*2.2. Pre-clinical AAV-mediated Gene Therapy for PFIC3*

ABCB4 deficiency seems the best option to investigate the feasibility of gene therapy as a treatment option for PFIC. This is because of the availability of a suitable animal model and a previous study showing that partial correction by hepatocyte transplantation provides therapeutic correction [52]. The *Mdr*2 <sup>−</sup>/<sup>−</sup> (*Abcb4*−/−) mouse has been demonstrated to be a relevant model for PFIC3 [10,53]. The severity of the pathology depends on the mouse strain, with FVB mice displaying a more severe phenotype compared to the C57Bl/6 mice. Adjustments via the diet, such as cholate supplementation, increase the mouse bile hydrophobicity which worsens the disease phenotype [54]. In a recent study we investigated the feasibility of AAV8-h*ABCB4* mediated correction in adult *Abcb*4 <sup>−</sup>/<sup>−</sup> mice with a C57Bl/6 background [55]. Upon vector administration, the efficacy was monitored over time until 6 months using dietary cholate administration to mimic the human bile toxicity [54]. This study provided proof of concept by demonstrating long-term correction, as shown by normalization of the liver damage parameters AST, Alanine Transaminase (ALT) and Alkaline Phosphatase (ALP), and the absence of fibrosis. Restoring a sufficient Phosphatidylcholine (PC) content in bile is a pre-requisite for prolonged correction using AAV. In case of insufficient correction, the ongoing hepatocyte proliferation will result in loss of AAV vectors, thereby further reducing PC presence in bile and thus increasing the liver damage. This was demonstrated by Weber et al. with *Abcb4*−/<sup>−</sup> FVB mice displaying a more severe phenotype [56]. These mice were treated with an AAV vector, consisting of a codon-optimized h*ABCB4* at week 2 after birth and the effect was monitored for 12 weeks. In males, this treatment appeared effective, but in 50% of the females the correction was lost. To overcome this, a second cohort was injected two times, first at week 2 and three weeks later a second dose was given. This protocol provided prolonged correction and is, with regard to severity of the pathological model, more comparable to PFIC3 patients. Both studies do indicate that long-term correction of a disease causing hepatocyte proliferation using non-integrating AAV vectors is feasible but only if the efficacy is sufficient.

In PFIC3 patients, the disease onset is observed in young children and therefore ideally the therapy should be given at an early stage. This complicates the use of a nonintegrating vector like AAV. Several studies have shown that the initial correction is lost over time when treating neonatal or juvenile animals [49,50]. Moreover, the induction of proliferation upon (partial) loss of correction will contribute to episomal transgene loss and will accelerate the decrease in therapeutic efficacy. This hurdle can be overcome by strategies that aim for integration of the therapeutic transgene. Siew et al. tested integration of a therapeutic construct, consisting of a liver specific promoter and a codonoptimized human *ABCB4* transgene, flanked by the piggyBac transposase short terminal repeats [57]. Co-administration of this construct with an AAV2/8 vector containing a piggybac expression cassette to juvenile FVB *Abcb*4 <sup>−</sup>/<sup>−</sup> mice resulted in the integration of the therapeutic construct providing lifelong hABCB4 expression and correction of

the disease. Integration is non-random, but transposons do cause integration in close proximity of actively transcribed gene regions, transcription start sites, and open chromatin structures [58,59]. None of the analyzed unique integration regions were linked to genes known to play a role in hepatocellular carcinomas [59]. Nevertheless, this cannot be excluded. Targeting the transgene integration to a safe genomic locus could overcome this safety issue.

An obvious target to ensure integration in an active region in hepatocytes is the Albumin locus. A recent study aimed at Homologous Directed Repair (HDR) mediated integration of a promoterless Factor IX coding region in this locus without using a nuclease [60]. As the episomal construct lacks a promoter, and a self-cleaving protein is added to the transgene, its expression is regulated by the expression of albumin. The absence of a promoter and a nuclease increases safety of this "generide" strategy. The basis of this approach is to deliver a Factor IX encoding transgene flanked by two homologous arms to the albumin gene with complimentary sequences coding for the desired integration region. Using the generide strategy Muro et al. demonstrated that this could be used to correct a metabolic disorder, albeit with a low efficacy [61]. Inducing double strand breaks in this locus will stimulate HDR and integration efficacy. In a follow up study, increased correction was established by combining the generide approach with an AAV-*saCas9* guided to the albumin locus [62]. In all studies neonatal mice were used, modelling the use early after birth, when the liver is actively growing and hepatocytes are proliferating. The latter is a major advantage because during cell division the HDR system is active whereas it is inactive in quiescent cells [63]. In this study, life-long therapeutic correction of the pathology was established with a treatment shortly after birth. Further, no off-target integrations were seen in predicted sites which underlines the safety profile of this gene modulation system [62]. This targeted integration strategy seems a feasible option to treat PFIC shortly after birth.
