*2.3. Prospects of AAV-Mediated Gene Therapy for PFIC1, 2, 4, 5, and 6*

The feasibility of in vivo AAV-mediated non-integrating gene therapy has only been reported for PFIC3. In this disease, the detergent activity of bile salts, that causes the pathology, can be neutralized by partial correction of phosphatidylcholine levels in bile [64]. Establishing ABCB4 expression in a sufficient percentage of the hepatocytes will prevent further damage, halt disease progression and stop hepatocyte proliferation, thereby, preventing the loss of episomal AAV vector genomes caused by cell division [55,56]. AAVmediated in vivo gene therapy using non-integrating strategies seems less feasible for other types of PFIC.

In the case of PFIC1 and 2, their pathophysiology is driven by individual cellular stressors. These stresses are caused by decreased membrane integrity (in PFIC1) [65] and intracellular bile salt accumulation (in PFIC1, and 2) [66]. Therefore, all hepatocytes need to be corrected to stop damage induced hepatocyte proliferation. The high vector dose to ensure transduction of all hepatocytes may cause liver toxicity. The subsequent proliferation to compensate hepatocyte loss will result in loss of non-integrated AAV vector copies causing loss of correction. For both types an integrating gene therapy strategy, preventing the loss of the vector genome upon cell division, seems a pre-requisite for sustained correction. The current state of the art integrating AAV-mediated gene therapy systems, as described above, have a low efficacy but when performed in neonatal mice the percentage of corrected cells can go up to 24% of total hepatocytes [62]. This indicates therapeutic correction in PFIC1, 2 and 3, seems feasible because of the survival benefit of cells with a correctly integrated copy of the relevant gene. In contrast to the episomal AAV vector genomes, these integrated genes are copied during cell division and transferred to both daughter cells. Due to this transfer of the survival benefit to their descendants a limited number of corrected hepatocytes will repopulate the liver as shown for transplanted hepatocytes [52,67].

The pathophysiologic mechanisms causing liver damage in PFIC4, 5 and 6 have been clarified more recently. In PFIC4, the lack of functional TJP2 results in claudin mislocalization leading to paracellular bile leakage [68]. Patients with PFIC5, are deficient for the nuclear FXR, that has a central role in bile acid (BA) synthesis and homeostasis. Since FXR signaling is also needed for the expression of both ABCB4 and ABCB11, its deficiency results in lack of expression of these two transporters resulting in intracellular BA accumulation as seen in PFIC1 and 2 [69]. Recently, PFIC6 has been described to be caused by bi-allelic missense mutations in *MYO5B*. This protein plays a central role in intracellular transport of membrane proteins and these MYO5B mutants lead to ABCB11 mislocalization and, as a consequence, results in intracellular accumulation of bile salts [7,9]. Based on the pathophysiology of these three disease types, only integrating gene therapy strategies seem suited as possible treatment because the survival benefit of corrected hepatocytes may lead to liver repopulation and subsequent correction of the disease.

Animal models are essential to investigate the feasibility by showing proof of concept of gene therapy for these types of PFIC. For PFIC4, a TJP2 deficient model is needed. The whole body knock-out has been generated but appeared to be lethal during the embryonal period [11]. Therefore additional models, such as a hepatocyte specific knockout by crossing a *TJP2* gene floxed mouse with an Alb-*Cre* mouse or even a conditionally inducible model need to be generated [70]. FXR-deficient mice have been generated, but in addition to a liver phenotype, this model displays a wide array of symptoms in agreement with the central role of this nuclear receptor in many processes. In addition to impaired liver regeneration, increased hepatic tumorigenesis and cholestasis, intestinal pathology, atherosclerosis and neurological malfunction is seen [71–74]. This suggests that gene therapy that only targets the liver may be partly therapeutic. Mouse models for MYO5B-deficiency have been generated and the predominant symptom in this model is Microvillus Inclusion Disease (MID). A recent study shows that the MID mouse, having total body knock-out of MYO5B, is not suitable to model cholestasis. The aberrant protein trafficking to the apical membrane of hepatocytes, resulting in the PFIC6 phenotype, is caused by missense mutations affecting the motor domain but not by complete MYO5Bdeficiency [75]. The presence of wildtype MYO5B partially corrects the mislocalization of apical proteins in a hepatoma cell line, suggesting gene addition therapy seems a feasible approach. Proof of concept for this strategy would require a new mouse model expressing one of the specific missense mutations identified in PFIC6 patients.

The genome of wild-type AAV consist of a 4.8 Kb long single stranded DNA chain. Although AAV vectors can package a somewhat longer genome, longer genomes result in less efficient packaging or packaging of partial constructs. The maximal capacity that allows efficient packaging is limited to 5.2 Kb [76]. The coding sequences for the proteins deficient in the different types of PFIC are 3753 bp for hATP8B1 (PFIC1), 3963 bp for hABCB11 (PFIC2), 3858 bp for hABCB4 (PFIC3), 3570 bp for hTJP2 (PFIC4), and 1458 bp for hFXR (PFIC5). In PFIC6, the canonical spliced MYO5B variant consist of 5544 bp, but smaller splice variants consist of 2889 and of 1257 bp. As only specific missense mutations in the motor domain of MYO5B cause cholestasis, gene addition therapy of such a smaller splice variant, if functional, may in theory be an option, but only in combination with deleting the expression of the endogenous mutated MYO5B protein. The MYO5B domain binding Rab11 plays a crucial role in the mislocalization of apical proteins in hepatocytes resulting in PFIC6 [75]. This suggests that using AAV-mediated gene therapy to knock-out the expression of this domain using for instance CRISPR/Cas could be an option. Such a treatment would not require homologous repair nor the delivery of a donor template, that both limit the efficacy of in vivo gene correction therapy. Importantly, besides the transgene, AAV constructs consist of one poly A consensus sequence, a promoter, when using a non-integrating strategy, or two homologous arms, in the case of an integrating approach, and the inverted terminal repeats needed for packaging. These requirements make the development of both integrating and non-integrating therapeutic constructs within the limited packaging capacity of AAV challenging, but feasible for PFIC1 to 5. The

absence of intrahepatic cholestasis in patients with complete MYO5B deficiency, suggest that a gene therapy that blocks the expression of MYO5B with motor domain missense mutations may partly correct the mislocalization of apical proteins and in theory could be a feasible approach to treat PFIC6.
