Most of maximal human cell engraftment was described on two transgenic models with highly immunosuppressed mouse models, uPA+/+
mice and Fah−/−
mice. This engraftment could become sufficient to talk of “humanized liver,” with human hepatic functions and human pharmacological proprieties [12
The first is the uPA+/+ mouse model. To our knowledge, there was no publication of human liver cell transplanted in a Nude mouse uPA model.
The second transgenic Fah−/−
mouse model was used by Azuma et al. [27
] in Fah−/−
Nude mice model to create humanized liver, but they did not obtain sufficient human cell engraftment in this model and had to use a more immunosuppressed transgenic model (Fah−/−
- mice) reinforced by uPA-adenovirus administration intravenously 24–48 hours before human hepatocyte transplantation, associated to NTBC withdrawal over five days after transplantation.
To our knowledge, no success in humanized livers in Nude mice has been reported, most likely because of an insufficient immunosuppression of this mouse.
In other uPA+/+
immunosupressed mouse models, such as Rag2−/−
mice, human hepatocytes can engraft, although at lower rate than murine hepatocytes [79
]. In such a model, Dandri et al. [28
] reported 8 weeks after human hepatocyte transplantation a 2 to 10% of repopulation by human cells. In the Fah−/−
mouse model, human hepatocytes could repopulate Fah−/−
mouse liver with a range from 5 to 34%, 12 weeks after transplant [26
But these two models have specific disadvantages, directly linked with their concept; uPA+/+
mice, because of their inborn metabolic abnormally, have a poor breeding efficiency, a quite narrow window in time to perform transplantation (on neonatal age, between1 to 3 weeks of age, before they die of severe bleeding), and a renal disease that still exists despite hepatocyte transplantation [19
]. They can return to wild type by inactivation of the gene [59
]. Moreover, there is a continuous and progressive hepatic parenchyma injury, possibly via activation of plasminogen and modified activity of matrix metalloproteinase: thus, this modified metabolism can interfere with liver cell growth and distort a physiopathologic model [30
Fah deficient mice also have inconveniences: their metabolic pathway leads to development of liver hepatocellularcarcinomas [82
], requiring treatment by NTBC-diet repeated cycles to prevent tumor formation and maintain long-term survival. This diet cancels the natural selective advantage that triggers xenogenic cells proliferation and can give possible bias in results interpretation. To increase human hepatocyte repopulation efficiency, some teams use a transfection of uPA gene, which adds the same issues as encountered in uPA+/+
Another strategy is represented by liver suicide model: it consists to transfect recipient liver hepatocytes with an apoptotic gene; this gene being under the control of herpes virus type 1 thymidine kinase (HSV TK): it can be activated by administration of Gancyclovir, and thus could destroy specifically cells targeted with this suicide gene [83
Hasegawa et al. [30
] obtained an NOG (NOD/SCID/Il2Rg−/−
mouse expressing HSV TK transgene in their liver. By inducing apoptosis of liver recipient cells five days before transplantation of human hepatocyte, they observed a high index of repopulation (average up to 43%), correlated with elevated human albumin in plasma, and functional human hepatocyte.
Douglas et al. [31
] who used uPA-SCID model optimized it to achieve total replacement of uPA+/+
hepatocytes by human hepatocytes, by associating mouse liver suicide to uPA-SCID model, so that Ganciclovir treatment could induce conditional selective murine hepatocyte death in humanized SCID-uPA mouse liver. But unexpected, mice survival was not increased by humanized liver, whatever the repopulation index (32 to 87%).
As illustrated by the results of Douglas et al, the limits of the humanized liver reside intrinsically in the principle of xenogenic transplantation: whatever the models, even in the best repopulation performance such as uPA+/+
transgenic mouse, xenogenic murine models gave best engraftment results compared to human cells transplantation [29
]. Interestingly, it does not seem possible to avoid a percentage of nearly complete failure of human cells engraftment in mouse liver, for each experiment, even in the studies reporting very high repopulation index [28
This disparity between animals of the same study could be due to imperfect immunosuppression [84
] or inadequation between murine and human metabolism [74
In favor of the first hypothesis, Tateno et al. [85
] showed that when human hepatocyte engraftment in uPA-SCID mouse results in more than 50% repopulation, this high repopulation index leads to death of recipient. This mortality can be corrected by a treatment that blocks human complement factor activity.
In favor of the second hypothesis are the results of Su et al. [26
] of a failure to induce chimerism in a Fah−/−
mouse transplanted with human hepatocyte, in 6 over 14 mice despite an immunosuppressive drug (FK506). Also in favour of this hypothesis is the observation that mouse survival was not increased by humanized liver, whatever the repopulation index (32 to 87%), in the uPA-SCID model of liver failure challenged with liver suicide gene activation [31
]. There seems to be an incompatibility directly linked to animal species differences: although human and murine cells can create narrow cellular junctions, morphologically and architecturally subnormal links confirming integration of xenogenic cells but human cells will develop unexplained glycogen storage or steatosis anomalies [81
Therefore, the hypothesis that humanized liver would allow performing experiments not feasible on humans thus helping to predict pharmacotoxicological and pathobiological effects in humans needs additional demonstration. But, even in this case, transgenic models would mimic only limited aspects of human clinical liver failure, and one should be aware that humanized liver may not be transposable to real human physiology.
As a last comment, the use of an immunodeficient animal model could also be a bias in these liver cell transplantation studies. In fact, immune cells could modulate liver regeneration [87
]. Their absence in immunodeficient animals could modify the liver response to acute injury (demonstrated by Strick-Marchand et al. [90
]) and to chronic injury, as well as engraftment of hepatic cells. In the other hand, the use of immunosuppressive treatment for transplanted nonimmunodeficient animal models could also interact with liver physiology [91
] and become another bias.
This raises the question of the bias of using xenogenic models, with respect to the risk taken by xenogenic animal models not reflecting human physiology.