for such preparations is limited for broader clinical use by the low numbers of donors, this approach can be seen only as a preliminary examination of the feasibility of the clinical use of human FL liver cells. Larger clinical studies may be possible only if in vitro expansion technologies are advanced or if selected and transplanted hepatic stem cells are proven to expand to a sufficient number in vivo.

In conclusion, we have described a human FL cell isolation method that could be of interest for future regenerative medicine research, for in vitro drug toxicity/metabolism studies, and for the development of cell expansion systems. This could potentially lead to in vitro expansion studies and provide an innovative cell source for CT and extracorporeal liver support.69


The authors thank Dr. L. Alio and Dr. R. Ferraro (Department of Gynecology and Obstetrics, Civico Hospital, Palermo, Italy) for their help and support with donated tissue procurement. The authors also appreciate the technical assistance of S. Pasqua.