Adeno-Associated Virus Serotype 8-Mediated Genetic Labeling of Cholangiocytes in the Neonatal Murine Liver.

Determination of the cellular tropism of viral vectors is imperative for designing precise gene therapy. It has been widely accepted that transduction of hepatocytes using adeno-associated virus serotype 8 (AAV8) is a promising approach to correct inborn errors in neonates, but the type of neonatal hepatic cells transduced by AAV8 has not been thoroughly investigated. To address this question, we used a reporter mouse that carries Cre recombinase (Cre)-inducible yellow fluorescent protein (YFP). Our analysis primarily focused on cholangiocytes, given their pivotal roles in normal liver function and disease. We treated RosaYFP/+ mice at postnatal day 2 (P2) with AAV8-cytomegalovirus (CMV) promoter-Cre and analyzed livers at P10 and P56. The vast majority of HNF4α+ hepatocytes were labeled with YFP at both time points, and 11.6% and 24.4% of CK19+ cholangiocytes were marked at P10 and P56, respectively. We also detected YFP+ cells devoid of hepatocyte and cholangiocyte markers, and a subset of these cells expressed the endothelial and fibroblast marker CD34. Next, we used the hepatocyte-specific thyroxine-binding globulin (TBG) promoter. Surprisingly, AAV8-TBG-Cre marked 6.8% and 30.9% of cholangiocytes at P10 and P56, respectively. These results suggest that AAV8 can be a useful tool for targeting cholangiocytes in neonatal livers.

Previous studies have demonstrated that the transduction efficiency and tropism of AAV vectors depend on several variables [8, 14,15]. These include capsid serotypes, promoters, vector dosage, routes of injection, and the timing of injection, suggesting the importance of characterizing responses to AAV in the neonatal liver. It is becoming increasingly clear that non-hepatocytes in the liver such as cholangiocytes, endothelial cells, and fibroblasts contribute to normal liver function as well as disease progression [16][17][18][19][20], implying that non-specific infections of non-parenchymal cells may lead Pharmaceutics 2020, 12, 351 2 of 9 to side effects. Conversely, if AAV8 transduces non-parenchymal cells, this may lead to a broader application of AAV8 vectors. Therefore, a precise understanding of the cellular tropism of AAV8 can allow for improved study design and gene therapy. However, whether AAV8 has the capability to target non-hepatocytes at neonatal ages remains unclear.
Therefore, we hypothesized that AAV8 can be used for transducing non-hepatocytes in the neonatal liver and aimed to characterize the cellular tropism of AAV8 using Cre recombinase driven by the cytomegalovirus (CMV) and thyroxine-binding globulin (TBG) promoters [21]. The CMV promoter leads to ubiquitous expression, while the latter has been used to target hepatocytes [6,21]. Given the critical role of biliary epithelia in liver disease and normal bile flow [17,20], this study aimed to determine hepatic cell types transduced by AAV8 following neonatal injections with a focus on cholangiocytes (biliary epithelial cells). Our study reveals a novel use of AAV8 (i.e., a tool for the genetic labeling of cholangiocytes in the neonatal murine liver).

Immunofluorescence
Paraffin-embedded formalin-fixed sections were dewaxed and rehydrated sections were subjected to antigen retrieval followed by incubation in blocking solution (3% normal donkey serum and 0.25% triton X-100 in phosphate-buffered saline) for 1 hour at room temperature. Next, sections were treated with primary antibodies (Table 1) for overnight at 4 • C and secondary antibodies for 2 h at room temperature. 4 ,6-diamidino-2-phenylindole (DAPI) was used for nuclei staining.

Quantification and Data Analysis
In P10 livers the average numbers of cells analyzed per animal were 4448 ± 443 hepatocyte nuclear factor 4 alpha (HNF4α)+ hepatocytes and 398 ± 105 cytokeratin 19 (CK19)+ cholangiocytes (mean ± SD). In P56 livers the average numbers of cells analyzed per animal were 14,859 ± 2421 HNF4α+ hepatocytes and 1076 ± 246 CK19+ cholangiocytes (mean ± SD). The percentage of YFP-marked cells within the HNF4α+ and CK19+ populations were calculated, and all data were expressed as mean ± SD. Image J (National Institutes of Health, Bethesda, Maryland) was used for cell counting [25]. Prism 8 (GraphPad Software, Inc., San Diego, CA, USA) was used for Student's t-test and Mann-Whitney U test. A p-value of smaller than 0.5 was considered significant in this study.

Neonatal Injection of AAV8-CMV-Cre Labels Cholangiocytes
To determine the transduction specificity of AAV8 in neonatal livers, we used Rosa YFP/+ reporter mice [22]. Although the Rosa26 locus is ubiquitously expressed, transcriptional stop sequences block expression of yellow fluorescent protein (YFP). Removal of floxed stop sequences by Cre recombinase leads to permanent labeling of transduced cells with YFP. Therefore, we treated postnatal day 2 (P2) Rosa YFP/+ mice with AAV8-CMV-Cre and livers were analyzed at P10 ( Figure 1A). To determine the efficiency of transduction, we performed immunostaining for YFP, hepatocyte marker HNF4α, and cholangiocyte marker CK19. HNF4α and CK19 expressions were mutually exclusive, and the majority of hepatocytes were labeled as expected, while the control vector AAV8-CMV-RFP did not lead to YFP expression ( Figure 1B,C). Surprisingly, we also detected Cre/YFP-marked cholangiocytes at P10 with 11.6% ± 7.8% (mean ± SD) labeling efficiency ( Figures 1C and 2).   To determine whether cholangiocytes remain labeled at a later time point, we treated Rosa YFP/+ animals with AAV8-CMV-Cre at P2 and analyzed the liver at P56 ( Figure 1A). While most hepatocytes were labeled at this time point, 24.4% ± 7.5% of CK19+ cells were also labeled ( Figures 1D and 2). Interestingly, there was a statistically significant increase in the percentage of labeled cholangiocytes from P10 to P56 (Figure 2). YFP-labeled cholangiocytes also expressed additional markers for cholangiocytes, epithelial cell adhesion molecule (EPCAM) and osteopontin (OPN) (Figure 3) [26][27][28].
Our results indicate that neonatal injection of AAV8 can be used to transduce a substantial fraction of cholangiocytes.

AAV8-CMV-Cre Labels CD34-Expressing Cells
We observed that a subset of YFP-marked cells expressed neither HNF4α nor CK19 ( Figure 1C, white arrow). Therefore, we performed immunostaining of P10 livers for additional markers (Figure 4). Livers were stained for the fibroblast marker alpha smooth muscle actin (αSMA) and CD34 that is expressed by the endothelium of central and portal veins as well as portal fibroblasts [16,29]. Although rare, we confirmed the presence of CD34+αSMA-cells labeled with YFP. On the other hand, none of the CD34-αSMA+ cells were marked, suggesting that AAV8 transduces distinct subsets of cells. This result indicates that AAV8-CMV-Cre labels CD34-expressing cells in addition to cholangiocytes.

AAV8-CMV-Cre Labels CD34-Expressing Cells
We observed that a subset of YFP-marked cells expressed neither HNF4α nor CK19 ( Figure 1C, white arrow). Therefore, we performed immunostaining of P10 livers for additional markers (Figure 4). Livers were stained for the fibroblast marker alpha smooth muscle actin (αSMA) and CD34 that is expressed by the endothelium of central and portal veins as well as portal fibroblasts [16,29]. Although rare, we confirmed the presence of CD34+αSMA-cells labeled with YFP. On the other hand, none of the CD34-αSMA+ cells were marked, suggesting that AAV8 transduces distinct subsets of cells. This result indicates that AAV8-CMV-Cre labels CD34-expressing cells in addition to cholangiocytes.
We observed that a subset of YFP-marked cells expressed neither HNF4α nor CK19 ( Figure 1C, white arrow). Therefore, we performed immunostaining of P10 livers for additional markers (Figure 4). Livers were stained for the fibroblast marker alpha smooth muscle actin (αSMA) and CD34 that is expressed by the endothelium of central and portal veins as well as portal fibroblasts [16,29]. Although rare, we confirmed the presence of CD34+αSMA-cells labeled with YFP. On the other hand, none of the CD34-αSMA+ cells were marked, suggesting that AAV8 transduces distinct subsets of cells. This result indicates that AAV8-CMV-Cre labels CD34-expressing cells in addition to cholangiocytes.

Neonatal Injection of AAV8-TBG-Cre Labels Cholangiocytes
We then hypothesized that a hepatocyte-specific promoter would suppress the labeling of cholangiocytes by AAV8. To test this hypothesis, we used the TBG promoter to drive the expression of Cre ( Figure 5A). When Rosa YFP/+ mice were treated with AAV8-TBG-Cre at P2, the vast majority of hepatocytes were labeled, while the control vector AAV8-TBG-LacZ did not lead to the expression of YFP

Neonatal Injection of AAV8-TBG-Cre Labels Cholangiocytes
We then hypothesized that a hepatocyte-specific promoter would suppress the labeling of cholangiocytes by AAV8. To test this hypothesis, we used the TBG promoter to drive the expression of Cre ( Figure 5A). When Rosa YFP/+ mice were treated with AAV8-TBG-Cre at P2, the vast majority of hepatocytes were labeled, while the control vector AAV8-TBG-LacZ did not lead to the expression of YFP ( Figure 5B-D). Surprisingly, 6.8% ± 4.2% and 30.9% ± 5.2% of CK19+ cholangiocytes were labeled at P10 and P56, respectively ( Figure 5C,D and Figure 6). The percentage of labeled cholangiocytes increased from P10 to P56 in line with the previous pattern observed with the CMV promoter ( Figure 2). The percentage of AAV8-TBG-Cre-marked CK19+ cells at P56 was slightly higher than the percentage of CK19+ cells marked by AAV8-CMV-Cre although the difference was not statistically significant (TBG: 30.9% ± 5.2% and CMV: 24.4% ± 7.5%, p = 0.08). This phenomenon might be attributable to the silencing of the CMV promoter [9].
In conclusion, our data indicate AAV8-CMV-Cre and AAV8-TBG-Cre can be used for the genetic labeling of cholangiocytes in neonatal murine livers.
In conclusion, our data indicate AAV8-CMV-Cre and AAV8-TBG-Cre can be used for the genetic labeling of cholangiocytes in neonatal murine livers.

Discussion
This study investigated the cellular tropism of the AAV8 vector in the neonatal murine liver and revealed that (a) non-hepatocytes such as CK19+ cells and CD34+ cells can be transduced by AAV8 and (b) the TBG promoter can be used to label cholangiocytes. Both AAV8-CMV-Cre and AAV8-TBG-Cre marked CK19+ cholangiocytes with YFP when injected into Rosa YFP/+ reporter mice at P2, with elevated labeling efficiency at P56 as compared to P10. One possible explanation of this increased percentage of labeled cholangiocytes is that AAV8-expressed Cre continues to mark CK19+ cholangiocytes after P10. Indeed, studies suggest that although AAV is primarily episomal, the expression of the transgene can last up to several months [15,30]. The association between the stability of transgenes and transduction efficiency requires further investigation.
To interpret outcomes of gene therapy and avoid adverse effects, it is critical to understand the cellular tropism of vectors in the neonatal liver. Our data not only provide insight into AAV-based therapy but also can be applied for fast genetic modulation of non-hepatocytes since the generation of genetically engineered mouse models can be time-consuming and expensive. However, our injection

Discussion
This study investigated the cellular tropism of the AAV8 vector in the neonatal murine liver and revealed that (a) non-hepatocytes such as CK19+ cells and CD34+ cells can be transduced by AAV8 and (b) the TBG promoter can be used to label cholangiocytes. Both AAV8-CMV-Cre and AAV8-TBG-Cre marked CK19+ cholangiocytes with YFP when injected into Rosa YFP/+ reporter mice at P2, with elevated labeling efficiency at P56 as compared to P10. One possible explanation of this increased percentage of labeled cholangiocytes is that AAV8-expressed Cre continues to mark CK19+ cholangiocytes after P10. Indeed, studies suggest that although AAV is primarily episomal, the expression of the transgene can last up to several months [15,30]. The association between the stability of transgenes and transduction efficiency requires further investigation.
To interpret outcomes of gene therapy and avoid adverse effects, it is critical to understand the cellular tropism of vectors in the neonatal liver. Our data not only provide insight into AAV-based therapy but also can be applied for fast genetic modulation of non-hepatocytes since the generation of genetically engineered mouse models can be time-consuming and expensive. However, our injection protocol has some limitations. First, neonatal injection of AAV is a suitable strategy for Cre-mediated permanent modification of the genome but not ideal for transient overexpression since the viral genome will be diluted as the liver expands [11]. Also, YFP-marked CD34+ cells were rare with a labeling efficiency of less than 1% (data not shown). Our protocol is not suitable for the modulation of the entire CD34+ population but may be useful for tracing the fate of individual CD34+ cells.
Nevertheless, this study has established a novel strategy for labeling non-hepatocytes in neonatal livers. Our data also revealed a need for reevaluating the specificity of hepatocyte promoters, especially when the vector is administered at young ages. For therapeutic application of our transduction strategy, future research should investigate the use of cholangiocyte-specific promoters and species dependence of cellular tropism.

Conclusions
This study demonstrates that the AAV8 vector is a valuable tool for transducing cholangiocytes in vivo, and that the CMV and TBG promoters can be used to drive Cre expression in cholangiocytes. Our findings potentially provide a basis for fast and simple genetic engineering of cholangiocytes.