Liu et al. ^[25] also published on prime editing for liver disease. They worked on the correction of the E342K mutation in the SERPINA1 gene which causes Alpha-1 antitrypsin (A1AT) deficiency. The E342K mutation is the most common one causing this disease. It is the same mutation that Schene et al. ^[24] failed to correct, highlighting that pegRNA design optimization is crucial. Liu et al. ^[25] first used PE2 to introduce the mutation into HEK293T cells. The classic PE2 requires an NGG PAM. Unfortunately, there were none near the site to be edited. Thus, they selected one that was more distant. Their pegRNA had a 20 nt spacer, a 13 nt PBS, and a 27 nt RTT. This pegRNA and PE2 achieved only 1.9% editing in HEK293T cells. However, with the same pegRNA and PE3, they achieved 9.9% editing. With this pegRNA and PE* ^[25] (PE that has a nuclear localization signal optimization), they obtained 6.4% of editing with PE2* and 15.8% with PE3*. They then attempted to correct in vivo a mutation in PiZ mice ^[25] (mouse model for the mutation E342K in the SERPINA1 gene). To do so, they delivered prime editing’s plasmid DNA by hydrodynamic tail vein injection. They obtained 2.1% editing with PE2 and 6.7% with PE2*. AAVs capacity being at most 5 kb ^[32], prime editing cannot be delivered using a single AAV. To address that issue, the authors tested to deliver the split-intein prime editors via AAVs by tail vein injection. By injecting a low dose of dual AAV8-PE3 (2 × 10¹¹ viral genome), they detected 0.6% of editing after two weeks, 2.3% after six weeks, and 3.1% after ten weeks. However, the authors hypothesized that because the PAM is far from the edited site, the efficiency of prime editing is not at its best and that the utilization of a nearer PAM will lead to greater efficiency. In the same month that Liu et al. ^[25] released their study, Habib et al. ^[26] published an article on prime editing in hiPSCs (human induced pluripotent stem cells) by generating a doxycycline-inducible prime editing platform. They also attempted to correct by prime editing the E342K mutation (1024 G > A) in the SERPINA1 gene in hiPSCs derived from a patient with alpha 1-antitrypsin (A1AT) deficiency. First, with their doxycycline-inducible prime editing platform using PE3, the authors demonstrated that creating hiPSC cell lines containing a desired mutation for the HEK3 and RNF2 locus was possible. They obtained similar conclusions by comparing their pegRNA constructs for these loci with the results in hiPSCs of David Liu’s group (the creators of prime editing) in HEK293T cells. For example, both groups concluded that, compared to the HEK3 locus, efficient prime editing at the RNF2 site required a longer PBS sequence. That conclusion suggests that the optimum PBS length depends on the sequence and, thus, might have the same optimal length even when used in different cell types. Moreover, transition substitutions (position +1 to +33), transversion substitutions (position +1 to +33), small indels (1–3 bp at +10 position), large deletions (up to 80 bp) and large insertions (up to 42 bp) are possible at the HEK3 target site in hPSCs. However, Habib et al. ^[26] failed to introduce small indels at the +17 and +21 positions for a mutation in the RNF2 gene. Another exciting aspect of Habib et al.’s ^[26] study is the section explaining why the PE3 system generates indel mutations. To understand what the main factor was, the authors tested five different scenarios: a single DSB induced by wild-type (wt) Cas9 and a sgRNA, a single nick induced by nCas9-RT and a sgRNA, a double DSB induced by wt Cas9 and two sgRNAs, a double nick induced by nCas9-RT and two sgRNAs, and finally the PE3 system (double nick induced by nCas9-RT, one pegRNA, and one sgRNA). From this experiment, the authors concluded that the unintended indels are not directly caused by double nicking but by the combinatory activity of the RT and the pegRNA. In November 2021, Genesis Lung released his thesis ^[27] on the correction by prime editing of the E342K mutation in the SERPINA1 gene that causes the A1AT deficiency. Lung worked on the same mutation as Schene et al. ^[24], Liu et al. ^[25], and Habib et al. ^[26], whose work has been described above. The last three mentioned authors failed to demonstrate a detectable correction of the mutation ^[25]. They used a PE that required an NGG PAM, which caused the targeted nucleotide to be far from the cut site, thus hindering good efficiency. In their original paper, Anzalone et al. ^[20] proposed an RTT length between 10 and 16 nt, which is much shorter than what Liu et al. ^[25] used (27 nt) ^[26]. Lung ^[27] attempted to correct the E342K mutation in the SERPINA1 gene using a prime editor using another PAM. He first designed a variant of the prime editor that could use different PAMs (NGG, NGA, NGC) and conjugated them with pegRNAs that had different PBS (13 or 17 nt) and RTT lengths (between 10 and 20 nt). Liu reached up to 3% editing for the variants using an NGA PAM in HEK293T cells containing a lentivirus cassette with the E342K mutation. The best pegRNA had a spacer of 20 nt, a PBS of 13 nt, and an RTT of 20 nt. With PE3, he reached 3 to 5% editing. Those experiments were conducted by transfecting plasmids DNA with Lipofectamine 2000. He also tried to transfect the prime editing components in their RNA form, which can improve efficiency ^[33]. However, in this case, the efficiency was not significantly different between the use of DNA or RNA. He next tried to correct the E342K mutation in human primary fibroblasts. By using NGA-PE2, he obtained 1.99% of editing. In his thesis, Lung also described that overexpression of the three prime repair exonuclease 2 (TREX2) gene might decrease prime editing efficiency. It has been shown that this nuclear protein has 3′–5′ exonuclease activities ^[34]. Non-engineered pegs being really sensible to degradation by exonuclease, the presence of TREX2 may decrease prime editing efficiency by degrading pegs from their 3′ extremity, thus reducing the available number of pegs. Hence, the knockout of this gene may help prime editing. It is to note that since Lung’s master’s thesis is not a peer-reviewed article. Therefore, those results should be taken with caution. Böck et al. ^[28]. worked on the editing of two other loci, dnmt1 and Pah^enu2, both of which code for proteins expressed in the liver. They first tested two variations of the prime editor, the intein-split PE2∆RnH ^[28] and its unsplit version. AAV8 delivered the split version, and the unsplit version, being larger, was delivered by human adenoviral vector 5 (AdV). The latter virus can contain a larger cargo but is much more immunogenic. With the split version, they achieved 15% editing in vivo of the dnmt1 locus. Since they assumed that the unsplit version would give a better result, the authors also tested it in vivo, and they obtained much better results when the prime editor was not split. In neonatal mice, they obtained 58.2% of editing; in adult mice, they obtained 35.9%. Next, they tested the correction of a mutation at the Pah^enu2 locus (F263S, c.835T > C), this mutation leading to phenylketonuria, a liver disease. They first tested their pegRNAs in an HEK293T cell line that had stably integrated exon 7 of the Pah^enu2 allele. Their two best constructs had a spacer of 20 nt, a PBS of 13 nt, and an RTT of 16 or 19 nt. These two pegRNAs resulted in an editing of nearly 20%. The one with a 19 nt RTT had fewer off-targets; this pegRNA was thus chosen for the in vivo experiments. For the in vivo experiments, a mouse model of phenylketonuria (Pah^enu2 mouse model ^[28]) was used. First, with the split version of the prime editor delivered by AAV8, they obtained meager results (less than 1% with PE2 and less than 2% with PE3). They then used the unsplit version of the prime editor delivered by an AdV. With the PE2 version in adult mice, they obtained only 2% editing. However, the results were better when they treated neonate mice. With PE2, they obtained 6.9% editing, and with PE3, they obtained 11.1% editing. This percentage was sufficient to lead to a therapeutic reduction of blood phenylalanine, without even inducing detectable off-target mutations and without leading to prolonged liver inflammation. Their results are encouraging. This project’s major problem in pursuing clinical translation is that they deliver a massive dose of virus (7 × 10¹⁴ vector genome/kg). In addition to being very expensive, using that amount of virus results in the induction of the immune system. They however noted that the percentage of edited hepatocytes was maintained even 12 weeks after the injection. Jiang et al. ^[29] optimized prime editing to delete and replace long genomic sequences. The authors combined the PE with two pegRNAs, and they called it the PE-Cas9-based deletion and repair (PEDAR) method. They tested their method to remove a 1.38 kb pathogenic insertion in the FAH gene in a mouse model of tyrosinemia and replace it with a 19 bp sequence. They successfully delete the fragment and repair the deletion junction to restore FAH expression in the liver. They detected FAH-expressing hepatocytes on PEDAR-treated liver sections and obtained 0.76% of correction. Even if this is low, edited hepatocytes gained a growth advantage and eventually repopulated the liver. Indeed, forty days after the treatment, widespread FAH patches were observed in PEDAR-treated mouse liver sections and edited hepatocytes showed normal morphology. Kim et al. ^[30] also worked on hereditary tyrosinemia type 1. They studied the correction by prime editing of a G > A point mutation (position +10 in their pegRNA) in the FAH gene in chemically derived hepatic progenitors (CdHs) from a mouse model of hereditary tyrosinemia (HT1 mice) ^[30]. After the treatment by prime editing, they grafted those cells into the liver of HT1 mice to study the repopulation of the liver by the corrected cells. First, the authors generated CdHs from HT1 mice. Next, they electroporated the cells with PE3, a sgRNA nicking at position -4, and a pegRNA that contained a spacer of 20 nt, a PBS of 11 nt, and an RTT of 15 nt. They obtained 2.3% editing without any off-target effects. Next, these authors tested the possibility of an ex vivo therapeutic transplantation of a corrected HT1-mCdHs-PE3b (chemically derived hepatic progenitors that are from HT1 mice and that have been treated with PE3b) cell population into the livers of HT1 mice. Because the bulk population of cells had a sufficient editing efficiency, they did not need to isolate cell clones. They thus directly grafted the bulk population of treated cells in the liver of HT1 mice. HT1-mCdHs-PE3b transplanted mice survived for more than 160 days, compared to control mice injected with PBS that died before 90 days. Liver damage of transplanted mice had also significantly decreased. After 140 days, the authors showed that the FAH-positive cell population in the HT1-mCdHs-PE3b transplanted liver repopulated the liver. Because of the repopulation of those cells, the percentage of edits in the liver increased from 2.3 to 34.3%. Jang et al. ^[31] also studied the correction by prime editing of a mutation in the FAH gene. They worked on the G > A point mutation at the last nucleotide of exon 8. This mutation causes exon 8 skipping and results in loss of function of FAH, which causes hereditary tyrosinemia type 1. They first tested some pegRNAs in mutant Fah target sequence–containing HEK293T cells. The best one had a spacer of 20 nt, and the edit was at position +10 from the cut site. By using PE2, they obtained 18.7% of editing in vitro. They next tested their pegRNA in vivo in a mouse model of hereditary tyrosinemia type 1 (Fah^mut/mut). The authors tested both PE2 and PE3. The first good news was that treated mice survived till the end of the experimental period (40 days for experiments with PE3 and 60 days for experiments with PE2), unlike the sick control mice which showed substantial weight loss and died before day 30 of the experiment. With these results, they demonstrated that the treatment prevented the mice from losing weight and prolonged their survival. Quantitative analyses were conducted following the mouse sacrifice. In the PE3-treated Fah^mut/mut mice, they observed that 12% of the Fah mRNA contained the exon 8, compared to the control Fah^mut/mut mice that did not have any. The frequency of FAH+ cells in the liver was also quantified. At day 40, PE3-treated mice had on average 61% FAH+ cells. At day 60, PE2-treated mice had on average 33% FAH+ cells. They also determined the percentage of editing by deep sequencing. They observed 11.5% editing in PE3-treated mice and 6.9% in PE2 treated-mice. This percentage was lower than the frequency of FAH+ cells because most hepatocytes are polyploid ^[35] and because hepatocyte DNA was mixed with nonparenchymal cell DNA. Therefore, the frequency of FAH+ cells is more indicative of clinical improvement. The authors also investigated the presence of indels at or near the targeted nucleotide and at potential off-target sites of the pegRNA. No off-target mutations were detected. In short, for liver diseases, prime editing has been used to correct mutations causing DGAT1-deficiency, bile salt export pump deficiency, alpha-1-antitrypsin deficiency, phenylketonuria, and tyrosinemia type 1. Prime editing has also been used to generate cell or animal models for liver cancer, bile salt export pump deficiency, alpha-1-antitrypsin deficiency, and a liver disease caused by a mutation in the DNMT1 gene.