Comparison of the Opn-CreER and Ck19-CreER Drivers in Bile Ducts of Normal and Injured Mouse Livers

Inducible cyclization recombinase (Cre) transgenic mouse strains are powerful tools for cell lineage tracing and tissue-specific knockout experiments. However, low efficiency or leaky expression can be important pitfalls. Here, we compared the efficiency and specificity of two commonly used cholangiocyte-specific Cre drivers, the Opn-iCreERT2 and Ck19-CreERT drivers, using a tdTomato reporter strain. We found that Opn-iCreERT2 triggered recombination of the tdTomato reporter in 99.9% of all cholangiocytes while Ck19-CreERT only had 32% recombination efficiency after tamoxifen injection. In the absence of tamoxifen, recombination was also induced in 2% of cholangiocytes for the Opn-iCreERT2 driver and in 13% for the Ck19-CreERT driver. For both drivers, Cre recombination was highly specific for cholangiocytes since recombination was rare in other liver cell types. Toxic liver injury ectopically activated Opn-iCreERT2 but not Ck19-CreERT expression in hepatocytes. However, ectopic recombination in hepatocytes could be avoided by applying a three-day long wash-out period between tamoxifen treatment and toxin injection. Therefore, the Opn-iCreERT2 driver is best suited for the generation of mutant bile ducts, while the Ck19-CreERT driver has near absolute specificity for bile duct cells and is therefore favorable for lineage tracing experiments.


Introduction
The development of inducible Cre transgenic mouse strains has made it possible to generate cell type-specific knockout and lineage tracing experiments. Such experiments generally require two genetically modified loci. The first locus contains a gene or an artificial transgene that is flanked by two loxP recombination sites, a so-called "floxed" allele. The second locus is a transgene or a knock-in where a cell type-specific promoter drives the expression of the Cre recombinase. Often this transgene encodes a tamoxifen-inducible Cre where Cre is fused to the ligand binding domain of the Estrogen

Tamoxifen and CCl 4 Injections
Tamoxifen (T5648; Sigma, St. Louis, MI, USA) was dissolved in corn oil at a concentration of 10 mg/mL and injected intraperitoneally for five consecutive days at a dose of 80 mg/kg bodyweight into mice aged 6 to 8 weeks. Mice from uninjured groups were sacrificed after either a 3-day or a 3-week washout period following the last tamoxifen injection. Injured groups received an intraperitoneal injection of CCl 4 (1 mL/kg bodyweight, in corn oil) after a 3-day or a 3-week washout period. Mice were sacrificed 3 days after CCl 4 injection. Mice with liver injury but without tamoxifen washout received CCl 4 on day 2 of the 5 days of tamoxifen injection. These mice were sacrificed on day 5, the last day of tamoxifen injection. Vehicle control mice were injected with corn oil on five consecutive days, mice were sacrificed 3 days after corn oil injection.
Hnf-4α + hepatocytes were quantified by counting 18 pictures from two liver sections per mouse. This amounted to about 5000 Hnf-4α + cells per section (4697 ± 704). The number of Hnf-4α + ,tdTomato + double positive cells was divided by the total number of Hnf-4α + cells for each mouse. The resulting fraction was expressed as percentage and used as a measure for R26-tdTomato recombination.
To determine the efficiency of R26-tdTomato recombination in cholangiocytes, per mouse 20 bile duct regions of all sizes were visualized in one liver section and Ck19 + cells were counted. On average 400 (397 ± 69) Ck19 + cells were counted per mouse. Every Ck19 + cell was checked for tdTomato expression. The number of Ck19 + ,tdTomato + double positive cells was then divided by the total number of Ck19 + cells, expressed as a percentage, and used as a measure for recombination efficiency.

Statistical Analysis
Two-way ANOVA with interaction was performed on percentages of Hnf-4α + cells that were also tdTomato + to test for differences in specificity between genotypes and between washout types. In the two-way ANOVA for uninjured mice genotypes included Opn-CreER-tdTomato and Ck19-CreER-tdTomato and washout groups included 3-day washout or 3-week washout. In the two-way ANOVA for CCl 4 -injured mice genotypes included Opn-CreER-tdTomato and Ck19-CreER-tdTomato and washout types included 3-day washout, 3-week washout, or no washout. Tukey's range test was performed as a post-hoc test to compare all possible pairs of means in both ANOVAs. Two-way ANOVA was performed on percentages of Ck19 + cells that were also tdTomato + to test for differences in efficiency between genotypes and treatment groups. Genotypes included Opn-CreER-tdTomato and Ck19-CreER-tdTomato, and treatment groups included uninjured or CCl 4 -treated after either a 3-day or 3-week washout following tamoxifen injections, resulting in a total of 8 groups. Tukey's range test was performed as a post-hoc test to compare all possible pairs of means. A p-value of 0.05 and lower was considered to be statistically significant for all tests. All statistical tests were performed using R-Studio (Version 1.1.463; R studio Inc., Boston, MA, USA).
First, we evaluated whether the two Cre drivers induced recombination in the absence of tamoxifen. Liver sections were immunostained for Ck19 to mark all cholangiocytes, and were imaged for Ck19 and tdTomato expression. After corn oil injection (vehicle), Opn-CreER-tdTomato mice showed tdTomato expression in 2% of cholangiocytes (9/388). The 9 tdTomato + cholangiocytes were found in 7 different bile ducts and mostly appeared as single cells. In Ck19-CreER-tdTomato corn oil injected mice, 13% of cholangiocytes were tdTomato + (54/418) (Figure 1b). The 54 tdTomato + cells were found in 8 different bile ducts and appeared as large patches of cells, presumably clones, in large bile ducts. This amounted to 1.80 independent recombination events (clones) per 100 cholangiocytes for Opn-CreER T -tdTomato mice and 1.91 for Ck19-CreER T -tdTomato mice observed at 8 weeks of age. We did not observe tdTomato expression in any other liver cell type in either strain. Thus, while both strains showed some degree of leakiness, Ck19-CreER T had high numbers of recombined cells in the uninduced condition. First, we evaluated whether the two Cre drivers induced recombination in the absence of tamoxifen. Liver sections were immunostained for Ck19 to mark all cholangiocytes, and were imaged Mice were injected with tamoxifen on five consecutive days and sacrificed three days or three weeks later. (e) Percentage of tdTomato + Ck19 + cells quantified in liver sections at different timepoints after tamoxifen injection (n = 7-8 mice per experiment; 20 bile duct regions per mouse). The difference between the two genotypes was significant 3 days (p = 0.001) and 3 weeks (p = 0.001) after tamoxifen injection. Scale bars: 20 µm.
Next, we evaluated the recombination efficiency induced by tamoxifen administration. We injected five doses of tamoxifen on consecutive days and analyzed tdTomato expression three days after the last injection. In Ck19-CreER-tdTomato mice, tdTomato expression was highly mosaic and we did not detect any duct that was entirely composed of tdTomato + cells (Figure 1b, Supplemental Figure S1). Rather, the percentage of tdTomato + cholangiocytes per bile duct was variable, and in most ducts less than half of the Ck19 + cells expressed tdTomato (Figure 1b,c). Quantification of a large number of bile ducts showed that only 35% of the Ck19 + cholangiocytes (932/2676) expressed tdTomato (Figure 1d,e; Table 1). This number did not increase over time and three weeks after tamoxifen injection still only about 28% of the Ck19 + cells expressed tdTomato (862/3124). In contrast, virtually all bile duct cells expressed tdTomato in Opn-CreER-tdTomato mice (Figure 1b). Quantification showed that tdTomato was detected in 99.96% and 99.94% of the cholangiocytes (2722/2723; 3298/3300) three days and three weeks after tamoxifen injection, respectively (Figure 1d,e; Table 1). These data show that the Opn-iCreER T2 driver efficiently recombines the floxed STOP cassette whereas the Ck19-CreER T driver has limited recombination efficiency when using R26-tdTomato as a reporter. During this analysis, we also observed that heterozygous Ck19-CreER-tdTomato mice had reduced Ck19 expression. This is probably due to the fact that the Ck19-CreER T construct is a knock-in into the Ck19 locus causing loss of function ( Figure S2) [2]. Homozygous Ck19-CreER-tdTomato mice lacked Ck19 expression altogether ( Figure S2). On the other hand, neither Ck19 nor Opn expression was affected in heterozygous or homozygous Opn-CreER-tdTomato mice consistent with the transgenic nature of the Opn-iCreER T2 construct ( Figure S2) [5].

The Ck19-CreERT and Opn-iCreERT2 Drivers are Highly Specific towards Cholangiocytes
To determine the frequency of recombination in hepatocytes, we stained the Ck19-CreER-tdTomato and Opn-CreER-tdTomato mice for the hepatocyte-specific marker Hnf-4α and determined overlap with tdTomato expression. No tdTomato + hepatocytes were detected in Ck19-CreER-tdTomato mice three days (0/61190) or three weeks (0/76513) after the last tamoxifen injection ( Figure 2; Table 2). Opn-CreER-tdTomato mice had very few tdTomato + hepatocytes, namely 5 hepatocytes out of 67462 (0.0074%) three days after tamoxifen injection and 7 hepatocytes out of 77086 (0.0091%) three weeks after tamoxifen treatment ( Figure 2; Table 2). These tdTomato + hepatocytes usually appeared as single cells and their location was not restricted to a specific zone. All tdTomato + Ck19cells were positive for Hnf-4α, and none of those cells co-stained for the hepatic stellate cell marker Desmin or the immune cell marker Cd45 (Figure S3), confirming earlier reports [17]. We conclude that the Ck19-CreER T driver had absolute specificity in these experiments while the Opn-iCreER T2 driver had extremely low frequency of recombination in cells other than cholangiocytes.

Ectopic Expression of the Cre Drivers after Liver Injury
Liver injury can affect gene expression in diverse liver cells. As such, ectopic Opn expression was observed in hepatocytes, HSCs, and immune cells after liver injury [13,15,18,19]. We therefore assessed the specificity of the Ck19-CreER T and the Opn-iCreER T2 drivers after liver injury caused by carbon tetrachloride (CCl 4 ) toxicity. To do this, we injected a single dose of CCl 4 into mice either during the tamoxifen treatment, or three days or three weeks after the end of the tamoxifen treatment. We then quantified the number of Hnf-4α + cells that expressed tdTomato three days after the CCl 4 injection (Figure 3a-c). As shown by H&E staining, CCl 4 induced necrosis in pericentral hepatocytes ( Figure  S4). In the Ck19-CreER-tdTomato mice, we found only one of over 100,000 hepatocytes that expressed tdTomato (1/21016; 0/44034; 0/40768 tdTomato + hepatocytes for the three conditions). This result indicates that CCl 4 does not induce ectopic expression of the Ck19-CreER T transgene (Figure 3a-c, Table 2). Indeed, Ck19 staining of CCl 4 -injected mice showed that Ck19 expression remained restricted to cholangiocytes after liver injury (Figure 3d). Similarly, extremely few tdTomato + hepatocytes (0/35801; 7/30984) were found in Opn-CreER-tdTomato mice when CCl 4 was injected three days or three weeks after tamoxifen administration (Figure 3a,b). However, when CCl 4 was injected during tamoxifen treatment, 0.33% of hepatocytes (81/24192) in the Opn-CreER-tdTomato mice expressed tdTomato (Figure 3c-d; Table 2). Thus, elevated ectopic recombination in hepatocytes was observed when the injury was concomitant with the tamoxifen administration. This is consistent with the ectopic expression of the endogenous Opn gene after different types of liver injury [13,19]. No tdTomato expressing HSCs or immune cells were detected in either mouse model, as shown by immunostaining for Desmin and Cd45 respectively ( Figure S5). Recombination efficiencies in cholangiocytes were comparable between injured and non-injured conditions (Table 1). Altogether, these data show that Ck19-CreER T and Opn-iCreER T2 retain their high specificity towards cholangiocytes after liver injury. However, CCl 4 induces ectopic Opn expression in hepatocytes and therefore a 3-day washout period between tamoxifen and toxin treatment is required to maintain this high level of specificity of the Opn-iCreER T2 driver when using R26-tdTomato as a reporter.

Discussion
In this study we performed a thorough analysis of the efficiency and specificity by which the Ck19-CreER T and Opn-iCreER T2 drivers trigger recombination of the R26-tdTomato reporter in the liver.
We followed a standardized tamoxifen regimen and compared tdTomato expression in homeostatic livers and livers with acute toxic injury. Our results show that the Ck19-CreER T driver has near absolute specificity for cholangiocytes under normal and injury conditions. However, Ck19-CreER T has relatively low efficiency such that only about 32% of cholangiocytes underwent recombination of the R26-tdTomato reporter. On the other hand, the Opn-iCreER T2 driver showed nearly 100% efficiency in recombining the R26-tdTomato reporter in cholangiocytes. Notably, other studies reported lower levels of recombination with the Opn-iCreER T2 driver, probably because lower amounts of tamoxifen were injected and a less sensitive reporter strain was used (Rosa26-loxP-STOP-loxP-YFP) [5,17]. In any case, our findings indicate that Opn-iCreER T2 drives LoxP site recombination more efficiently than the Ck19-CreER T driver and is therefore more suitable to perform knockout studies in bile ducts.
Despite the fact that Opn-iCreER T2 is a stronger recombination driver than Ck19-CreER T , one should keep in mind that efficiency of creating knockout cells may be lower than 100% in knockout studies. Recombination of one allele is sufficient to produce a positive signal from the R26-tdTomato reporter, yet generation of homozygous mutant cells may require recombination of both alleles. In addition, the sensitivity to Cre recombination for a floxed allele may be lower than for the tdTomato reporter line, which is known to be very sensitive [20]. Determination of the recombination efficiency is therefore important to interpret the results of knockout experiments.
Our study showed Cre activity in the absence of tamoxifen. Recombination occurred in 13% of cholangiocytes in the Ck19-CreER T -tdTomato line and in 2% in the Opn-iCre-ER T2 -tdTomato line. Despite its high specificity, the CK19-CreER T might therefore be unsuited for lineage tracing experiments since temporal control over Cre recombination is partially lost. These baseline levels of Cre recombination have to be considered when performing experiments.
Some hepatocytes ectopically activated Opn-iCreER T2 and expressed the tdTomato reporter. This number was minimal, since less than 1 in 10,000 hepatocytes (0.01%) was positive for tdTomato in all conditions. This extremely low level of Cre recombination in hepatocytes may not be sufficient to affect the phenotype of a full knockout in bile ducts. Nevertheless, this low background recombination may make the Opn-iCreER T2 driver unsuited for lineage tracing experiments that relay on absolute specificity for cholangiocytes, yet the driver can be used to generate mutant bile ducts.
The origin of the occasionally observed tdTomato + hepatocytes is unclear. One possibility is that these hepatocytes were derived from cholangiocytes by trans-differentiation. However, we think that this scenario is improbable because we did not detect signs of a transition from cholangiocytes to hepatocytes, like expression of Opn or Ck19 in tdTomato + Hnf-4α + hepatocytes, at any analyzed time point, and because the tdTomato + hepatocytes were not accumulated around bile ducts as would be expected if they were derived from cholangiocytes. Rather, the tdTomato expression in sparse hepatocytes may be caused by ectopic activation of the promoter driving Cre in hepatocytes. Indeed, expression of Opn is ectopically activated in hepatocytes upon liver injury, indicating that also the Opn-iCreER T2 construct may be ectopically activated [13,19]. However, non-specific recombination driven by the ectopic activation of CreER in hepatocytes could be avoided when liver injury was induced three or more days after the end of the tamoxifen administration. This washout period was thus sufficient to lower the levels of tamoxifen below the threshold required to activate the CreER T2 that was ectopically expressed in hepatocytes [21]. The length of such a washout period will however depend on the tamoxifen regimen and on the injury model applied, and should therefore be re-examined when using a different experimental set-up. In addition, high doses of tamoxifen can cause liver injury and induce the expression of cholangiocyte-specific markers such as Sox9 in hepatocytes [22]. Thus, tamoxifen toxicity may contribute to the ectopic recombination of tdTomato in hepatocytes.
Ck19-CreER T was generated by a knock-in of the CreER T ORF at the start codon of the Ck19 gene. This produced a Ck19 loss-of-function mutation and indeed we observed reduced amounts of Ck19 expression in Ck19-CreER T heterozygous mice. Homozygous Ck19-CreER T mice are viable and appear to have normal bile ducts, although without Ck19 expression [23]. The possibility that reduced expression of Ck19 might impact cholangiocyte function or specific post-injury phenotypes needs to be further evaluated. The Opn-iCreER T2 mice were created by random integration of a bacterial artificial chromosome (BAC) containing the Opn gene with iCreER T2 inserted. The insertion of this construct into the mouse genome did not cause detectable deleterious effects, as homozygous Opn-iCreER T2 mice are viable and did not show altered bile duct morphology or cholangiocyte gene expression. Thus, if desired, both drivers can be bred to homozygosity.

Conclusions
Our data indicate that the Opn-iCreER T2 driver is best suited for the generation of mutant bile ducts, while the Ck19-CreER T driver has near absolute specificity for bile duct cells and is therefore favorable for lineage tracing experiments. Importantly however, it has to be considered in this model that Cre recombination is present in the absence of tamoxifen.

Conflicts of Interest:
The authors declare no conflict of interest.