Adult Camk2a gene reinstatement restores the learning and plasticity deficits of Camk2a knockout mice

Summary With the recent findings that mutations in the gene encoding the α-subunit of calcium/calmodulin-dependent protein kinase II (CAMK2A) causes a neurodevelopmental disorder (NDD), it is of great therapeutic relevance to know if there exists a critical developmental time window in which CAMK2A needs to be expressed for normal brain development, or whether expression of the protein at later stages is still beneficial to restore normal functioning. To answer this question, we generated an inducible Camk2a mouse model, which allows us to express CAMK2A at any desired time. Here, we show that adult expression of CAMK2A rescues the behavioral and electrophysiological phenotypes seen in the Camk2a knock-out mice, including spatial and conditional learning and synaptic plasticity. These results suggest that CAMK2A does not play a critical irreversible role in neurodevelopment, which is of importance for future therapies to treat CAMK2A-dependent disorders.


INTRODUCTION
Intellectual disability (ID), a condition defined by an IQ below 70, is a predominant feature of neurodevelopmental disorders (NDDs) affecting approximately 1% of the global world population (Maulik et al., 2011;McKenzie et al., 2016). Because of next-generation sequencing, many genes have now been implicated in this clinical condition (Gilissen et al., 2014;Vissers et al., 2016). One of the genes recently implicated in ID is CAMK2A, which encodes the alpha-subunit of calcium/calmodulin-dependent protein kinase II (CAMK2A), a highly abundant kinase in the brain that has been shown to play a critical role in hippocampus-dependent synaptic plasticity, learning, and memory in mice (Silva et al., 1992a(Silva et al., , 1992b. In humans, a similar role for CAMK2A has been suggested, as all individuals carrying mutations in CAMK2A suffer from a NDD with mild to severe ID, and some from autism spectrum disorder (Akita et al., 2018;Chia et al., 2018;Kü ry et al., 2017).
With the diagnosis for the CAMK2A-dependent NDD established, doors open toward identifying a potential therapy. However, the important question that emerges is whether the genetic mutation causes alterations during early neurodevelopment that might be irrevocable, and thus dictate a critical window for therapeutic intervention (Krol and Feng, 2018;Meredith et al., 2012). The current literature shows that the answer to this question depends strongly on the gene involved in the NDD and the behavior tested (Aceti et al., 2015;Clement et al., 2012;Guy et al., 2007;Lang et al., 2014;Mei et al., 2016;Mielnik et al., 2020;Silva-Santos et al., 2015). For example, in a mouse model for Rett syndrome, reinstatement of the gene Mecp2 in adulthood improves all phenotypic behavior, but seldom back to wild type performance (Guy et al., 2007;Lang et al., 2014). However, in other models for NDDs, the ability to rescue phenotypes was also shown to depend on the behavior tested, as exemplified by Grin1, Shank3, Ube3a, and Syngap1 (Aceti et al., 2015;Clement et al., 2012;Mei et al., 2016;Mielnik et al., 2020;Silva-Santos et al., 2015). Social and cognitive deficits are largely, but not completely rescued upon adult reinstatement of Grin1 (Mielnik et al., 2020) and fully rescued upon Shank3 reinstatement (Mei et al., 2016), but motor and anxiety deficits are only partially rescued upon Grin1 reinstatement (Mielnik et al., 2020) and not rescued upon Shank3 reinstatement (Mei et al., 2016). In a mouse model for Angelman Syndrome, adult reinstatement of Ube3a does not rescue behavioral phenotypes found in the Ube3a mouse (Silva-Santos et al., 2015). However, reinstatement of Ube3a at postnatal day 21 (P21) rescues the motor skill deficits on the rotarod test, but not deficits iScience Article in anxiety-related or intrinsic behavior, suggesting an early critical window for these forms of behavior (Silva-Santos et al., 2015). In a Syngap1 haploinsufficiency model, cognitive behavior cannot be rescued upon adult reinstatement, whereas motor and anxiety-related behavior can only be rescued upon very young reinstatement, at P1 but not P21 (Aceti et al., 2015;Clement et al., 2012). Interestingly, pharmacological treatment with a GSK3-b inhibitor of the Syngap1 mouse model during the critical period of P10-16 does rescue social, cognitive, and anxiety-related behavior, and partially motor behavior (Verma et al., 2021). Finally, the only phenotype that could be rescued upon adult reinstatement in all NDD mouse models discussed here is the electrophysiological plasticity (Guy et al., 2007;Mei et al., 2016;Ozkan et al., 2014;Rotaru et al., 2018;Silva-Santos et al., 2015).
CAMK2A starts to be expressed around P1, which means that it is not required for prenatal neurodevelopment (Bayer et al., 1999). Whereas in the adult brain CAMK2A has been shown to play a critical role in normal brain functioning, as adult deletion of Camk2a is equally detrimental for learning and plasticity as germline deletion (Achterberg et al., 2014), not much is known about a possible critical period for CAMK2A during the early postnatal development and whether the phenotypes seen in adult mice could potentially be rescued upon CAMK2A expression.
To assess whether there is a role for CAMK2A in postnatal neurodevelopment and if a critical period for CAMK2A expression exists, we generated an inducible Camk2a mouse model, in which we can reinstate Camk2a at the time of our choosing. We show that adult reinstatement of Camk2a rescues all behavioral and electrophysiological phenotypes seen in the Camk2a knockout mice. These results indicate that absence of CAMK2A during development does not lead to irreversible alterations in brain development, and that potential therapies for CAMK2A-related disorders do not require specific early time windows.

Generation of an inducible Camk2a mouse
To study the role of CAMK2A during neurodevelopment, we generated a novel mouse model. A transgenic cassette containing Camk2a exon 2 fused to the tdTomato gene with a transcriptional stop at the end, flanked with loxP sites, was inserted between exon 1 and 2 of the endogenous Camk2a gene (referred to as CAMK2 Lox -Stop -Lox (CAMK2 LSL )), allowing for temporally controlled re-expression of Camk2a upon Cre-mediated deletion of the transgenic cassette. Heterozygous CAMK2 LSL mice were then crossed with the transgenic tamoxifen-inducible CAG-Cre ESR line, to obtain both the CAMK2 LSL and the inducible CAMK2A LSL ;CAG-Cre ESR mouse line ( Figure 1A). As control, wild-type littermates negative or positive for Cre-expression were taken along (CAMK2A WT and CAMK2A WT ;CAG-Cre ESR , respectively).
In the absence of Cre-driven recombination, CAMK2A expression in adult CAMK2A LSL mice is abolished and replaced by tdTomato expression driven by the endogenous CAMK2 promotor ( Figure 1B). Administration of tamoxifen (0.1 mg/mL, four consecutive days) to adult mice (>8 weeks old), successfully deleted the tdTomato transgene and induced CAMK2A re-expression throughout the brain ( Figure 1B). To get an insight into the level of CAMK2A protein expression upon reinstatement, a western blot was performed on

OPEN ACCESS
hippocampal and cortical tissue of the mice 3-4 weeks after gene reinstatement. Western blot analysis revealed 89% and 83% of CAMK2A expression upon reinstatement in the hippocampus and cortex, respectively, of CAMK2A LSL ;CAG-Cre ESR mice, with low levels of tdTomato still present ( Figure 1C; all statistics are given in the figure legends). Full blots are displayed in Figure S1.
CAMK2A is known to form heteromeric holoenzyme complexes with CAMK2B consisting of 12-14 subunits (Bhattacharyya et al., 2016;Shen et al., 1998). As we started to express CAMK2A here in the adult brain, having the mice to develop with only CAMK2B present, we tested whether adult expressed CAMK2A forms heteromeric holoenzymes again with CAMK2B. Four weeks after tamoxifen injections, we immunoprecipitated CAMK2A from cortex tissue and immunoblotted for both isoforms. CAMK2B was successfully pulled down in CAMK2A WT and CAMK2A LSL ;CAG-Cre ESR mice, but not in CAMK2A LSL , indicating correct formation of heteromeric holoenzymes consisting of CAMK2A and CAMK2B subunits, upon Camk2a reinstatement ( Figures 1D and S1). Note that the input showed similar CAMK2B expression in all groups, confirming previous findings (Elgersma et al., 2002) that loss of CAMK2A does not result in upregulation of CAMK2B.

Adult Camk2a reinstatement rescues learning and plasticity phenotypes
Camk2a knockout mice are known to have severe impairments in spatial as well as associative learning (Chen et al., 1994;Elgersma et al., 2002;Silva et al., 1992a). To confirm that our CAMK2A LSL mice behave as a true knockout, and whether adult gene reinstatement can rescue these phenotypes, 12-16-week-old mice were injected with tamoxifen and, after 5 weeks, spatial and associative learning was assessed using the Morris water maze and contextual and cued fear conditioning ( Figure 2A).
Mice were trained to find the hidden platform of the Morris water maze for 5 days. Although latencies to find the platform decreased (with some fluctuations) over time for all genotypes, the overall reduction in latency to find the platform was less in CAMK2A LSL mice compared with CAMK2A WT , suggesting potentially impaired procedural learning ( Figure 2B). Latencies to find the platform are not very reliable to assess spatial learning, as just missing the platform could immediately lead to longer latencies. To better test whether the mice had used a spatial learning strategy to learn the platform position, we performed a probe trial, removing the platform from the water bath. CAMK2A WT and CAMK2A WT ;CAG-Cre ESR mice showed clear preferences to spend time in the target quadrant, proving they successfully learned the location of the platform (Figures 2C and 2D). In contrast, CAMK2A LSL mice did not show any preference for the target quadrant confirming impaired spatial learning. Mice in which the Camk2a gene was reinstated during adulthood (CAMK2A LSL ;CAG-Cre ESR ) showed clear preference for the target quadrant, equivalent to CAMK2A WT mice, showing that spatial learning was normal upon adult Camk2a gene reinstatement ( Figure 2D). Importantly, the swim speed was similar between genotypes ( Figure 2E).
For the fear conditioning paradigm, on day 1, mice were allowed to explore the fear conditioning box for 150 s after which they were exposed for 20 s to a tone, which ended simultaneously with a 2 s mild foot shock. After 24 h, on day 2, mice were placed back in the same box and contextual memory was assessed iScience Article by measuring the amount of time the mice showed freezing behavior ( Figure 2F). Confirming the previously published impaired associative learning (Achterberg et al., 2014;Chen et al., 1994), CAMK2A LSL mice spent significantly less time freezing compared with the other genotypes ( Figure 2G). On day 3, mice were placed in a different context and were presented with the tone to test for amygdala-dependent cued learning. Again, CAMK2A LSL mice showed impaired associative learning, spending significantly less time freezing compared with the other genotypes. Similarly as in the Morris water maze experiment, adult expression of CAMK2A induced a full rescue of the associative learning phenotype, as CAMK2A LSL ;CAG-Cre ESR mice showed equal freezing time compared with CAMK2A WT mice both in contextual and cued fear conditioning ( Figure 2G).
It is generally accepted that synaptic plasticity, i.e. the strengthening of synapses, underlies learning and memory (Shaw, 1986). Indeed, long-term potentiation (LTP) at the Schaffer collateral-CA1 synapse in the hippocampus is known to be impaired in Camk2a knockout mice (Elgersma et al., 2002;Hinds et al., 1998;Silva et al., 1992b). With all the cognitive phenotypes being rescued upon adult Camk2a gene reinstatement, we hypothesized that also synaptic plasticity would be rescued upon adult reinstatement. We first measured basal synaptic transmission at the Schaffer collateral-CA1 synapse, where the CAMK2A LSL showed similar strength compared with CAMK2A WT (Figures 3A and 3B). As we found a small Cre effect in the fEPSP in CAMK2A WT mice ( Figure 3B), we analyzed the slopes of the input/output curve, to assess the strength of basal transmission, which was similar between all groups ( Figure 3B inset). Also paired pulse facilitation (PPF) was found to be similar between all groups ( Figure 3C). Upon the induction of LTP, we found, consistent with previous findings (Elgersma et al., 2002;Silva et al., 1992b), a significant impairment in hippocampal LTP in the CAMK2A LSL mice compared with the other genotypes. Moreover, confirming our hypothesis, LTP was completely normalized in the CAMK2A LSL ;CAG-Cre ESR mice, proving a rescue also of synaptic plasticity upon adult Camk2a gene reinstatement ( Figures 3D  and 3E).

Adult Camk2a reinstatement rescues intrinsic behavioral phenotypes
Most of the phenotypes tested above depend on hippocampal plasticity, which appears to not have a critical developmental window, but remains important throughout life (Guy et al., 2007;Ozkan et al., 2014;Silva-Santos et al., 2015). Hence, we also wanted to assess the role of CAMK2A in more intrinsic behavior, which has been shown to have an early critical developmental time window during which alterations can cause irreversible damage (Silva-Santos et al., 2015). For this purpose, we performed a behavioral battery of different tests, previously shown to be sensitive to a critical treatment window to obtain full reversal (Silva-Santos et al., 2015;Sonzogni et al., 2018Sonzogni et al., , 2019Sonzogni et al., , 2020. Similar to the previous experiments, gene reinstatement in this cohort was induced in adult 11-19-week-old mice, and after 5 weeks, a behavioral battery was performed. No behavioral phenotype was found in CAMK2A LSL compared with CAMK2A WT mice in the open field, rotarod, or marble burying test ( Figure S2). However, in both the nest building and forced swim test, we found CAMK2A LSL mice to show significantly altered behavior compared with CAMK2A WT mice. In the nest building paradigm, CAMK2A LSL mice used less material to build their nest compared with the other genotypes and in the forced swim test, CAMK2A LSL mice showed decreased immobility compared with CAMK2A WT mice ( Figures 4A and 4B). Surprisingly, both phenotypes were rescued in the CAMK2A LSL ;CAG-Cre ESR mice as they showed similar behavior compared with CAMK2A WT mice, suggesting that although the absence of CAMK2A during early developmental periods affects these intrinsic behaviors, adult CAMK2A expression is beneficial to rescue these behaviors ( Figures 4A and 4B). This could indicate that the impaired circuits that underlie the nest building and forced swim test deficits in Camk2a mice are different from the circuits that underlie these deficits in Ube3a mice (Silva-Santos et al., 2015).

DISCUSSION
With the current advance in genetic diagnosis, causing a quickly expanding list of rare NDDs and the upcoming possibility to treat these disorders with different forms of gene therapy, such as antisense oligonucleotides, the need to understand the optimal timing for treatment becomes increasingly important. CAMK2-related disorder is one of these recently discovered disorders, and our aim in this study was to understand whether a developmental critical time window exists for CAMK2A to be expressed for normal brain development and function. For this purpose, we generated an inducible CAMK2A knockout model, where we could restore CAMK2A expression in adulthood and assessed the behavior. iScience Article imply that loss of CAMK2A during development does not cause irretrievable distortion to neural circuits in the brain, as we fully rescue the behavioral and electrophysiological phenotypes assessed here.
To our knowledge, this is the only NDD mouse model described in literature, where all phenotypes can be fully normalized upon adult gene reinstatement. In a subset of the NDD mouse models, adult gene reinstatement rescues some but not all behavioral deficits (e.g. upon reinstatement of Shank3 (Mei et al., 2016)), rescues observed phenotypes partially (e.g. upon reinstatement of Mecp2 or Grin1 (Lang et al., 2014;Mielnik et al., 2020)), or only rescues the behavioral deficits if reinstated at a juvenile age (e.g. upon reinstatement of Ube3a or Syngap1 (Aceti et al., 2015;Silva-Santos et al., 2015)). The reverse is also true, as only adult deletion of Camk2a completely phenocopies the deficits observed in germline knockout mice (Achterberg et al., 2014), whereas adult deletion in other NDD models do either not or only partially copy the observed behavioral abnormalities (e.g. upon Syngap1, Mecp2, or Ube3a deletion; Gemelli et al., 2006;Ozkan et al., 2014;Sonzogni et al., 2019), or only show the phenotypes upon juvenile deletion (e.g. upon Ube3a deletion; Sonzogni et al., 2019), proving a critical window for protein expression of this gene. The results presented here suggests that CAMK2A does not play a role during development. However, it was recently shown that deletion of Camk2a and Camk2b simultaneously from germline, is lethal at P1, the moment that CAMK2A starts to be expressed (Kool et al., 2019). With the Camk2a and Camk2b single knockouts being completely viable, this finding does suggest an important early postnatal developmental role also for CAMK2A, but one that can be compensated for by CAMK2B. Our results now imply that this compensation is of crucial importance for early postnatal brain development, preventing irreversible alterations in the neuronal circuit and therefore allowing for full phenotypic recovery in the adult mice.
To conclude, we show that CAMK2A expression is required during learning or other types of behavior, independent of its expression throughout development. This could be of crucial importance for patients that carry CAMK2A loss-of-function mutations (Chia et al., 2018;Kü ry et al., 2017) and opens doors toward the potential use of therapies that increase CAMK2A expression.

Limitations of the study
In this study, we generated a novel mouse model, allowing us to assess the effect of expressing CAMK2A in adult mutant mice that lack CAMK2A throughout their life. We found that adult expression of CAMK2A iScience Article completely normalizes behavior and plasticity in mice. However, we can only draw conclusions on the assays performed in this study. Previously, several additional phenotypes have been reported, such as the cerebellum-dependent vestibular and ocular reflex (VOR and OKR) learning, shown to be affected in CAMK2A knockout mice (Hansel et al., 2006). Similarly, we have only measured plasticity at the CA3-CA1 synapse in the hippocampus. Whether plasticity and other electrophysiological measures shown to be affected in CAMK2A knockout mice (e.g. Coultrap et al., 2014;Hansel et al., 2006;Hardingham et al., 2008) are also normalized, remains to be determined. Finally, we have to be careful in directly translating these findings to the clinic as in our mouse model, CAMK2A is completely absent during development. Most patients, however, express a mutant form of CAMK2A (Akita et al., 2018;Kü ry et al., 2017).

STAR+METHODS
Detailed methods are provided in the online version of this paper and include the following: iScience Article Adult (>8 weeks) CAMK2A LSL , CAMK2A LSL ;CAG-Cre ESR , CAMK2A WT and CAMK2A WT ;CAG-Cre ESR mice of both sexes were used for the experiments. During the experiments, the experimenter was blind for the genotypes. All mice were group housed at 22 G 2 C, except for the nest building test when they were single caged. They were on a 12/12h light/dark cycle with food and water available ad libitum. All experiments were performed during the light phase. All animal experiments were conducted in accordance with the European Commission Council Directive 2010/63/EU (CCD project license AVD101002017893), and all described experiments and protocols were subjected to ethical review (and approved) by an independent review board (IRB) of the Erasmus MC.
All data reported in this paper will be shared by the lead contact upon request.

Generation of mouse line
To generate the CAMK2A LSL mouse model, the construct used to generate the Camk2a floxed mouse line was used as a starting point (Achterberg et al., 2014). From this construct the 5 0 flanking arm of exon 2 and the 3 0 flanking arm containing exon 2 of Camk2a were obtained and cloned in a vector containing the cassette of Camk2a-exon2 fused to tdTomato, followed by the woodchuck hepatitis virus post-transcriptional regulatory element (WPRE), flanked by LoxP sites, and a neomycin cassette for positive selection, flanked by Frt sites. All exonic sequences were sequenced to verify that no mutations were introduced accidentally. The targeting construct was linearized and electroporated into E14 ES cells (derived from 129P2 mice). Cells were cultured in BRL cell-conditioned medium in the presence of leukemia inhibitory factor. After selection with G418 (200 mg/mL), targeted clones were identified by PCR (long-range PCR from neomycin resistance gene to the region flanking the targeted sequence). A clone with normal karyotype was injected into blastocysts of C57BL/6J mice. Male chimeras were crossed with female C57BL/6J mice and resulting offspring was further backcrossed into the C57BL/6J background. To obtain the mice used for experiments, heterozygous CAMK2 LSL females (15 times backcrossed in C57BL/6J) were crossed with Tg(CAG-cre/Esr1*)5Amc (CAG-Cre ESR ) male mice (MGI:2182767) (17 times backcrossed in 129S2/SvPasCrl). Of the F1 offspring, the heterozygous CAMK2A LSL ;CAG-cre ESR were crossed with heterozygous CAMK2A LSL mice, obtaining the resulting experimental group: CAMK2A LSL , CAMK2A LSL ;CAG-Cre ESR , CAMK2A WT and CAMK2A WT ;CAG-Cre ESR . Adult (>8 weeks) mice of both sexes were used for the experiments. During the experiments, the experimenter was blind for the genotypes. All mice were group housed at 22 G 2 C, except for the nest building test when they were single caged. They were on a 12/12h light/dark cycle with food and water available ad libitum. All experiments were performed during the light phase. All animal experiments were conducted in accordance with the European Commission Council Directive 2010/63/EU (CCD project license AVD101002017893), and all described experiments and protocols were subjected to ethical review (and approved) by an independent review board (IRB) of the Erasmus MC.

Tamoxifen treatment
To induce Cre-mediated deletion, mice received intraperitoneal injections of 0.1 mg Tamoxifen per gram body weight for four consecutive days at 11-19 weeks old. Tamoxifen was freshly dissolved at a concentration of 20 mg/mL in sunflower oil beforehand.