Ankyrin-G is an intracellular scaffolding protein critical for the formation and maintenance of the AIS, where it recruits multiple cytoskeletal elements, ion channels, transporters, and cell adhesion molecules that are necessary for normal AIS function (Nelson and Jenkins, 2017; Rasband, 2010; Leterrier, 2018). We, therefore, decided to use ankyrin-G as a backbone to design a conditional mouse model to intrinsically label the AIS (see Methods). Ankyrin-G tolerates the addition of a GFP tag, since ankyrin-G-GFP can rescue multiple functions of ankyrin-G in Ank3 null neurons, including scaffolding of voltage-gated sodium and potassium channels, clustering of the spectrin cytoskeleton, and recruitment of inhibitory synapses (Jenkins et al., 2015; Fréal et al., 2016; Tseng et al., 2015). To intrinsically label ankyrin-G with GFP, we utilized a FLEx cassette system to flip out the last exon of Ank3 and replace it with the last exon of Ank3 fused with the coding sequence for GFP (Figure 1). Given the heterogeneous expression of ankyrin-G in various cell compartments across multiple tissues, constitutive tagging of the endogenous Ank3 allele would make it difficult to visualize ankyrin-G-dependent structures. By contrast, our system allows us to only GFP-label ankyrin-G in cells expressing Cre recombinase, allowing the visualization of ankyrin-G in genetically and temporally defined cell populations.

When generating GFP fusion proteins, it is critical to confirm that GFP does not interfere with normal protein function. To this end, we utilized the linker-GFP sequence from the rescue plasmids used in multiple studies from our group and others to examine the function of the three main classes of brain ankyrin-G isoforms, 190, 270, and 480 kDa, respectively. Importantly, these GFP fusions have been shown to be sufficient to rescue Ank3 knockout or silencing in the formation of the AIS, noR, GABAergic synapses, and dendritic spines (Jenkins et al., 2015; Fréal et al., 2016; Tseng et al., 2015; Smith et al., 2014).

To characterize the ankyrin-G-GFP (ank-G-GFP) mouse, we utilized several of the most common recombination systems and experimental paradigms used in cellular neuroscience and confirmed intrinsic AIS features in different neuronal cell types in vitro, ex vivo, and in vivo across multiple brain regions (Figure 2A). Ank-G-GFP mice were first used to generate hippocampal OTC, which were then exposed to AAV expressing Cre-recombinase under the control of the synapsin promotor (AAV5-pmSyn1-EBFP-Cre). After 48 hr, neurons exposed to the virus showed GFP expression along their AIS (Figure 2B, left panel). We confirmed the GFP signal as originating from AIS by staining against βIV-spectrin (Figure 2B, middle panel). Not all AIS in OTC neurons were GFP⁺, likely due to the transfection efficiency with AAV-Cre. By directly injecting viruses, and by crossbreeding ank-G-GFP mice with different Cre driver lines, we overcame this limitation and used acute slices or cryosections to verify efficient labeling of the AIS. The injection of ank-G-GFP animals with a synapsin-Cre-tdTomato AAV resulted in ank-G-GFP⁺-AIS (Figure 2C, left panel), which again were βIV-spectrin⁺ as well (Figure 2C, middle panel). Of note, the somatic envelope and proximal dendrites were also visible via the ank-G-GFP signal (Figure 2C, asterisks in left panel), highlighting ankyrin-G expression in the somatodendritic compartment of mature neurons (Tseng et al., 2015; Nelson et al., 2020). Likewise, using a combination injection of AAV1-hDlx-Flex-dTomato-Fishell_7 and AAV1-hSyn.Cre.WPRE.hGH, allowed visualization of ank-G-GFP⁺-AIS in hippocampal interneurons via tdTomato expression (Figure 2D).

Mating of ank-G-GFP mice with a line expressing Cre recombinase under the control of the Calcium/calmodulin-dependent protein kinase type II alpha chain (CaMKIIα) promoter resulted in reproducible reporter detection in CaMKII-expressing neurons in layer II neocortical pyramidal neurons (Figure 2E). Neurons not expressing CaMKII were negative for ank-G-GFP, but as expected, additional staining against βIV-spectrin (Figure 2E, middle panel) resulted in labeling of all AIS in the section, allowing for clear distinction between the different populations (ank-G-GFP⁺-AIS vs ank-G-GFP^–-AIS).

Ank-G-GFP expression is not limited to typical cortical neurons (pyramidal cells, interneurons). We found reliable ank-G-GFP expression in retinal whole-mount preparations, both in peripheral and central retinal ganglion cells (RGC; Figure 2F; ank-G-GFP x CaMKII-Cre animals). Colabeling of ank-G-GFP⁺ AIS with antibodies directed against native ankyrin-G in RGC indicated that the vast majority of RGC express CaMKII and consequently, their AIS are positive for both GFP and the intrinsic AIS marker ankyrin-G (Figure 2F).

Interneuron populations showed similar selective labeling when using appropriate Cre driver lines. Mating of ank-G-GFP animals with parvalbumin (PV)-Cre mice highlighted exclusive GFP signal in AIS that belonged to PV⁺ interneurons as shown in ex vivo acute slices of the cerebellum (Figure 2G, left panel). Again, double labeling against βIV-spectrin indicated distinct GFP expression only in those neurons that expressed PV (Figure 2G, middle panel). In all experiments outlined above, double labeling against the characteristic AIS markers βIV-spectrin and ankyrin-G, respectively, served as control of AIS identity of labeled structures and we found that ank-G-GFP reporter expression always colocalized with these markers.

The 480 kDa isoform of ankyrin-G is also localized to nodes of Ranvier (noR) in myelinated axons where it serves as a scaffold for voltage-gated ion channels similar to those found at the AIS (Jenkins et al., 2015; Jenkins and Bennett, 2002; Dzhashiashvili et al., 2007). Consequently, the ank-G-GFP construct should also be discernible in this axonal microdomain. In mice expressing ank-G-GFP in CaMKIIa-positive excitatory neurons, noR in the alveus of hippocampal CA3 were visible as distinct puncta that co-localized with other nodal markers such as ankyrin-G and Na_v1.6 (Figure 3A). Of note, only CaMKII-expressing neurons showed ank-G-GFP signals in noR (Figure 3A, green circles). Ank-G-GFP^– noR were still characterized by the colocalization of ankyrin-G and Na_v1.6 (Figure 3A, blue circles). To confirm that the expression of ank-G-GFP does not disrupt the nodal structure, we examined the localization of the paranodal Contactin-associated protein (Caspr) (Rios et al., 2000). Immunostaining against GFP, ankyrin-G, and Caspr revealed normal localization of ank-G-GFP and Caspr to the node and paranode, respectively, in myelinated axons of CaMKII-positive neurons (Figure 3B). The quantification of node shape and fluorescence signals of Na_v1.6 and ankyrin-G showed no differences between GFP⁺ and GFP^- nodes (Figure 3C, Figure 3—figure supplement 1A and B). We note that these groups belong to genetically distinct subgroups of cells, but consistent node lengths and Na_v1.6 levels indicate that the GFP label does not grossly disrupt node structure. Furthermore, GFP label intensities within GFP⁺ nodes did not interfere with Na_v1.6 levels (Figure 3—figure supplement 1C, top panel). Instead, we found a strong correlation between GFP and ankyrin-G in GFP⁺ nodes, demonstrating the specificity of our model (Figure 3—figure supplement 1C, bottom panel).

The AIS and its niche are characterized by a number of intra- and extracellular components contributing to its unique structure and function (Leterrier, 2018). To investigate whether the fusion of GFP to endogenous ankyrin-G has any impact on core AIS features, we selected additional targets for immunolabeling (Figures 4 and 5) and compared whether they are affected by the GFP modification. We found that the length and distance of AIS (measured by βIV-spectrin signal) did not differ between neighboring GFP⁺ and GFP^- CA1 pyramidal cells (Figure 4—figure supplement 1A). Length measurements using the live ankyrin-GFP label in combination with post-fixed ankyrin-G or βIV-spectrin of the same AIS showed a strong correlation with one another (Figure 4—figure supplement 1B–C).

A major feature of the AIS is its endowment with voltage-gated ion channels, the anatomical prerequisite for proper AP generation and propagation (Kole et al., 2008; Kole et al., 2007; Bender and Trussell, 2009). Colabeling of the sodium channel isoform Na_v1.6 with βIV-spectrin in ank-G-GFP⁺ AIS in cortical layer V pyramidal neurons showed homogeneous signals of Na_v1.6 across the length of neighboring GFP⁺ and GFP^- AIS. Note that previously published data found a preference of Na_v1.6 for distal expression at the AIS in the layer V prefrontal cortex (instead of layer II somatosensory cortex; Van Wart et al., 2007; Van Wart et al., 2007 Figure 4A, Figure 4—figure supplement 1D). Likewise, a similar expression of K_v2.1 was observed in GFP⁺ and GFP^- AIS (Figure 4B, Figure 4—figure supplement 1E).

The AIS is also characterized by the expression of several cell adhesion molecules, chief among them neurofascin-186 (NF-186), which is important for the recruitment of ion channels to the AIS during development (Hedstrom et al., 2007; Alpizar et al., 2019). Potential disruption of normal AIS assembly in our reporter model may have an impact on the assembly of the extracellular niche at the AIS, however, immunostaining against NF-186 in ank-G-GFP⁺ AIS produced the expected co-staining pattern (Figure 4C). The AIS-related proteins TRIM46, organizing the fasciculation of microtubules in the AIS (Fréal et al., 2019; Harterink et al., 2019) and FGF14, regulating voltage-gated ion channels at the AIS (Pablo and Pitt, 2017), were also detected in ank-G-GFP⁺ neurons throughout the neocortex of ank-G-GFP x CaMKII-Cre mice (Figure 4D and E).

Next, we investigated whether the cisternal organelle (CO), an AIS-specific intra-axonal Ca²⁺-store, remained unperturbed by ank-G-GFP expression. The actin-binding protein synaptopodin (synpo) has been shown to be a reliable and reproducible marker for the CO in the AIS (Hanemaaijer, 2020; Schlüter et al., 2017; Sánchez-Ponce et al., 2012). Immunostaining with antibodies against GFP, synpo, and βIV-spectrin in layer V pyramidal neurons of primary motor cortex of ank-G-GFP x CaMKII-Cre mice revealed labeling of the CO within the AIS of ank-G-GFP⁺-neurons (Figure 5A) with the typical synpo⁺ spine apparatus as part of the dendritic domain visible in all samples (Figure 5A, magenta puncta in the background). Three-dimensional reconstruction of ank-G-GFP⁺ AIS indicated numerous synpo clusters residing within the axonal compartment as previously reported (Schlüter et al., 2017; Schlüter et al., 2019; Bas Orth et al., 2007). To test whether GFP-expression had an impact on the normal distribution of synpo clusters in pyramidal neurons, we quantified cluster number per AIS and compared GFP^- and GFP⁺ AIS from the same ank-G-GFP x CaMKII-Cre mice. No significant difference between GFP^- and GFP⁺ AIS was observed (GFP^- AIS 3.1±1.7 clusters vs GFP⁺ 3.3±1.7 clusters; n=50 AIS from three animals, p=0.6444, unpaired t-test; Figure 5A).

Another hallmark of the AIS is its axo-axonic innervation pattern, which we tested using immunostaining against ank-G-GFP, the vesicular GABA transporter (vGAT), a reliable presynaptic marker (Bragina and Conti, 2018), and βIV-spectrin (Figure 5B). Data revealed that indeed, ank-G-GFP⁺-AIS were contacted by vGAT⁺ axo-axonic interneurons (Figure 5B). Conceivably, GFP-expression in ank-G-GFP mice could have an impact on the number of putative GABAergic synapses at the AIS. Therefore, we quantified the number of vGAT⁺ puncta along GFP^- and GFP⁺ AIS from the same ank-G-GFP x CaMKII-Cre mice. We found no significant difference between these groups (GFP^- AIS 20.7±2.3 vGAT puncta vs GFP⁺ 20.8±2.2 vGAT puncta; n=50 AIS from three animals, p=0.7917, unpaired t-test; Figure 5B). We note that GFP^- and GFP⁺ AIS belong to genetically different cell types (defined by CamKIIa expression) that might not necessarily feature the same synpo or vGAT distributions.

A further characteristic of the AIS is its occasional ensheathing in microglial processes, which is the case for approximately 4% of all AIS in rodent cortex (Baalman et al., 2015). We, therefore, performed immunostaining against the microglial marker Iba1 in ank-G-GFP × CaMKII mice and found scarce, but clearly distinguishable interaction of select GFP⁺ AIS with microglial processes in S1 layer II/III pyramidal neurons (Figure 5C). In those cases, the majority of the microglial cell body was in contact with the soma of a pyramidal neuron and extended one or more processes alongside the AIS towards its distal end (Figure 5C).

The nanostructure of the AIS scaffold in cortical neurons exhibits a stereotypical subcellular organization, in which a periodic spacing of actin and spectrin rings intercalates with voltage-gated ion channels at approximately 170–190 nm distance, depending on the cell type studied (Leterrier et al., 2015; Schlüter et al., 2019; Akter et al., 2020). Both concentration and spatial arrangement of ion channels are thought to be crucial for AP generation and modulation (Lazarov et al., 2018; Hu et al., 2009; Inda et al., 2006; Kress et al., 2010) and, therefore, constitute a hallmark of excitable neuronal microdomains such as the AIS.

Here, we aimed to verify that the introduction of GFP into this scaffold did not disturb its spatial organization. Dissociated hippocampal neurons derived from the ank-G-GFP line without prior Cre exposure were infected with lentiviruses expressing functional Cre-recombinase (+Cre) or a deficient version of Cre without recombinase activity as control (ΔCre), both fused to nuclear GFP (nGFP) (Figure 6). Conventional confocal microscopy after immunostaining against the AIS marker βIV-spectrin showed that ΔCre neurons exhibited no GFP staining at their βIV-spectrin⁺ AIS (Figure 6A left), whereas intrinsic GFP labeling of the AIS was achieved in cultures exposed to functional Cre-recombinase (Figure 6A middle). Using Stimulated Emission Depletion (STED) microscopy, we assessed the nanoscale distribution of ankyrin-G, both via intrinsic GFP label as detected by the FluoTag-X4-GFP nanobody (Figure 6A right, upper panel) as well as via antibody staining directed against the C-terminus of ankyrin-G (Figure 6A right, lower panel). To further elucidate the nanoscale distribution of AIS proteins in control vs Cre-recombinase exposed cells, we performed immunostaining against the C-terminus of ankyrin-G (Figure 6B, left panel), βIV-spectrin (Figure 6B, middle panel), and the voltage-gated potassium channel K_v1.2 (Figure 6B, right panel), which has been shown to cluster at the AIS (Lorincz and Nusser, 2008). Comparing ΔCre to +Cre neurons, we found an equally reproducible staining for all three markers (compare upper and lower rows in Figure 6B). Autocorrelation analysis verified the characteristic 190 nm periodic organization of the proteins investigated (Figure 6C; left = ankyrin G, middle = βIV-spectrin, right = K_v1.2). Further, an autocorrelation amplitude analysis, which calculates the difference between the autocorrelation value at 190 nm and the average values at 95 nm and 285 nm, respectively, was performed, showing no significant difference between untreated, ΔCre or +Cre ank-G-GFP cultures (Figure 6D; left = ankyrin G, middle = βIV-spectrin, right = K_v1.2). In summary, our data indicate that ank-G-GFP⁺-AIS develop normally and upon Cre-activation, retain their nanoscale organization.

In order to utilize the ank-G-GFP line to study electrophysiological consequences associated with AIS function and plasticity, we verified that the fusion protein itself does not change the basic excitatory properties of ank-G-GFP⁺ neurons in different systems. Therefore, we investigated whether cells expressing the fusion protein exhibit different electrophysiological properties as compared to wild-type cells. We used two common model systems: (i) dissociated hippocampal neurons infected either with Cre recombinase or a recombination-deficient ΔCre virus (Figure 7A and B), and (ii) hippocampal neurons in acute slices of adult mice that were generated from the ank-G-GFP x CaMKII-Cre matings. These were compared to acute slices from both ank-G-GFP littermate controls not exposed to Cre, and standard wild-type C57Bl/6 J mice (Figure 7C and D).

In dissociated hippocampal neurons, we first determined passive membrane properties such as resting membrane potential (RMP) and input resistance (R_s) and found no significant difference between neurons expressing Cre-recombinase and ΔCre controls (Figure 7B). Similarly, AP properties, including amplitude, half-width, voltage threshold, maximum rise time, and rheobase were unaffected (Figure 7B). Plotting the input/output relation with a comparison of the maximum slope f’ (max) of the I/F-curve further provided evidence that the GFP fusion construct does not alter the firing properties of infected neurons in vitro.

The same electrophysiological parameters were tested in ex vivo hippocampal acute slices from three groups: (i) wild-type mice, (ii) ank-G-GFP mice without Cre exposure, and (iii) ank-G-GFP x CamKII-Cre mice (Figure 7C and D). We found no evidence of an impact of the GFP fusion protein on any of the investigated passive and active properties (Figure 7C and D, Supplementary file 1A).

Previous post-hoc studies of AIS plasticity indicated that the AIS scales its length, and sometimes position, in response to changes in network activity (reviewed in Jamann et al., 2018). Using the new live reporter, we tested whether the same principles apply to ank-G-GFP⁺-AIS. We employed previously reported experimental conditions of network alterations Jamann et al., 2021; Grubb and Burrone, 2010; Engelhardt et al., 2018 to elicit structural AIS changes ex vivo (Figure 8A). To that end, hippocampal OTC were prepared from ank-G-GFP mice at P4 and transfected with an AAV pmSyn1- EBFP-Cre virus at DIV 1 to trigger GFP expression.

Since previous research indicated that AIS elongation is a process that requires a longer time frame of network silencing (days to weeks; Jamann et al., 2021; Gutzmann et al., 2014), we first blocked synaptic transmission via NMDA receptors by maintaining OTC in the presence of 10 mM MgSO₄ for 20 d. Under these conditions, AIS elongated significantly (Ctrl 27.9±3.0 µm vs MgSO₄ 32.1±2.4 µm, n=6 OTC, 50 AIS per OTC; Figure 8A and B). The same effect was observed when cultures were maintained in MgSO₄ for 10 d (Ctrl 32.1±3.5 µm vs MgSO₄ 35.9±1.2 µm, n=6 OTC, 50 AIS per OTC; Figure 8C). Induction of increased network activity by exposing OTC to chronic depolarizing conditions (6 mM KCl) for 20 d and 10 d, respectively, did not result in significant length changes (KCl_{(10 d)} 29.9±1.8 µm, KCl_{(20 d)} 27.0±2.7 µm, n=6 OTC, 50 AIS per OTC; Figure 8B and D, one-way ANOVA and Tukey’s test for B, unpaired t-tests for C and D).

AIS plasticity is proposed as a homeostatic mechanism (Wefelmeyer et al., 2016; Jamann et al., 2021) and should be reversible as network activity returns to baseline. We, therefore, tested the reversibility of AIS plasticity by applying rescue conditions. After initial exposure to the activity-altering conditions for 10 d, OTC were cultured under normal growth conditions for an additional 10 d (to DIV 20; Figure 8A). In MgSO₄-treated OCT, AIS length returned to baseline after rescue (Ctrl MgSO₄ 27.9±3.0 µm vs rescue 30.6±2.3 µm; n=6 OTC, 50 AIS per OTC; Figure 8C). Rescue paradigms in KCl-treated cultures did not affect AIS length as expected (Ctrl KCl 27.9±3.0 µm vs rescue 28.9±1.9 µm; n=6 OTC, 50 AIS per OTC; Figure 8D, all tests performed as unpaired t-test).

AIS shortening after increased network activity occurs in a rapid time frame (1 hour) and is bidirectional with a return to baseline after an initial length change (Jamann et al., 2021). To test whether rapid AIS plasticity could be observed live in ank-G-GFP mice, we applied a different strategy. Ank-G-GFP mice were injected with an AAV5-CaMKIIa-Cre virus triggering Cre expression in a subset of hippocampal pyramidal neurons (Figure 8E). After 3 wk of virus expression, stable ank-G-GFP signals were detected in a subset of pyramidal neurons using epifluorescence in acute slices. Using pipette application of 15 mM KCl targeting individual neurons, we observed rapid changes in fluorescent signals based on the intrinsic GFP (Figure 8E). Direct application of high potassium resulted in a relative length reduction only of the AIS associated with the affected neuron, while all surrounding labeled AIS maintained their initial length (n=4 pairs of sample and control AIS; 10 min Ctrl vs KCl: p=0.0068; 20 min Ctrl vs KCl: p=0.0045; 30 min Ctrl vs KCl: p=0.0130; paired t-tests, Figure 8E). Interestingly, within minutes of the KCl application to the axon hillock, we observed the partial disappearance of the proximal onset of an individual AIS (Figure 8E, insets). This was quite striking since the majority of previous studies on this topic could only access AIS length and position changes in a post-hoc manner and never in the same cell. Using patch-clamp recordings, we verified that neurons tolerate and survive the treatment in the respective time frame. We here demonstrate that actual structural AIS changes do occur rapidly within 10 min and that they can be visualized and monitored within the same neuron.

As outlined above, previous studies investigating AIS plasticity relied on post-hoc analysis in fixed sections. We conducted the ultimate test of the ank-G-GFP reporter using live cell imaging in intact animals in a longitudinal approach.

Three different in vivo systems were applied to test the reliability, reproducibility, and stability of intrinsic ank-G-GFP signals in vivo. First, we imaged and tracked the same pyramidal neuron AIS in S1 for 1 wk (Figure 9A–C). After implanting cranial windows in 3 mo-old ank-G-GFP × CaMKII-Cre animals, the AIS of layer II/III neurons in S1 were imaged based on their intrinsic GFP signal, which proved to be robust and easy to visualize (Figure 9A–C). We compared the same region of interest at a 7 d interval and analyzed the length of the same individual AIS (Figure 9B and C). Quantification revealed that the same AIS maintained their average length reliably (week 1, magenta: 17.3±3.9 µm vs week 2, green: 16.8±3.9 µm, paired t-test, n=30 AIS; Figure 9B). This indicated that neither previous imaging nor potential physiological alterations in a 1 wk time frame changed GFP localization in such a way that it caused global length differences. Plotting of mean AIS length from week 1 and week 2 indicated a strong and significant correlation among individual AIS lengths measured in the same ROI (r²=0.7892, p<0.0001, n=30 AIS; Figure 9B).

In addition to constitutive expression of ank-G-GFP in intercrossed transgenic mice, AIS could also be monitored in vivo using a viral approach in two additional systems: (1) specific subpopulations of cross-callosal pyramidal neurons of the caudal forelimb region of M1 (Figure 9D) and (2) principal neurons of deep brain areas, like the basolateral amygdala (BLA) via gradient reflective index (GRIN) lenses (Figure 9E).

Our live imaging data provides evidence that the ank-G-GFP model is a powerful tool to study excitable membrane domains in living animals in future studies of neural development, animal behavior, and disease progression.

We used the Flip-excision (FLEx) system (Schnütgen et al., 2003) to introduce a conditional GFP-tag into the final exon of ankyrin-G, the major scaffolding protein of the AIS (Figure 1A; Leterrier, 2018). A FLEx cassette containing the last exon of Ank3 (ENSMUSE00001313132, exon 42) followed by a copy of the last exon of Ank3 fused with the coding sequence for eGFP in the reverse orientation was generated using standard molecular methods. The entire cassette was flanked by alternating loxP and lox2272 sites. In the presence of Cre recombinase, the cassette orientation is irreversibly flipped by sequential recombination and excision of two of the lox sites (Figure 1B). The result is the expression of an ankyrin-G-GFP fusion product from the endogenous locus only in cells expressing Cre recombinase (Figure 1B). The linker sequence between Ank3 and GFP is the same sequence used in previously described 190-, 270-, and 480-kDa ankyrin-G-GFP plasmids (Kole and Stuart, 2012; Jenkins et al., 2015; Kizhatil and Bennett, 2004; Zhang and Bennett, 1998). Importantly, these ankyrin-G-GFP fusion products have been able to rescue all functions of ankyrin-G tested to date, suggesting that the addition of GFP to the ankyrin-G carboxy-terminus does not affect ankyrin function (Jenkins et al., 2015; Tseng et al., 2015; Kizhatil and Bennett, 2004; Zhang and Bennett, 1998; He et al., 2014; He et al., 2012; Jenkins et al., 2013). A neomycin resistance cassette, driven by the phosphoglycerate promoter (PGK-neo) and flanked by flippase recognition target (FRT) sites, was inserted between exon 42 and the 5’ LoxP site. The linearized construct was introduced into 129X1/SvJ ES cells by electroporation, and G418-resistant clones were screened using polymerase chain reaction (Figure 1C). One of the primers was anchored outside of the 5’ homology arm to allow the detection of the FLEx cassette within the correct chromosomal location. ES cells selected for the presence of the FLEx cassette using G418 were injected into C57BL/6NHsd blastocysts. A high percentage of chimeric animals were obtained and bred to C57BL/6NHsd mice to produce heterozygous animals. Mutant mice were backcrossed to C57BL6/J mice (Jackson Laboratory) and were compared to C57BL6/J mice as WT controls. These mice have been deposited at the Jackson Laboratory (B6;129S6-Ank3tm1Pmj/J, JAX strain # 038118).

To confirm correct genotypes, ear punch biopsies were derived from adult breeding animals and tail cuts from all pups used for OTC preparation, using standard PCR procedures. Tissue samples were digested (25 mM NaOH, 0.2 disodium EDTA), centrifuged for 1 min at 14,000 rpm, and incubated at 95° C. After addition of 40 mM Tris-HCl and subsequent centrifugation at 14,000 rpm for 2 min, DNA was amplified from gDNA using the forward primer JDH71 AG-GFP-F 5’CTACAACCAATGGGGATCGTTAAC’3 and reverse primer JDH72 AG-GFP-R 5’TTAGGAAGGAGAAATGGGTGAGAG3’.