A novel Axin2 knock‐in mouse model for visualization and lineage tracing of WNT/CTNNB1 responsive cells

Summary Wnt signal transduction controls tissue morphogenesis, maintenance and regeneration in all multicellular animals. In mammals, the WNT/CTNNB1 (Wnt/β‐catenin) pathway controls cell proliferation and cell fate decisions before and after birth. It plays a critical role at multiple stages of embryonic development, but also governs stem cell maintenance and homeostasis in adult tissues. However, it remains challenging to monitor endogenous WNT/CTNNB1 signaling dynamics in vivo. Here, we report the generation and characterization of a new knock‐in mouse strain that doubles as a fluorescent reporter and lineage tracing driver for WNT/CTNNB1 responsive cells. We introduced a multi‐cistronic targeting cassette at the 3′ end of the universal WNT/CTNNB1 target gene Axin2. The resulting knock‐in allele expresses a bright fluorescent reporter (3xNLS‐SGFP2) and a doxycycline‐inducible driver for lineage tracing (rtTA3). We show that the Axin2 P2A‐rtTA3‐T2A‐3xNLS‐SGFP2 strain labels WNT/CTNNB1 responsive cells at multiple anatomical sites during different stages of embryonic and postnatal development. It faithfully reports the subtle and dynamic changes in physiological WNT/CTNNB1 signaling activity that occur in vivo. We expect this mouse strain to be a useful resource for biologists who want to track and trace the location and developmental fate of WNT/CTNNB1 responsive stem cells in different contexts.


Multiple reporter lines allow visualization of cells with active
WNT/CTNNB1 signaling, either directly (e.g., GFP) or indirectly (e.g., lacZ). Some strains report active WNT/CTNNB1 signaling using Anoeska Agatha Alida van de Moosdijk and Yorick Bernardus Cornelis van de Grift contributed equally to this work. F I G U R E 1 Design and generation of the Axin2 P2A-rtTA3-T2A-3xNLS-SGFP2 knock-in allele. (a) Overview of WNT/CTNNB1 reporter and lineage tracing strains that are available from public repositories. Information was retrieved from the International Mouse Strain Resource (IMSR) at http://www.findmice. org. Strains are subdivided by four criteria: whether it is a random transgenic insertion (Tg) or a targeted insertion (Rosa26 or Axin2 locus); whether the targeted Axin2 allele still produces a functional AXIN2 protein; whether the strain carries a reporter that can directly visualize WNT/CTNNB1 responsive cells; whether the strain can be used for lineage tracing experiments. n.a., not applicable, FP, fluorescent protein, asterisk "*" requires Cre-mediated removal of a stop-cassette before the reporter becomes functional. (b) Schematic representation of the mouse Axin2 locus (chr11:108920349-108950781, mm10 coordinates). Most existing Axin2 knock-in strains target the 5 0 end of the gene by introducing the knock-in cassette at the start codon (ATG) in exon 2. This disrupts the endogenous Axin2 coding sequence. (c) Cartoon depicting the Axin2 P2A-rtTA3-T2A-3xNLS-SGFP2 targeting construct. A multicistronic targeting cassette was cloned immediately upstream of the Axin2 stop codon (TGA) in exon 11. The PGK-Neo-polyA cassette, flanked by FRT sites (indicated by triangles), was used for selection of embryonic stem cells and was removed prior to establishing the colony. (d) Targeted locus after removal of the PGK-Neo-polyA selection cassette. The approximate location of the primers used for the genotyping PCR depicted in (g) is indicated with equilibrium arrows. (e) The 3 0 knock-in allele is designed to give rise to a single transcript in which the Axin2 5 0 UTR, coding sequence and 3 0 UTR are left intact. (f) Following translation, the self-cleaving P2A and T2A sequences ensure that the polypeptide is cleaved into a fully functional AXIN2 protein, a doxycycline activatable rtTA3 driver for lineage tracing, and a bright green fluorescent protein that localizes to the nucleus (3xNLS-SGFP2). (g) Genotyping PCR of wildtype, heterozygous and homozygous animals. Wildtype allele = 257 bp, knock-in allele = 461 bp an artificial reporter construct with concatemerized TCF/LEF sites, analogous to the first in vivo WNT/CTNNB1 reporter mouse, TOP-GAL (Cho, Smallwood, & Nathans, 2017;DasGupta & Fuchs, 1999;Ferrer-Vaquer et al., 2010;Maretto et al., 2003). Although these transgenic strains continue to be highly useful, they likely do not fully recapitulate the endogenous pattern of WNT/CTNNB1 signaling activity due to integration of this reporter cassette into a random or exogenous locus (Cho et al., 2017;Yu, Liu, Costantini, & Hsu, 2007).
Moreover, different expression patterns can be detected when multiple lines are compared in the same tissue (Al Alam et al., 2011).
Although truly universal WNT/CTNNB1 target genes are rare, the negative feedback regulator Axin2 has been shown to reliably report cells with active WNT/CTNNB1 signaling (Lustig et al., 2001). Accordingly, it can be used to track the developmental fate of WNT/ CTNNB1 responsive stem cells in postnatal tissues (Van Amerongen, Bowman, & Nusse, 2012).
All existing models suffer from some drawbacks. Specifically, none of the available strains combine the expression of a sensitive fluorescent reporter gene and a lineage tracing driver from an unperturbed, physiological locus. Therefore, we generated a novel mouse strain, Axin2 P2A-rtTA3-T2A-3xNLS-SGFP2 , in which the knock-in cassette allows both direct visualization and doxycycline-dependent lineage tracing of WNT/CTNNB1 responsive cells. Our model faithfully recapitulates the endogenous Axin2 expression pattern, while leaving normal Axin2 expression intact.

| RESULTS AND DISCUSSION
We designed our new Axin2 knock-in allele to meet specific criteria.
First, we targeted the 3 0 end of the gene to maintain endogenous Axin2 expression and to preserve both 5 0 and 3 0 regulatory control (Figure 1b-d, Figure S1). Second, we wanted a single allele to function as both a reporter and a lineage tracing driver (Figure 1e,f). Third, we designed the reporter to serve as a direct readout of Axin2 activity and to be suitable for live cell imaging. For this reason, we incorporated a fluorescent protein rather than a lacZ reporter gene. Based on prior experience, we expected Axin2 expression to be relatively low. Therefore, we concentrated the reporter signal in the nucleus by fusing the SGFP2 gene to a strong nuclear localization signal (3xNLS).
Wildtype, heterozygous and homozygous Axin2 P2A-rtTA3-T2A-3xNLS-SGFP2 mice were born at the expected mendelian ratios from heterozygous intercrosses on a C57BL/6 background ( Figure 1g, Table S1). We have not detected any phenotype (including differences in weight, fertility and lifespan up to 12 months) associated with homozygosity of the knock-in allele (data not shown). Of note, we also did not observe any shortening of the skull (data not shown), a craniosynostosis phenotype that has been reported in homozygous Axin2 lacZ mice (Yu et al., 2005). This suggests that our 3 0 knock-in allele retains full functionality, in contrast to 5 0 Axin2 knock-in alleles, which by design disrupt endogenous Axin2 expression (de Roo et al., 2017;Lustig et al., 2001;Van Amerongen, Bowman, et al., 2012) and which may be haploinsufficient depending on the genetic background (Table S2).
For functional validation of the allele, we isolated primary mouse embryonic fibroblasts (MEFs) from E13.5 embryos that were heterozygous for the Axin2 P2A-rtTA3-T2A-3xNLS-SGFP2 allele. Only cells with activated WNT/CTNNB1 signaling are predicted to induce Axin2 and to express rtTA3, as well as a bright green fluorescent protein in the nucleus (Figure 2a,b). Quantitative RT-PCR analysis confirmed that  Figure S2).
Next, we tested if rtTA3 was capable of activating a tetO-responsive promoter in a doxycycline and CTNNB1-dependent manner (Figure 2i,j).
In MEFs heterozygous for both the Axin2 P2A-rtTA3-T2A-3xNLS-SGFP2 allele and a tetO-responsive FLAG-tagged transgene, FLAG-tagged protein expression was only induced in the presence of both CHIR99021 and doxycycline, but not by either treatment alone ( Figure 2j).
To determine if the knock-in allele properly reports WNT/ CTNNB1 signaling during embryonic development, we isolated embryos from stages at which defined Axin2 expression domains were previously determined using RNA in situ hybridization (Jho et al., 2002). A dim SGFP2 signal could be detected in freshly isolated heterozygous or homozygous, but not wildtype embryos under a dissecting microscope ( Figure S3). Homozygous embryos consistently showed a brighter GFP signal than heterozygous embryos, suggesting bi-allelic expression of Axin2.
At E8.5, SGFP2 was expressed in the head folds and the posterior neural tube (Figure 3a-c). At E10.5, we detected SGFP2 expression in the developing limb bud, in the branchial arches, and along the dorsal neural tube, including the roof plate of the developing brain ( Figure 3d-f). These are the same sites where endogenous Axin2 was previously shown to be expressed via RNA in situ hybridization (Jho et al., 2002). At E12.5, prominent SGFP2 expression was visible in the mammary placodes (Figure 3g Using wholemount confocal microscopy, individual SGFP2-positive nuclei could readily be distinguished in these embryos at the indicated sites ( Figure 3c,e,f,i, Movies S1 and S2). Together, these experiments confirm that Axin2 P2A-rtTA3-T2A-3xNLS-SGFP2 reports endogenous WNT/CTNNB1 signaling in the developing embryo.
To test functionality of the lineage tracing module in vivo, we generated compound Axin2 P2A-rtTA3-T2A-3xNLS-SGFP2 ; tetO-Cre; Rosa26 mTmG mice and labeled WNT/CTNNB1-responsive cells in E7.5 embryos by a single intraperitoneal injection of doxycycline into timed pregnant females. At E10.5, triple-heterozygous embryos showed prominent recombination of the Rosa26 mTmG reporter in the posterior half of the embryo (Figure 3j-l), consistent with posterior expression of Axin2 at the time of induction at E7.5 (Jho et al., 2002). Thus, F I G U R E 2 Legend on next page. Postnatally, SGFP2 expression was readily visualized in the small intestinal crypt. The strongest signal was found at the crypt bottom in so-called crypt-base-columnar cells (CBCs), which are known to be WNT/CTNNB1 responsive (Barker et al., 2007). These cells are readily identified by their distinct shape and presence among SGFP2-negative Paneth cells (Figure 4a,b). In adult mice, a gradient of SGFP2 signal extended into the upper portion of the crypt (Figure 4a).
Axin2 expressing cells in the intestine could also be detected by immunohistochemical staining of formalin fixed paraffin embedded (FFPE) tissue sections with an anti-GFP antibody ( Figure S4).
In adult triple-heterozygous Axin2 P2A-rtTA3-T2A-3xNLS-SGFP2 ; tetO-Cre; Rosa26 mTmG mice, Cre/lox mediated recombination of the Finally, we established primary 3D small intestinal organoid cultures from Axin2 P2A-rtTA3-T2A-3xNLS-SGFP2 animals. One heterozygous and one homozygous organoid line were maintained for extended cultures of more than 6 months, during which the reporter remained stably expressed and did not get silenced ( Figure 5, Figure S8). Similar to what we observed in embryos, the levels of SGFP2 were higher in homozygous than in heterozygous organoids ( Figure S8a We also demonstrate that the new, doxycycline-inducible lineage tracing driver is suitable for embryonic and postnatal traces ( Figures 3 and 4). At the same time, our model does not fully F I G U R E 2 Functional validation of the Axin2 P2A-rtTA3-T2A-3xNLS-SGFP2 allele. (a and b) Cartoons depicting the response of the knock-in allele to downstream WNT/CTNNB1 signaling, either by activation by WNT proteins in vivo (a) or by treatment with CHIR99021, a specific GSK3 inhibitor (b). Cells with active WNT/CTNNB1 signaling will induce Axin2, resulting in the expression of 3xNLS-sGFP expression (depicted as green nuclei). (c-f) Dot plots showing the dose-dependent induction of Axin2 mRNA expression in wildtype (c) and heterozygous (d) mouse embryonic fibroblasts (MEFs), as measured by qRT-PCR in n = 3 independent MEF isolates. Rpl13a was used as a reference gene, values in the DMSO treated control were set to 1. (e and f) Same as for (c and d), but showing rtTA3 (e) and SGFP2 (f) mRNA expression in heterozygous MEFs. (g) Confocal microscopy images of fixed mouse embryonic fibroblasts (MEFs), showing the dose-dependent induction and direct detection of SGFP2. Scalebar is 100 μm. (h) Quantification of the experiment described and depicted in (g). Fixed MEFs were counterstained with DAPI to allow nuclear segmentation, after which the relative SGFP2 expression levels were calculated by correcting for the fluorescence intensity of the DAPI signal. This experiment was performed for n = 3 independent MEF isolates. One experiment is shown here. The results for two additional MEF isolates are shown in Figure S2. (i) The rtTA3 transcriptional activator is induced together with SGFP2, but only becomes active in the presence of doxycycline (DOX). This results in activation of tetO-driven transgenes (tetO-Tg). (j) Western blot showing the WNT/CTNNB1dependent induction of SGFP2 and the WNT/CTNNB1-and DOX-dependent induction of a FLAG-tagged protein expressed from a tetOresponsive allele in whole cell lysates from MEFs after treatment with different combinations of CHIR99021 (5 μM) and DOX (1 μg/ml). These MEFs were isolated from embryos carrying both the Axin2 P2A-rtTA3-T2A-3xNLS-SGFP2 knock-in allele and a tetO-responsive FLAG-Wnt5a transgene recapitulate the results that we previously obtained with the 5 0 Axin2 lacZ or 5 0 Axin2 CreERT2 knock-in models (Lim et al., 2013;Van Amerongen, Bowman, et al., 2012). Specifically, both lacZ reporter gene expression and Cre ERT2 driven recombination were detected in a larger set of adult cell types than sGFP2 reporter expression and rtTA3-driven Cre/lox recombination in our new 3 0 knock-in model.
This once again underscores that care should be taken when interpreting results from genetic reporters and, in general, when comparing expression patterns across different mouse strains. The discrepancy between the different Axin2 models can be explained by one or more of the following reasons. First, our new knock-in model preserves both 5 0 and 3 0 transcriptional and translational regulatory control. It thus fully recapitulates endogenous Axin2 expression. The direct detection of a fluorescent reporter reveals how low these expression levels really are, especially after birth. This is in agreement with the low-level and low-amplitude oscillations of Axin2 expression detected by others (Sonnen et al., 2018) and something that may be obscured by the enzymatic amplification that occurs during X-gal staining in Axin2 LacZ mice. It may also explain reduced labeling efficiency in postnatal lineage tracing experiments. Second, the tetO-Cre line used for our studies is not leaky, but may display a high threshold to activation. In combination with the low Axin2 (and thus low rtTA3) expression this may result in insufficient Cre activity to recombine the Rosa26 mTmG allele.
Whether rtTA3 or tetO-Cre, or a combination of both, is the rate lim- iting factor remains to be tested. Labeling efficiency might be increased in homozygous Axin2 P2A-rtTA3-T2A-3xNLS-SGFP2 animals or in combination with a different tetO-Cre strain.
Summarizing, our bright fluorescent nuclear SGFP2 reporter offers straightforward detection of Axin2 expression and is expected to be compatible with live cell imaging, nuclear segmentation and thus automated cell tracking in embryonic tissues and organoids. The Axin2 P2A-rtTA3-T2A-3xNLS-SGFP2 reporter is sensitive to direct changes in endogenous WNT/CTNNB1 signaling and allows monitoring of WNT/CTNNB1 sensitive cells. As such, our model can be a useful tool for the scientific community to simultaneously visualize and trace cells with active WNT/CTNNB1 signaling both in vivo and in vitro.
The Axin2 P2A-rtTA3-T2A-3xNLS-SGFP2 mouse line was generated by Ozgene Pty Ltd (Bentley WA, Australia). The final targeting construct, containing 5.5 kb and 2.8 kb homology arms, was linearized and electroporated into Bruce 4 [B6.Cg-Thy1] ES cells (Köntgen, Süss, Stewart, Steinmetz, & Bluethmann, 1993). Homologous recombinant ES cell clones (7/95) were identified by Southern hybridization and 2 independent clones were injected into goGermline blastocysts (Koentgen et al., 2016). Male chimeric mice were obtained and (d) Wholemount E10.5 homozygous embryo imaged under a fluorescence dissecting microscope. Arrowheads indicate expression in the developing limb bud and branchial arches, along the dorsal neural tube, and in the roof plate of the developing brain. Scalebar is 500 μm. The same embryo is depicted in Figure S2. (e and f) Wholemount confocal microscopy images of the roof plate (e) and forelimb bud (f) of an E10.5 heterozygous embryo. Individual SGFP2-positive nuclei can be distinguished and are indicated with arrowheads. Scalebar is 50 μm for (e) and 100 μm for (f). (g) Wholemount E12.5 homozygous embryo imaged under a fluorescence dissecting microscope. Scalebar is 1 mm. (h) Close up of the area highlighted in (g). Arrowheads indicate expression of the fluorescent reporter in the mammary buds. Scalebar is 500 μm. (i) Maximum intensity projection of a wholemount confocal microscopy Z-stack of the E12.5 mammary bud, showing the difference in GFP intensity between a heterozygous (HET, top) and homozygous (HOM, bottom) embryo. Nuclear SGFP2 signal can be detected in both the epithelial bud and the surrounding mesenchyme. A 3D rotation of the entire Z-stack is provided in Movies S1 (HET) and S2 (HOM). (j) Timeline for the experiment depicted in (l). Lineage tracing in WNT/CTNNB1-responsive cells was induced in utero at E7.5 by administering a single intraperitoneal injection of doxycycline (DOX) to pregnant females. Embryos were isolated and imaged at E10.5. (k) Lineage tracing principle using the Rosa26 mTmG reporter. Prior to Cre-mediated recombination, all cells express a membrane-localized red fluorescent protein (mT). Cre recombination results in a permanent switch to expression of a membrane-localized green fluorescent protein (mG). (l) Composite image showing Cre/lox dependent recombination in the posterior half of an Axin2 P2A-rtTA3-T2A-3xNLS-SGFP2 ; tetO-Cre; Rosa26 mTmG triple heterozygous embryo. Red signal reflects nonrecombined cells (mT), green signal reflects recombined cells (mG). Note that expression of membrane-bound GFP in the Rosa26 mTmG reporter is driven by a strong CAGGS promoter, such that with the imaging and filter settings chosen the weaker endogenous 3xNLS-SGFP2 signal of the Axin2 knock-in allele itself is not detected. The embryonic timed matings and tracing experiments were performed in two independent litters for all timepoints. Scalebar is 500 μm

| PCR genotyping
Ear clips or yolk sacs were lysed either overnight (ear clips) or for 2 hr (yolk sacs) at 55 C in 200 μl of Direct PCR tail lysis buffer (Viagen) supplemented with 200 μg/ml Proteinase K (20 mg/ml stock solution).
Proteinase K was inactivated by incubating the samples at 85 C for 15-45 min. Samples were cooled to room temperature and spun down (2 min at 14,000 rpm), after which 1 μl of the supernatant was used as input for a standard 20 μl PCR reaction with 0.4 μl of Phire II polymerase (ThermoFisher, #F-124S). PCR conditions were as follows: initial denaturation at 98 C for 30 s, followed by 30 or 35 cycles of denaturation at 98 C for 5 s, annealing at the relevant temperature for 5 s, extension at 72 C for 10 s, followed by a final extension step of 72 C for 1 min. Samples were cooled to 16 C and analyzed on a 2% agarose gel in standard TAE buffer. Primer sequences, annealing temperatures, number of cycles and expected band sizes are detailed in Table S3.

| Isolation and culture of mouse intestinal organoids
Mouse intestinal organoids were established as previously described (Sato et al., 2009). Crypts were isolated from the entire length of the The asterisk indicates dead material excreted into the organoid lumen that gives autofluorescence in the SGP2 channel. SGFP2 is depicted in a red look up table for better visual representation of the subtle differences in combination with nuclear and membrane stainings. Note that this organoid line is also heterozygous for the Rosa26 mTmG allele, which is used as a membrane marker in panels (c-j). (b) Close-up of the highlighted crypt section from (a). Scalebars are 30 μm (a) and 15 μm (b). For (a and b) a total of 17 organoids were imaged in a single experiment. A representative image is shown. (c) Confocal microscopy image of two crypts from a fixed Axin2 P2A-rtTA3-T2A-3xNLS-SGFP2 HOM ; Rosa26 mTmG HET small intestinal organoid. Membrane-tdTomato is depicted in gray and 3xNLS-SGFP2 expression in red. 3xNLS-SGFP2 is expressed in a gradient along the crypt-axis, the highest expression is located at the crypt bottom and gradually declines when cells move upward. Scalebar is 30 μm. (d) Heat map showing the relative Axin2 expression superimposed on the confocal image in (c). The heat map depicts the 3xNLS-sGPF2 signal relative to a DRAQ5 nuclear staining (not shown) and is imposed on the area of the nuclei. This overlay corrects for signal intensity differences due to imaging depth and reveals differences in gene expression. Scalebar is 30 μm. For panels (c and d), a total of 14 organoids were imaged in two independent experiments. A representative image is shown. (e and f) Confocal microscopy images of crypts from fixed Axin2 P2A-rtTA3-T2A-3xNLS-SGFP2 HOM ; Rosa26 mTmG HET (3xNLS-SGFP2, red) small intestinal organoids stained with nuclear dye DRAQ5 (gray). The organoids were treated with either DMSO for 24 hr (e), CHIR99021 for 24 hr (f), or IWP2 for 48 hr (g) before fixation. Scalebar is 30 μm. (h-j) Heat map showing relative Axin2 reporter levels superimposed on the confocal images of (e and f). The dotted lines indicate the cutoff for which the nuclei were excluded from further analysis. Note that this relative scale is not directly comparable to the one depicted in (d). Scalebar is 30 μm. For panels (e-j) a total of 51 organoids were imaged in two independent experiments (n = 14 for DMSO, n = 29 for CHIR99021, n = 8 for IWP2). Representative images of all conditions are shown small intestine of one Axin2 P2A-rtTA3-T2A-3xNLS-SGFP2(HOM) ; Rosa26 mTmG(HET) animal (used for Figure 5), one Axin2 P2A-rtTA3-T2A-3xNLS-SGFP2 (HOM) animal and one Axin2 P2A-rtTA3-T2A-3xNLS-SGFP2(HET) animal (used for the comparison of heterozygous and homozygous organoids in Figure S8) and used to establish individual organoid lines. Organoids were cultured in 10 μl Matrigel droplets (Corning) in culture medium containing advanced DMEM/F12 (ThermoFisher Scientific) supplemented with 100 U/ml Penicillin/Streptomycin (ThermoFisher Scientific), 2 mM Glutamax (ThermoFisher Scientific), 10 mM HEPES (ThermoFisher scientific), 1× B27 supplement (ThermoFisher Scientific) and 1.25 mM N-acetylCysteine (Sigma Aldrich), freshly added EGF (50 ng/ml, PeproTech), recombinant murine Noggin (100 ng/ml, PeproTech) and recombinant murine R-spondin 1 (500 ng/ml, Sinobiological Inc.) at 37 C and 5% CO 2 . For passaging, cell culture medium was removed and Matrigel was broken into small pieces by scraping, followed by vigorous pipetting with ice-cold advanced DMEM/F12. Crypts were centrifuged at 200g for 5 min at 4 C. The supernatant was carefully removed, the pellet was resuspended in Matrigel and plated on pre-warmed plates. Medium was refreshed every other day and organoids were split once a week in a 1:3 ratio.

| Microscopy image analysis
Fluorescence microscopy images were processed in Fiji (Schindelin et al., 2012) using the Image 5D plugin and custom built approaches.
For the quantification of SGFP2 expression in MEFs (Figure 2c,d and Figure S2), ROIs surrounding single nuclei were selected manually in the DAPI channel using the magic wand tool. The SGFP2 nuclear signal was normalized over DAPI intensity. For each condition, three images were analyzed. This was repeated for three independent MEF lines. Data for one of the lines is shown in Figure 2, data for the other two lines are shown in Figure S2. For Figure 5, heatmaps were drawn as follows: ROIs surrounding single nuclei were selected manually in the DRAQ5 channel using the Freehand selection tool. The SGFP2 nuclear signal was normalized over DRAQ5 intensity. To determine the relative Axin2 expression a ratio of SGFP2 over DRAQ5 was calculated in Excel and imported in Fiji. An overlay was made that draws a heat map in selected ROIs based on this ratio with the ROI Color Coder plugin (Ferreira et al., 2017).
Color scheme choices: SGFP2 signal is shown in green, except for Figure 5 and Figure S6, where a red LUT was chosen to be able to overlay the SGFP2 and the nuclear signal. All lineage tracing experiments are depicted in green and red to maintain the original mTmG reporter set up (i.e., a switch from tdTomato to eGFP).

ACKNOWLEDGMENTS
We thank our animal caretakers for taking daily care of the mice, the