Refined protocols of tamoxifen injection for inducible DNA recombination in mouse astroglia

Inducible DNA recombination of floxed alleles in vivo by liver metabolites of tamoxifen (TAM) is an important tool to study gene functions. Here, we describe protocols for optimal DNA recombination in astrocytes, based on the GLAST-CreERT2/loxP system. In addition, we demonstrate that quantification of genomic recombination allows to determine the proportion of cell types in various brain regions. We analyzed the presence and clearance of TAM and its metabolites (N-desmethyl-tamoxifen, 4-hydroxytamoxifen and endoxifen) in brain and serum of mice by liquid chromatographic-high resolution-tandem mass spectrometry (LC-HR-MS/MS) and assessed optimal injection protocols by quantitative RT-PCR of several floxed target genes (p2ry1, gria1, gabbr1 and Rosa26-tdTomato locus). Maximal recombination could be achieved in cortex and cerebellum by single daily injections for five and three consecutive days, respectively. Furthermore, quantifying the loss of floxed alleles predicted the percentage of GLAST-positive cells (astroglia) per brain region. We found that astrocytes contributed 20 to 30% of the total cell number in cortex, hippocampus, brainstem and optic nerve, while in the cerebellum Bergmann glia, velate astrocytes and white matter astrocytes accounted only for 8% of all cells.

is predominantly metabolized via two pathways. In the first, TAM is converted by CYP3A to N-desmethyltamoxifen (NDM-TAM), one of the major metabolites. This metabolite undergoes multiple oxidation steps including 4-hydroxylation to endoxifen (END) by CYP2D6. 4-Hydroxylation of TAM to 4-hydroxytamoxifen (4-OH-TAM) via multiple CYPs represents the second metabolic route, with significant importance for experimental biology in mice. A small proportion of endoxifen appears to result from CYP3A-catalyzed N-demethylation of 4-OH-TAM 10,11,49,50 . (B) Samples (serum/brain) of treated C57BL/6NRj wild type (wt) mice were collected and were analyzed by LC-HR-MS/MS. (C,D) Mice were injected on post-natal day 28 with TAM (100 mg/kg) once (C) or for three consecutive days (D, red triangles). Brain and blood samples were collected at different time points (indicated with black arrows). After processing samples were analyzed by LC-HR-MS/MS for TAM and its derivatives NDM-TAM, 4-OH-TAM and END.
Tamoxifen-induced gene recombination. Transgenic mice were injected intraperitoneally with TAM in corn oil (100 mg/kg body weight, 10 mg/ml stock solution, Sigma, St. Louis, MO, USA) at the age of 4 weeks once per day for one, two, three or five consecutive days (Fig. 3A). In addition, mice were injected with TAM at different intervals (3× TAM with a pause of one day in between, Suppl. information Fig. 3A). Mice were analyzed 8 hours post injection (hpi), 48 hpi, 21 days post injection (dpi) or 200 dpi (Fig. 3A). For the pharmacokinetic analysis, wt mice received a single dose of TAM once or at three consecutive days (Fig. 1C,D). Samples were collected at 1 hpi, 4 hpi, 8 hpi, 24 hpi, 2 dpi, 5 dpi, 7 dpi, 14 dpi or 21 dpi. After three injections, tissue was collected 24 h after the last injection, corresponding to 3 d after the first injection. Similarly, when tissue samples were prepared at 18 d after the last injection, it corresponds to 21 d after the first injection (Fig. 1C,D). The biological half-lifes of TAM and its metabolites in blood or brain were determined after reaching maximum by non-linear regression (one-phase decay, GraphPad Prism 7, GraphPad Software Inc., La Jolla CA, USA). Over several years, we observed a mortality rate of 3.7% after three injections of TAM (n = 4.080, adult mice).

Quantitative Real Time PCR (qRT-PCR).
Quantitative RT-PCR (qRT-PCR, CFX96 Real-Time PCR Detection System; BioRad, Hercules, CA, USA) was performed to determine (1) the loss of floxed (recombined) alleles (loss) in various brain regions and (2) relative increase of recombined alleles (gain) (ctx/cb) using EvaGreen for quantification (Axon, Kaiserslautern, Germany). Loss primers were located up-and downstream of the 5′ loxP site (p2ry1, gabbr1, stop tdTomato) to amplify non-recombined alleles. Values (ΔCT) of floxed recombined samples were normalized to those obtained from samples of control floxed non-recombined animals. Gain primers were designed to flank 3′ and 5′ loxP sites leading to PCR amplification after successful recombination. ΔCT values of floxed recombined mice were normalized to mean ΔCT values after five consecutive TAM injections. Reactions were carried out in triplicates with neuregulin 1 type III (NrgIII) and β-actin (bact) as endogenous gene controls.  regarded them as also being normally distributed and applied the Grubb's test to identify outliers. These outliers were not included in the statistical calculations, but we depicted them in light red color to show the complete datasets. All data of Fig. 4 and Suppl. Fig. 7 were normally distributed.
Inter-group comparisons were performed by two-tailed Student t test using the GraphPad Prism 7 software (GraphPad Software Inc., La Jolla CA, USA). Data are represented as means ± SEM of natural replicates (mice) with *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001. Significant outliers were not included into calculations. But, to demonstrate the full data-set, they were indicated by light red color in Fig. 3 and Suppl. information Fig. 3. tdTomato+ and TO-PRO-3+ cells were determined in three female and three male mice. A total of nine detailed images (from two parasagittal slices per animal; 40× objective, in posterior, mid and anterior regions of the cortex) were taken. No gender differences were observed.
Data availability statement. The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.

Results
Fast uptake and clearance of tamoxifen and its bioactive metabolite 4-OH-TAM into the brain. To analyze the contribution and time course of tamoxifen (TAM) and its metabolites N-desmethyltamoxifen (NDM-TAM), 4-hydroxytamoxifen (4-OH-TAM) and endoxifen (END) for cell type-specific . Primers (for p2ry1 and gria1 gene) span both loxP sites with amplification of the flanked sequence after recombination. (D,E) Differences in recombination efficiency between both analyzed alleles, but also both brain regions were detected with the lowest cortical and cerebellar recombination at 8 hpi. Per injection protocol three to four mice (exception 200 d, n = 2) were analyzed (colored points) and ΔCT-values were normalized to the mean value of animals which received 5× TAM. The light red dots indicate significant outliers that were not considered for calculations. The error bars correlate to the SEM of the biological replicates (n = 2-4, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001). recombination, we determined their pharmacokinetic profile by LC-HR-MS/MS. After a single intraperitoneal injection of TAM, metabolite concentrations were determined in serum and brain at 1, 4, 8, 24 h and 2, 5, 7, 14 and 21 d post injection (hpi or dpi) (Fig. 1B,C). In addition, the concentrations were determined after injection of TAM for three consecutive days (Fig. 1B,D).
TAM and 4-OH-TAM showed fast partitioning in the brain reaching maximal levels as early as 8 hpi. 4-OH-TAM entered the brain equally fast, although it required prior enzymatic hydroxylation of TAM in the liver by cytochrome P450 cyclooxygenases (Fig. 2). Corresponding TAM samples from the serum displayed profound variability, which was less pronounced for 4-OH-TAM (Suppl. information Fig. 2). We attributed this to variations of tissue and blood absorption after individual intraperitoneal injections. Both, TAM and 4-OH-TAM were efficiently cleared from brain and blood within 7 dpi (Suppl. information Table 1, Fig. 2A,B, Suppl. information Fig. 2A,B). In the brain, TAM was approximately six times enriched in comparison to 4-OH-TAM at 8 hpi as well as at 48 hpi ( Fig. 2A,B). The other two metabolites, NDM-TAM and END, required three-to five-fold more time to reach their concentration peaks (at 24 hpi, Suppl. information Table 1, Fig. 2C,D, Suppl. information Fig. 2C,D). Nevertheless, both were cleared completely at 7 dpi (Fig. 2C,D, Suppl. information Fig. 2C,D). As expected from its lipophilic properties, TAM displayed the highest concentration in the brain followed by NDM-TAM, 4-OH-TAM and, at lowest, END (Fig. 2, Suppl. information Table 1).
Consecutive injections of TAM are used frequently to achieve higher recombination efficiencies. Therefore, we injected TAM for three consecutive days and determined the levels of TAM and its metabolites focusing at the time required for clearance ( Fig. 2E-H). All four compounds showed the highest concentration 24 h after the last injection (3 dpi). While TAM and 4-OH-TAM were only slightly enriched, NDM-TAM and END were about 5-fold enriched in comparison to single injections (Fig. 2E-H, Suppl. information Table 1). Obviously, consecutive injections increase the levels of TAM metabolites in the brain.
For the brain we determined clearance rates (t 1/2 ) of 9.9 h,15.8 h, 19.9 h and 33.2 h for TAM, 4-OH-TAM, NMD-TAM and END after a single injection, respectively. These rates increased after three injections to 26.1 h, 20.7 h, 30.5 h and 41.2 h. Although, concentration levels were highly variable in the serum, all four compounds were cleared with similar rate constants as in the brain (Suppl. information Fig. 2 and Table 1).
In summary, the time windows of optimal bioactivity of TAM and its metabolites are between 4 and 24 h post injection for single injections, and 4 h to 5 d for repeated injections over three consecutive days. All metabolites were effectively cleared to negligible levels at 7 dpi.

Maximal recombination levels are reached by 3 or 5 days of TAM injections in cerebellum or cortex, respectively.
To determine the in vivo efficacy of TAM injections to activate the Cre DNA recombinase, we studied the recombination of two floxed gene loci, p2ry1 (P2Y1 receptor) and gria1 (AMPA receptor subunit GluA1) in two different areas of the brain, cerebellum (cb) and cortex (ctx), using GLAST CreERT2 mice (high-affinity glutamate/aspartate transporter, GLAST, slc1a3; GLAST CreERT2/+ x gria1 fl/fl x p2ry1 fl/fl ; Fig. 3C). Mice received single injections of TAM (100 mg/kg body weight) on 1, 2, 3 and 5 consecutive days. Samples were analyzed at 8 hpi, 21 dpi and 200 dpi by qRT-PCR (gain), where successful recombination is indicated by an amplified PCR product (Fig. 3A,B). We also tested a two-day interval of three consecutive TAM injections (inspired by the rapid drop of 4-OH-TAM at 48 hpi) ( Fig. 2B and Suppl. information Fig. 3).
After three TAM injections, both alleles (p2ry1 and gria1) were already recombined at a percentage of 80 and more (n = 4, Fig. 3D,E, Suppl. information Table 2), in ctx as well as in cb. Almost 100% recombination was achieved after five consecutive injections. As expected from LC-HR-MS/MS data, increasing the time period between induction and analysis of recombination from 21 to 200 d did not change the recombination rate for both gene loci (n = 2, Suppl. information Table 2). Analyzing successful recombination already at 8 hpi after a single TAM injection revealed large differences in recombination within different brain regions, but also between different gene loci. In the ctx only a small percentage of floxed alleles was recombined (n = 3), while the recombination rate in the cb (n = 3) was higher. As expected from the longer bioavailability of its metabolites, higher degrees of recombination were observed at 21 dpi in comparison to 8 hpi (n = 3) after a single TAM injection (Fig. 3D,E, Suppl. information Table 2). Similarly, three TAM injections were more efficient than only two (n = 3) at 21 dpi (Fig. 3D,E, Suppl. information Table 2).
Since we observed fast clearance within less than 48 h, an interval protocol with one-day pause between injections (i.e. injections at every 2 nd day) was expected to be less efficient than a daily-injection protocol (Fig. 2, Suppl. information Fig. 3). Indeed, the interval protocol revealed less recombination in the ctx, (reduction of 32% for p2ry1 and 61% for gria1) in comparison to the five-consecutive-day injection protocol (n = 4, Suppl. information Fig. 3B). In cb, the interval protocol proved to be less efficient than the 5-consecutive-day protocol (by 20%). But, in general, cerebellar recombination was higher than cortical recombination (Suppl. information Fig. 3C).
At early time points after TAM injections, (8 hpi) large differences in recombination efficiencies between different alleles and brain regions were detected. In the cortex, recombination of the p2ry1 alleles was more than four times larger than gria1 alleles. In contrast, in the cerebellum no difference between the two alleles was detected (Fig. 3D,E, Suppl. information Table 2). Simultaneously, the recombination rate of the cerebellum was much higher than in the cortex. At 21 dpi, a single TAM injection induced more than twice as many recombined p2ry1 than gria1 alleles in the ctx. This difference disappeared, when three consecutive injections were applied (Fig. 3D,E). In general, both target genes were recombined faster in the cb than in the ctx after one or two injections, while comparable recombination was observed after three and five injections. Maximal levels of recombination require single TAM injections for 3 to 5 d depending on brain region and floxed allele.
To functionally evaluate recombination efficiencies, we induced expression of the fluorescent reporter protein tdTomato in GLAST-Cre ERT2 x R26-tdTomato (stop fl/fl tdTomato) mice (Suppl. information Figs 4, 5 and 6) and co-stained with astrocytic markers (GS, S100B, GFAP). After three consecutive injections, increasing numbers of recombined astrocytes could be identified throughout the brain (Suppl. information Figs 5 and 6). At 3.5 and SCIentIfIC RepoRts | (2018) 8:5913 | DOI:10.1038/s41598-018-24085-9 5 months post injection, a stabilized percentage of recombined cells was observed. The recombination specificity was the same in older mice (Supp. information Fig. 6). These data confirmed that three to five injections are sufficient to recombine the maximum number of cells in GLAST-Cre ERT2 mice of any age (Suppl. information Fig. 4).

GLAST-Cre ERT2 mediated DNA recombination can be used to quantify astroglial cell numbers. If
Cre ERT2 is expressed in a cell type-specific manner, the proportion of this cell type in a given region can be determined by qRT-PCR across a single loxP site of the non-recombined locus, since reduction of amplimer levels indicates successful gene deletion (Fig. 4, loss). Using the GLAST-Cre ERT2 mouse line, we compared recombination of three different floxed alleles (stop fl/fl tdTomato; p2ry1 fl/fl and gabbr1 fl/fl ) in different brain regions (bs, cb, ctx, hc, opt).
To confirm the qRT-PCR analysis, we also counted the number of recombined tdTomato-positive cells in the cortex (Fig. 4G,H). We found 21 ± 5% of all cells (TO-PRO labelled nuclei) were tdTomato-positive (Fig. 4H), a value well in-line with the observed cortical recombination (20 ± 2%). In addition, no significant differences of tdTomato-positive cells could be detected in the cortices of males and females (21 ± 6%, 20 ± 4%, respectively, Suppl. information Fig. 7). Our results show that qRT-PCR data of recombined loxP sites obtained from tissue homogenates correlates to counting of individual recombined cells. Therefore, the recombination rate reflects well the percentage of GLAST-positive cells (=astrocytes) in the respective brain regions.

Discussion
The inducible Cre ERT2 /loxP system is a highly effective genetic tool. By quantifying the bioavailability of tamoxifen (TAM) and its metabolites and by assessing genomic recombination using the GLAST-Cre ERT2 knockin mouse line, we provide a rationale for optimal protocols of recombination (Fig. 5). In addition, we show that qRT-PCR of the Cre ERT2 / loxP system can also be used to determine the proportion of defined cell types in various tissue regions (Fig. 5).
Our main goal was to establish a protocol with the best recombination efficiency using the lowest TAM concentration to minimize side effects. For that purpose, we analyzed the concentration of TAM and its metabolites in the brain tissue and serum of wt mice. TAM and 4-OH-TAM showed highest concentrations already very early at 8 hpi, while NMD-TAM and END peaked at 24 hpi. We found that at 8 hpi,a single injection of TAM generated high levels of the effective 4-OH-TAM (4.4 µg/g brain weight), while others 24 observed only 2.1 µg/g at 6 hpi and required an additional injection. Consecutive injections of low doses of TAM let all four substances accumulate, but they were efficiently excreted within 7 dpi, similar to a recent study, where different doses of TAM (up to 400 mg/kg body weight within 24 h) were compared 24 . Within the first 40 h, TAM and its metabolites were reduced by half after a single dose of TAM with TAM (9.9 h), 4-OH-TAM (15.8 h) and NMD-TAM (19.9 h) being cleared faster than END (33.2 h). Although TAM itself can be toxic, we observed a rather low mortality rate of 3.7% with a TAM concentration of 100 mg/kg, tested in many adolescent and adult mice (older than 21 days, n = 4,080). All results were obtained after intraperitoneal TAM injection which we prefer over oral gavage, due to the faster application and more precise dosage. The handling itself is also less stressful to the mice [24][25][26] . Others have also demonstrated a more efficient Cre activity in GLAST-CreERT2 knockin mice after i.p. injection of tamoxifen when compared to oral gavage 18 .
The minimal effective concentration of 4-OH-TAM has been determined at 38 ± 22 ng/g 24 . This value is slightly lower than the concentration we detected in the brain still at seven days when TAM was injected for three consecutive days (58.4 ng/g, 7 dpi). Long presence of 4-OH-TAM is important for genetic experiments in which extended gene excision is required. For fate mapping experiments, often short time windows for reporter gene activation are preferred. Our analysis demonstrates that such activation windows are shorter than five days after a single injection. At 5 dpi, only ineffective concentrations of 4-OH-TAM remain in the brain (1.3 ng/g).
The metabolites END and 4-OH-TAM have a 100 times stronger affinity to estrogen receptors in comparison to TAM 6,27,28 ; and both are almost comparable in their induction capability of CreERT2-mediated gene recombination 29,30 . Since we found higher levels of END (up to two-fold) in comparison to 4-OH-TAM after 5 to 7 dpi, at this time some recombination due to slow accumulation of END cannot be excluded.
Three to five days of consecutive injections of TAM induced maximal recombination, while fewer injections were less effective. Similarly ineffective was a protocol with a pause between injections (three injections with 48 h in between) since TAM and 4-OH-TAM were cleared quickly, as seen by LC-HR-MS/MS. Differences of gene recombination at different floxed alleles or brain regions disappear after 21 days, when at least three injections are given. Because of its solubility properties, TAM has to be dissolved in oil. Due to variable partitioning, we observe strong fluctuations of TAM in serum of different animals. However, in the brain, the levels of TAM and its metabolites vary significantly less, probably due to the regulatory function of the blood-brain barrier.
Our data show that the three-to-five-injection standard protocol gives maximal recombination levels with a low mortality rate of less than 4%.
Astrocytes are often considered the most abundant cell type in the brain, but precise numbers are missing. Approaches used in recent work cannot distinguish different neuronal or glial cell types. Only ratios of neuronal (NeuN-positive) and non-neuronal (NeuN-negative = glial) cells were estimated in different brain regions of various species [31][32][33][34][35] with about 50% of cells being glia in the mouse cerebral cortex. Inspired by our DNA recombination study, we provide an alternative approach that allows quantifying the proportion of individual glial cell populations. By continuing to use the GLAST-CreERT2 mice, we estimated the proportion of astrocytes in different regions of the murine brain. By setting the percentage of GLAST-CreERT2 mediated loss of floxed alleles as a measurement for the proportion of astrocytes (GLAST + cells) in a given brain tissue, we found 20% (brainstem), 8% (cerebellum), 22% (cortex), 30% (hippocampus) and 31% (optic nerve) of all cells account for astrocytes. In support of these results, direct cell counting of tdTomato + cells in the cortex revealed that 21% of all cells are astrocytes. In our experiments, we confirmed the high cell specificity of the GLAST-CreERT2 mouse line as extensively described 18 . Therefore, we interpreted GLAST + cells as astrocytes and neglected the different patterns of GLAST expression in astrocyte subpopulations or expression in neurogenic radial glia 36 . Regional differences of GLAST expression, for example in radial glial cells in the hippocampus (leading to recombined newly born granule cells), are not taken into account in our data sets 37,38 . Similar to our data, counting of S100B + cells in rat cortex revealed 18.5% astrocytes 39 . Our method relies on the fact that each cell contains the same genomic content including loxP sites. This approach can similarly be used for other glial cell types such as oligodendrocytes or microglia taking advantage of respective Cre DNA recombinase driver lines, e.g. by employing PLP-CreERT2 or CX3CR1-CreERT2 40,41 mouse lines (see also http://www.networkglia.eu/en/animal_models). We tested our approach at three different floxed alleles: p2ry1, gabbr1, tdTomato. Although, largely, the three loci displayed very similar recombination efficiencies, we observed differences particularly at the onset of genomic recombination (p2ry1, gria1). Such differences disappeared when plateau phases of maximal recombination were reached.
One caveat is that the various gene loci may have different recombination efficiencies. Recombination kinetics for the AMPA receptor locus gria1 was determined for the cerebellum on a precise time scale 42 . After 11 h about 50% of the recombination events occurred, i.e. three hours after the determined peak of TAM and 4-OH-TAM. In line with these data, we found that 33% recombined alleles for gria1 and 36% for the p2ry1 locus at 8 hpi. In contrast to the cerebellum, however, recombination in the cortex was lower with only 3% and 14% for gria1 and p2ry1, respectively. Since effective recombination does not only depend on the level of Cre ERT2 expression, but also on chromatin structure, we suggest that the differences in recombination kinetics between (1) gria1 and p2ry1 and (2) the different brain regions (ctx vs. cb) are caused by different chromatin structures, with the gria1 locus being less accessible than the p2ry1 locus. Local chromatin structure and consequently, the potential for gene expression, are regulated by a number of post-translational, covalent modifications of histone-amino terminals, like methylation or acetylation [43][44][45] . Conceptually, within euchromatin structures the degree of condensation or DNA accessibility varies depending on gene activity 46 . Different DNA modifications between p2ry1 and gria1, but also between ctx and cb could lead to different recombination efficiencies, as it has been shown for the Huntington's disease homolog allele (hdh), which recombined only in the brain and in testis; both tissues with highest levels of hdh mRNA in mice 47 . In contrast, usage of the same Cre-expressing line caused recombination in every tissue when other floxed alleles were used 47 . Hence, for each new allele-of-interest a thorough DNA recombination analysis should be performed to rensure maximal recombination. Concentrations of TAM and its derivatives after single TAM injections revealed that about 28% of the originally injected TAM could be detected in the brain while only 4% 4-OH-TAM, 6% NDM-TAM and negligible amount of END could be found at 8 hpi. While TAM (9%) and 4-OH-TAM (3%) decreased 24 hpi, NDM-TAM (9%) and END (3 × higher) increased their concentrations in the brain. (B) The optimized injection protocol for GLAST-Cre ERT2 mice for cortical and cerebellar recombination requires TAM injections for three to five consecutive days depending on brain region and floxed allele. (C) In a brain of a young adult mouse, astrocytes account for 20% (±2, n = 14) in the bs, 8% (±2, n = 13) in the cb, 22% (±2, n = 19) in the ctx, 30% (±3, n = 16) in the hc and 31% (±2, n = 17) in the optic nerve (setting GLAST-positive cells as astrocytes).
SCIentIfIC RepoRts | (2018) 8:5913 | DOI:10.1038/s41598-018-24085-9 In this study, we focused on young adult mice. It is quite conceivable to assume that recombination varies with aging. And indeed, it has been proposed that tamoxifen induces recombination more efficiently in younger mice 48 . However, when we studied reporter expression in adult Rosa26-reporter mice using different Cre ERT2 expressing mouse lines, we could not detect changes in age-dependent recombination efficiency or specificity. In contrast, we are convinced that epigenetic changes affect the chromatin structure at distinct loci and thereby determine recombination. The activity level of a given locus, more than age, gender or strain background itself, determine recombination, as discussed above. Therefore, our work is designed to offer a blueprint for urgently required future experiments that address the interdependence of physiological requirements at different ages, different animal activity levels, different gender and the activity of chromatin modifying enzymes.