A small-molecule TrkB ligand improves dendritic spine phenotypes and atypical behaviors in female Rett syndrome mice

Rett syndrome (RTT) is an X chromosome-linked neurodevelopmental disorder associated with intellectual disability and autism that affects approximately 1:10,000 females worldwide (Katz et al., 2012; Neul and Zoghbi, 2004). The majority of individuals with RTT carry loss-of-function mutations in the gene that encodes methyl-CpG-binding protein 2 (MeCP2), a transcriptional regulator that binds to methylated DNA sites and recruits transcriptional repressors and modulates chromatin compaction and its accessibility (Amir et al., 1999; Chahrour et al., 2008; Lyst and Bird, 2015; Nan et al., 1997; Percy and Lane, 2005).

The transcription, synthesis, intracellular transport and activity-dependent release of brain-derived neurotrophic factor (BDNF) are all impaired in a number of neurological disorders, including RTT and Huntington's disease (Gines et al., 2010; Hartmann et al., 2012; Li and Pozzo-Miller, 2014; Tapia-Arancibia et al., 2008). The relevance of BDNF deficiency to RTT pathogenesis is supported by the observations that Bdnf expression is directly regulated by MeCP2 in an activity-dependent manner (Chen et al., 2003; Martinowich et al., 2003; Zhou et al., 2006), BDNF levels are lower in multiple brain areas of Mecp2-deficient mice (Abuhatzira et al., 2007; Li et al., 2012; Schmid et al., 2012; Wang et al., 2006), and increasing BDNF levels via genetic or pharmacological manipulations improve some of the deficits observed in Mecp2-deficient neurons and mice (Chang et al., 2006; Chapleau et al., 2009; Ogier et al., 2007). BDNF signaling via its receptor, tropomyosin-related kinase B (TrkB, encoded by Ntrk2), plays a key role in neuronal and synaptic development, as well as in adult synaptic plasticity. BDNF increases the density of dendritic spines of CA1 pyramidal neurons in organotypic cultures of hippocampal slices through TrkB activation of the ERK1/2 pathway (Alonso et al., 2004; Tyler and Pozzo-Miller, 2001). Moreover, hippocampal pyramidal neurons of Mecp2 knockout (KO) mice have lower dendritic spine density (Chapleau et al., 2012), along with impaired BDNF-induced membrane currents and Ca²⁺ signals mediated by TRPC3 channels (Li et al., 2012), as well as reduced dendritic trafficking and activity-dependent release of BDNF-GFP (Xu et al., 2014). Although these studies indicate that BDNF deficiency is a key component in RTT pathogenesis (Katz, 2014; Li and Pozzo-Miller, 2014), the therapeutic potential of BDNF is limited by its low blood-brain barrier permeability and short plasma half-life (Poduslo and Curran, 1996).

An alternative to BDNF itself is to use synthetic small molecules that target the TrkB receptor as ligands. An established preclinical candidate is LM22A-4, a ‘mimetic’ of the BDNF loop-2 domain that activates TrkB and its downstream signaling pathways (Massa et al., 2010). Indeed, LM22A-4 improved disease phenotypes in mouse models of Huntington's disease (Simmons et al., 2013), Dravet's disease (Gu et al., 2022) and chemotherapy-induced cognitive decline (Geraghty et al., 2019). In female Mecp2 heterozygous (HET) mice, a 2-month treatment with LM22A-4 improved breathing irregularities (Kron et al., 2012; Schmid et al., 2012), reduced network hyperactivity in hippocampal slices, restored long-term potentiation (LTP) of excitatory synaptic transmission, and improved object location memory (Li et al., 2017). Similarly, a second-generation TrkB ligand based on LM22A-4 also improved breathing patterns and motor deficits in Mecp2 HET mice (Adams et al., 2020). In the present study, we tested the effects of LM22A-4 on dendritic spine density and size in hippocampal pyramidal neurons of female Mecp2-deficient mice, as well as in a machine-learning unbiased screen of open-field behaviors in female Mecp2 HET mice interacting with unfamiliar and familiar mice.

We first confirmed that enhanced yellow fluorescent protein (eYFP)-expressing hippocampal CA1 pyramidal neurons in organotypic slice cultures from postnatal day (P) 7 male Mecp2 KO mice had lower spine density than that of neurons in cultured slices from age-matched male WT mice (P=0.0032, unpaired two-tailed t-test; n=17 Mecp2 KO neurons/10 slices versus n=14 WT neurons/9 slices; Fig. 1A,B), as we reported previously (Chapleau et al., 2012). In addition, the volume of individual spines was larger in CA1 pyramidal neurons in slice cultures from Mecp2 KO mice [P≤0.0001, Kolmogorov–Smirnov (K-S) test; n=1458 Mecp2 KO spines/17 neurons/10 slices versus n=2139 WT spines/14 neurons/9 slices; Fig. 1C), as we described previously in dissociated cultures of hippocampal neurons from P1 male Mecp2 KO mice (Xu et al., 2017) and in CA1 pyramidal neurons in ex vivo slices from symptomatic P45-P65 male Mecp2 KO mice (Li et al., 2016).

Consistent with multiple reports in different experimental preparations, including ours in hippocampal slice cultures from neonatal rats (Alonso et al., 2004; Chapleau et al., 2012, 2008; Tyler and Pozzo-Miller, 2001), BDNF treatment (250 ng/ml, 48 h) increased spine density in CA1 pyramidal neurons in slice cultures from both male Mecp2 KO mice [P=0.0026, one-way ANOVA with Bonferroni's post hoc test (hereafter ANOVA-Bonferroni's); n=17 control Mecp2 KO neurons/10 slices versus n=13 BDNF Mecp2 KO neurons/5 slices] and WT littermates (P=0.0042, ANOVA-Bonferroni's; n=14 control WT neurons/9 slices versus n=15 BDNF WT neurons/5 slices; Fig. 1D). Also in line with prior reports, these effects of BDNF were reduced by treatment with the non-selective Trk receptor inhibitor K-252a (200 nM) in WT slice cultures (P=0.0088, ANOVA-Bonferroni's; n=13 BDNF+K-252a WT neurons/6 slices versus n=15 BDNF WT neurons/5 slices; Fig. 1D). Unexpectedly, K-252a treatment did not alter the effect of BDNF in Mecp2 KO neurons (P>0.9999, ANOVA-Bonferroni's; n=15 BDNF+K-252a Mecp2 KO neurons/6 slices versus n=13 BDNF Mecp2 KO neurons/5 slices) and in fact increased spine density by itself (P<0.0001, ANOVA-Bonferroni's; n=7 K-252a Mecp2 KO neurons/4 slices versus n=17 control Mecp2 KO neurons/10 slices; Fig. 1D), which may reflect an altered signaling balance between the low-affinity p75 receptor (encoded by Ngfr) and TrkB (Chapleau and Pozzo-Miller, 2012) in the absence of MeCP2.

As expected from its reported activation of TrkB receptors as a partial agonist (Massa et al., 2010) and the lower levels of BDNF in Mecp2 mutant mice compared to those in WT mice (Katz, 2014; Li and Pozzo-Miller, 2014), LM22A-4 treatment (500 nM, 48 h) increased spine density in pyramidal neurons in areas CA1 and CA3 of hippocampal slice cultures from male Mecp2 KO mice (P<0.0001, ANOVA-Bonferroni's; n=17 control Mecp2 KO neurons/10 slices versus n=10 LM22A-4 Mecp2 KO neurons/7 slices; Fig. 1E). Because of its partial agonism at TrkB receptors (Massa et al., 2010), LM22A-4 had no effect on spine density in pyramidal neurons from male WT mice (P>0.9999, ANOVA-Bonferroni's; n=12 control WT neurons/11 slices versus n=13 LM22A-4 WT neurons/9 slices), where typical BDNF levels outcompete LM22A-4 for TrkB binding. Similar to the effect of BDNF described above, K-252a and the selective TrkB inhibitor ANA-12 (Cazorla et al., 2011) both reduced the effect of LM22A-4 on spine density in Mecp2 KO neurons (P=0.0011, ANOVA-Bonferroni's; n=10 LM22A-4+ANA-12 Mecp2 KO neurons/7 slices versus n=10 LM22A-4 Mecp2 KO neurons/7 slices; Fig. 1E). Similar to the spinogenic effect of K-252a, ANA-12 also increased spine density by itself in Mecp2 KO neurons (P<0.0001, ANOVA-Bonferroni's; n=11 ANA-12 Mecp2 KO neurons/5 slices versus n=10 LM22A-4 Mecp2 KO neurons/7 slices; Fig. 1E).

To identify neurons of known genotypes in the ‘mosaic’ brain of female Mecp2 HET mice (due to X-chromosome inactivation), we crossed female Mecp2 HET mice with male mice expressing GFP-tagged MeCP2 (Schmid et al., 2012). For quantitative analyses of dendritic morphology, CA1 pyramidal neurons in ex vivo hippocampal slices from young adult mice were identified as MeCP2-expressing or MeCP2-lacking based on the nuclear presence of GFP, and were filled with biocytin through a whole-cell patch pipette, followed by confocal microscopy of Alexa Fluor 488-tagged streptavidin (Fig. 2A).

Mirroring the spine phenotype of symptomatic male Mecp2 KO mice (P45-P65), in which their density is similar to that in age-matched male WT mice (Li et al., 2016), spine density was similar in the two cellular genotypes in 6-month-old female MeCP2-GFP HET mice, and comparable to that in age-matched female WT mice (P>0.9999, ANOVA-Bonferroni's; n=35 MeCP2-expressing HET neurons/18 slices, n=24 MeCP2-lacking HET neurons/14 slices, n=9 WT neurons/6 slices; n=4-5 mice, with half receiving LM22A-4; Fig. S1). Unexpectedly, MeCP2-expressing CA1 pyramidal neurons in female MeCP2-GFP HET mice had dendritic spines with larger volumes than those of neighboring MeCP2-lacking neurons and CA1 pyramidal neurons in slices from age-matched female Mecp2 HET mice (P<0.0001, K-S test; n=521 spines/35 MeCP2-expressing neurons/18 slices in HET, n=257 spines/24 MeCP2-lacking neurons/14 slices in HET, n=145 spines/9 neurons/6 slices in WT; Fig. 2B). This observation suggests that dendritic spines of MeCP2-expressing neurons in the mosaic brain of female MeCP2-GFP HET mice retain their sensitivity to neuronal activity (Segal et al., 2000) under the conditions of heightened hippocampal activity observed in female Mecp2 HET mice (Li et al., 2017), which is comparable to that in the hippocampus of male Mecp2 KO mice (Calfa et al., 2011, 2015; Li et al., 2016).

Four- to six-month-old female MeCP2-GFP HET mice and their age-matched WT littermates were treated with either LM22A-4 (50 mg/kg in saline, twice daily) or vehicle (saline) for 4 weeks, as described previously (Kron et al., 2012; Li et al., 2017; Schmid et al., 2012). Consistent with the phenotype described above, LM22A-4 did not affect spine density (P>0.9999, ANOVA-Bonferroni's, n=48 MeCP2-expressing neurons/24 slices in LM22A-4 HET versus n=28 MeCP2-lacking neurons/18 slices in LM22A-4 HET; versus P=0.0423, K-S test, n=11 LM22A-4 WT neurons/9 slices; n=4-5 mice, with half receiving LM22A-4; Fig. S1). However, LM22A-4 reduced spine volume only in MeCP2-expressing pyramidal neurons of female MeCP2-GFP HET mice, reaching values comparable to those in MeCP2-lacking neurons in control HET mice (P>0.9999, K-S test; n=505 spines/48 MeCP2-expressing neurons/24 slices in LM22A-4 HET versus n=257 spines/24 MeCP2-lacking neurons/14 slices in control HET) as well as those in female control WT mice (P>0.9999, K-S test; n=505 spines/48 MeCP2-expressing neurons/24 slices in LM22A-4 HET versus n=145 spines/9 neurons/6 slices in WT; Fig. 2C-G).

Similar to its actions in vitro, LM22A-4 had no effect on spine volume in neurons of female WT mice (P>0.9999, K-S test; n=134 spines/11 neurons/9 slices in LM22A-4 WT versus n=145 spines/9 neurons/6 slices in control WT). Spine volume was similar in the neurons of control WT mice and LM22A-4-treated WT mice, and MeCP2-lacking neurons in LM22A-4-treated HET mice (P>0.9999, K-S test; n=145 spines/9 neurons/6 slices in control WT, n=134 spines/11 neurons/9 slices in LM22A-4 WT, n=303 spines/28 MeCP2-lacking neurons/18 slices in LM22A-4 HET; Fig. 2F). Spine volumes of MeCP2-expressing neurons in LM22A-4-treated HET mice were partially rescued to levels comparable to those of control WT and LM22A-4 WT (P<0.0001, K-S test; n=505 spines/48 MeCP2-expressing neurons/24 slices in LM22A-4 HET, n=134 spines/11 neurons/9 slices in LM22A-4 WT, n=145 spines/9 neurons/6 slices in control WT; Fig. 2G).

Four- and six-month-old female Mecp2 HET mice were used to test the effects of a 4-week treatment with LM22A-4 on behaviors relevant to RTT.

We used the standard three-chamber social interaction test (Moy et al., 2004) in 7-month-old mice, first for social preference (novel mouse inside an inverted pencil cup versus empty inverted pencil cup), and secondly for social memory [preference for prior novel mouse (familiar) mouse versus new novel (unfamiliar) mouse inside inverted pencil cups], as described previously (Phillips et al., 2019). Using the average interaction time from the whole 10-min session, we found no differences between the discrimination indices for sociability (P>0.9999, ANOVA-Bonferroni's; n=10 control WT mice versus n=10 control Mecp2 HET mice; Fig. 3A) and social memory (P>0.9999, ANOVA-Bonferroni's; control WT versus control Mecp2 HET mice; Fig. 3B). In line with the absence of differences between genotypes, a 4-week treatment with LM22A-4 did not affect any of these measures either in Mecp2 HET or in WT mice (P>0.9999, ANOVA-Bonferroni's; n=10 LM22A-4 Mecp2 HET mice, n=10 control Mecp2 HET mice, n=12 LM22A-4 WT mice, n=10 control WT mice; Fig. 3A,B).

Because female mice interact in shorter bouts with their preferred choice mainly at the start of the social preference test compared to males (Netser et al., 2017), the average interaction time from the whole 10-min session may not be the most accurate measure of their behavior. Thus, we compared their discrimination index on a minute-by-minute basis throughout the 10-min session. There were no differences in the sociability trial between female Mecp2 HET and WT mice (P>0.9999, ANOVA-Bonferroni's; control Mecp2 HET versus control WT; Fig. 3C). However, during the social memory trial, most female WT mice (10 of 10 mice) had a positive discrimination index within the first minute of the test (boxed data in Fig. 3D), suggesting that they could quickly discriminate between a familiar mouse and an unfamiliar mouse. On the contrary, most female Mecp2 HET mice (8 of 10 mice) took 5 min to reach a discrimination index similar to that of female WT mice (boxed data in Fig. 3D; P>0.9999, ANOVA-Bonferroni's; control Mecp2 HET versus control WT), indicating at least a delayed social memory. After those initial times, neither WT nor Mecp2 HET mice showed clear socially motivated behaviors (P>0.9999, ANOVA-Bonferroni's; control Mecp2 HET versus control WT; Fig. 3D). Furthermore, LM22A-4 had no effects on these dynamics of social interactions, neither during the sociability test nor on the social memory test (P>0.9999, ANOVA-Bonferroni's; among all groups; Fig. 3A-D). These observations are consistent with previous reports of female mice engaging differently during social interactions than males, which engage in longer interaction bouts (Karlsson et al., 2015; Li and Dulac, 2018; Netser et al., 2017; Rodriguez et al., 2023; van den Berg et al., 2015; Williamson et al., 2019).

Seven-month-old female Mecp2 HET mice covered significantly shorter distances during each of the 10-min sessions of both the sociability phase (empty cup versus novel mouse; P=0.0006, ANOVA-Bonferroni's; n=9 control WT mice versus n=10 control Mecp2 HET mice; Fig. S2A) and social memory phase (familiar mouse versus unfamiliar mouse; P=0.0027, ANOVA-Bonferroni's; n=10 control WT mice versus n=10 control Mecp2 HET mice; Fig. S2C) of the three-chamber test. These differences were not altered by LM22A-4 either in the sociability trial (P<0.0001, ANOVA-Bonferroni's; n=11 LM22A-4 WT mice versus n=9 LM22A-4 Mecp2 HET mice; Fig. S2A) or in the social memory trial (P=0.0052, ANOVA-Bonferroni's; n=11 LM22A-4 WT mice versus n=9 LM22A-4 Mecp2 HET mice; Fig. S2C). Similarly, 7-month-old female Mecp2 HET mice exhibited slower velocities than those of age-matched female WT mice in the sociability trial (P=0.0006, ANOVA-Bonferroni's; n=9 control WT mice versus n=10 control Mecp2 HET mice; Fig. S2B) and in the social memory trial (P=0.0028, ANOVA-Bonferroni's; n=10 control WT mice versus n=10 control Mecp2 HET mice; Fig. S2D). These differences were also not altered by LM22A-4 either in the sociability trial (P<0.0001, ANOVA-Bonferroni's; n=11 LM22A-4 WT mice versus n=9 LM22A-4 Mecp2 HET mice; Fig. S2B) or in the social memory trial (P=0.0053, ANOVA-Bonferroni's; n=11 LM22A-4 WT mice versus n=9 LM22A-4 Mecp2 HET; Fig. S2D). Finally, the total walking times during the 10 min of the unrestricted social assay (see below) were not different between genotypes and were not affected by LM22A-4 (P>0.9999, ANOVA-Bonferroni's; among all groups; Fig. 5). This indicates that locomotion deficits may be context dependent and could contribute to the lack of clear sociability and social memory phenotypes in female Mecp2 HET mice, although female WT mice without this locomotion deficit showed a rather mild preference in the standard three-chamber test, evident only in the first minute (Fig. 3).

Because the novel (unfamiliar) and familiar mice used as social stimuli were restrained under inverted pencil cups during the standard social interaction test (Moy et al., 2004), which could result in experimental confounds (e.g. stress vocalizations by stimulus mice), we performed an unbiased assay of naturalistic behaviors shown by a test mouse when interacting with an unfamiliar mouse and a littermate, all freely moving in a large arena (30×40 cm) for 10 min. The three mice were recorded from above, and their individual trajectories were tracked with Mouse Tracker (Motr) software (Ohayon et al., 2013), which was used to train the machine-learning model Janelia Automatic Animal Behavior Annotator (JAABA) (Kabra et al., 2013) with specific behaviors, thus producing an unbiased scoring of behaviors by the test mouse when interacting with freely moving unfamiliar and familiar mice. We successfully used this approach to reveal altered social behaviors in symptomatic male Mecp2 KO mice (Phillips et al., 2019).

Five-month-old female Mecp2 HET mice (n=6) showed more ‘face-following’ (P=0.00358) and more ‘nose-sniffing’ (P=0.0005) behavior than that of WT mice (n=6, ANOVA-Bonferroni's; Fig. 5A). A 4-week treatment with LM22A-4 reduced nose-sniffing behavior in Mecp2 HET mice (n=6) to levels comparable to those shown by WT mice (P=0.0487, ANOVA-Bonferroni's; control Mecp2 HET versus LM22A-4 Mecp2 HET), whereas treatment had no effects in WT mice (n=9; Fig. 5A).

In a slightly older cohort of 7-month-old female mice, there were no longer differences in face-following and nose-sniffing behavior (P>0.9999, ANOVA-Bonferroni's, n=12 Mecp2 HET and n=12 WT mice; Fig. 5B). Furthermore, two other behaviors altered in male Mecp2 KO mice, ‘rear sniffing’ and atypical ‘piggy-back jumping’ (Phillips et al., 2019) were unchanged between female Mecp2 HET and WT mice (P>0.9999, ANOVA-Bonferroni's; Fig. 5B). The same results were obtained using the average discrimination indices of the whole interaction time or broken down to a minute-by-minute basis (P>0.9999, ANOVA-Bonferroni's; among all groups; face following, Fig. 4A,B; rear sniffing, Fig. 4C,D; atypical piggy-back jumping, Fig. 4E,F; nose sniffing, Fig. 5B). Consistent with this lack of differences between genotypes, LM22A-4 had no effects on face-following, rear-sniffing, atypical piggy-back jumping and nose-sniffing behaviors shown by Mecp2 HET (n=12) and WT (n=12) mice, either using the 10-min average discrimination indices or in a minute-by-minute basis.

Five-month-old Mecp2 HET mice were ‘followed’ more than WT mice (P=0.0130, ANOVA-Bonferroni's; control WT versus control Mecp2 HET; Fig. 5A), which was also observed in 6-month-old mice (P=0.0003, ANOVA-Bonferroni's; control WT versus control Mecp2 HET; Fig. 5B). Unexpectedly, 6-month-old mice Mecp2 HET mice treated with LM22A-4 showed lower times being followed than those for control Mecp2 HET mice (P=0.0074, ANOVA-Bonferroni's; LM22A-4 Mecp2 HET versus control Mecp2 HET; Fig. 5B).

Five-month-old female Mecp2 HET mice exhibited longer periods of ‘aggressive’ behaviors (P<0.0001) and ‘chasing’ behaviors (P=0.0008) than those for age-matched WT mice (ANOVA-Bonferroni; control Mecp2 HET versus control WT; Fig. 5A). A 4-week treatment with LM22A-4 reduced the levels of chasing behaviors in Mecp2 HET mice to WT levels (P=0.0033), but it did not affect aggressive behaviors (P=0.0180, ANOVA-Bonferroni; LM22A-4 Mecp2 HET versus control Mecp2 HET; Fig. 5A). In contrast, no differences between genotypes nor effects of LM22A-4 on aggressive or chasing behaviors were observed in 7-month-old mice (P>0.05, ANOVA-Bonferroni's; Fig. 5B).

Five-month-old female Mecp2 HET mice were the target of aggressive chasing (P=0.0008) and aggressive behaviors (P=0.0010) for longer times than those for age-matched WT mice (ANOVA-Bonferroni's; control Mecp2 HET control versus control WT; Fig. 5A). Interestingly, the 4-week LM22A-4 treatment reduced the duration for which Mecp2 HET mice were being chased aggressively (P=0.0359, ANOVA-Bonferroni's; LM22A-4 Mecp2 HET versus control Mecp2 HET; Fig. 5A). However, LM22A-4 did not affect the duration of Mecp2 HET mice being the target of aggressive behaviors (P=0.1242, ANOVA-Bonferroni's; LM22A-4 Mecp2 HET versus control Mecp2 HET; Fig. 5A).

Seven-month-old female Mecp2 HET mice showed more ‘shuffling’ locomotion than that of age-matched WT mice (P=0.0035, ANOVA-Bonferroni's; control Mecp2 HET versus control WT; Fig. 5B), but this was not affected by LM22A-4 (P<0.0001, ANOVA-Bonferroni's; LM22A-4 Mecp2 HET versus control Mecp2 HET; Fig. 5B).

Five-month-old female Mecp2 HET mice showed shorter periods of ‘digging’ (P=0.0023) and ‘air-sniffing’ behaviors (P=0.0052) than those for WT mice (ANOVA-Bonferroni's; control Mecp2 HET versus control WT; Fig. 5A), but LM22A-4 did not affect these durations (P>0.9999, ANOVA-Bonferroni's; LM22A-4 Mecp2 HET versus control Mecp2 HET; Fig. 5B).

In the present study, we tested the in vitro and in vivo effects of the TrkB partial agonist LM22A-4 on dendritic spine density and volume in hippocampal pyramidal neurons of Mecp2-deficient mice, as well as in the classical three-chamber test of social behaviors and a novel machine-learning unbiased screen of open-field behaviors in female Mecp2 HET mice interacting with unfamiliar and familiar mice.

We first confirmed that LM22A-4 mimics the TrkB-dependent spinogenic effect of BDNF in organotypic slice cultures from neonatal male Mecp2 KO mice, but not in slice cultures from WT mice due to their typical levels of BDNF that outcompete the partial agonism of LM22A-4 on TrkB receptors. This property of LM22A-4 was described in the original paper (Massa et al., 2010) and in all subsequent reports using different experimental mouse models with reduced BDNF levels, including models of Huntington's disease (Simmons et al., 2013), traumatic brain injury-induced epilepsy (Gu et al., 2018), Dravet's disease (Gu et al., 2022), compulsive alcohol drinking (Warnault et al., 2016), non-arteritic anterior ischemic optic neuropathy (Ali Shariati et al., 2018) and chemotherapy-induced cognitive decline (Geraghty et al., 2019).

In the context of RTT, LM22A-4 also showed therapeutic effects in female Mecp2 HET mice: it improved breathing irregularities (Kron et al., 2014; Schmid et al., 2012) and mitigated network hyperactivity in hippocampal slices, which restored LTP of CA1 excitatory synaptic transmission in hippocampal slices and improved hippocampal-dependent object location memory (Li et al., 2017). Similarly, a second-generation TrkB ligand derived from LM22A-4 also improved breathing patterns and motor deficits in female Mecp2 HET mice (Adams et al., 2020). Because these studies identified therapeutic effects of a 1- to 2-month LM22A-4 treatment in 3- to 4-month-old female Mecp2 HET mice, we followed the same dosage regimen to assess the effects of LM22A-4 effects on dendritic spine phenotypes in hippocampal pyramidal neurons, and on an unbiased screen of atypical behaviors at an age where most RTT-like phenotypes appear in female Mecp2 HET mice (Samaco et al., 2013), including breathing apneas (Kron et al., 2014; Schmid et al., 2012) and altered hippocampal dependent memory (Li et al., 2017).

Using female MeCP2-GFP HET mice revealed a non-cell-autonomous consequence of Mecp2 deletion on dendritic spines in the mosaic brain of female RTT mice: CA1 hippocampal pyramidal neurons that express MeCP2 (i.e. WT cells) had larger dendritic spines than neighboring neurons lacking MeCP2 (i.e. mutant cells) and those in female WT mice. We interpret these larger spines to reflect an activity-dependent response of MeCP2-expressing neurons to the atypically heightened hippocampal network activity in female Mecp2 HET mice (Li et al., 2017). In contrast, MeCP2-lacking neurons may have lost their capacity for dendritic spine plasticity, similar to the saturation of LTP in CA1 pyramidal neurons in male Mecp2 KO mice, caused by atypically stronger excitatory synapses due to impaired synaptic trafficking of GluA1 (encoded by Gria1)-containing AMPA receptors (Li et al., 2016). Note that the expression of TrkB receptors is not affected by Mecp2 deletion (Li et al., 2012). These observations resemble the non-cell-autonomous consequences of Mecp2 deletion in the primary motor cortex of female Mecp2 HET mice, where Mecp2-expressing neurons have a lower spine density than that of both Mecp2-lacking neurons and neurons in female WT mice (Belichenko et al., 2009). Moreover, the size of the cell body of Mecp2-expressing neurons in female Mecp2 HET mice is smaller than that of neurons in female WT mice (Rietveld et al., 2015; Wither et al., 2013; reviewed in Ribeiro and MacDonald, 2020). Interestingly, LM22A-4 reduced the spine volume in MeCP2-expressing neurons to levels seen in female WT mice, which is consistent with its reduction of hippocampal hyperactivity that allowed typical LTP and restored object-location memory in female Mecp2 HET mice (Li et al., 2017).

Here, we investigated the effects of a 4-week LM22A-4 treatment on social behaviors in female Mecp2 HET mice, focusing on key aspects such as sociability, social memory and naturalistic behaviors. In the standard three-chamber social interaction test, no differences in sociability and social memory were found, with LM22A-4 treatment demonstrating no effect. Minute-by-minute analysis of the three-chamber social interaction test revealed a delayed social memory in Mecp2 HET mice in comparison to that in WT controls, an effect not mitigated by LM22A-4 treatment. The unrestricted assay highlighted the efficacy of LM22A-4 in reducing aggressive behaviors in 5-month-old Mecp2 HET mice, surmising its positive impact on disease manifestations. However, this effect was more limited in an older cohort (7-month-old mice), suggesting a potential impact of disease progression in Mecp2 HET mice. Notably, at 7 months of age, Mecp2 HET mice demonstrated increased shuffling locomotion, further emphasizing the progression of the observed effects. All in all, our study brings to light behavioral alterations in female Mecp2 HET mice, presenting comprehensive social dynamics across the disease progression. These findings demonstrate the potential therapeutic significance of LM22A-4, providing a promising avenue for investigation in the context of RTT and neurodevelopmental disorders.

The progression of behavioral phenotypes in female Mecp2 HET mice is not a simple process of all deficits equally worsening with age, resembling the onset and waning of different neuropsychiatric signs in individuals with RTT during the regression, pseudo-stationary and deterioration phases of the disorder (Samaco and Neul, 2011). Our results indicate that female Mecp2 HET mice display more aggressive behaviors at 5 months of age than age-matched female WT mice, which is absent at 7 months of age. Another example is the anxiety phenotype during the open-field test, which was enhanced in 3-month-old female Mecp2 HET mice, whereas this difference was absent at 5 months of age and returned later at 7 months (Ribeiro and MacDonald, 2022).

Taken together, these data indicate that social and aggression behaviors may be altered in female Mecp2 HET mice prior to their shuffling locomotion, and that LM22A-4 only improves some of the affected behaviors. Interestingly, social aggression was reported in male mice lacking Mecp2 in either Sim1-expressing or serotonergic PET1 (FEV)-expressing neurons (Samaco et al., 2009). Similarly, Mecp2 deletion in PET1-expressing neurons led to an aggressive phenotype in male mice, despite a lack of an anxiety phenotype, as tested in the open-field or ‘light-dark box’ assays (Fyffe et al., 2008). In a different study, male Mecp2 KO mice showed hyper-reactive escape and defensive behaviors in a ‘mouse defense test battery’ assay (‘predator avoidance’, ‘chase/flight’, ‘closed-door approach’, and ‘forced-contact’ tests), despite the lack of an anxiety phenotype, as tested in the ‘elevated plus’ maze and ‘elevated zero’ maze (Pearson et al., 2015). Intriguingly, a phenotype-based genetic association study revealed that MeCP2 protein levels in mice and MECP2 single nucleotide polymorphisms in individuals with schizophrenia are associated with social aggression behaviors (Tantra et al., 2014). Furthermore, studies in male mice with deletions or knockdown of either Bdnf or TrkB display enhanced aggressive phenotypes (Adachi et al., 2017; Ito et al., 2011; Lyons et al., 1999), enhanced anxiety and have increased body weight (Rios et al., 2001). Overall, our results demonstrate that female Mecp2 HET mice may have altered aggression phenotypes, which should be further explored in deeper detailed in future studies.

Genetic background also contributes significantly to behavioral phenotypes in RTT mice: female Mecp2 HET mice bred into two different mixes of genetic strains, FVB/N×129S6/SvEv and 129S6/SvEv×C57BL/6, showed altered sociability, contextual fear memory and passive avoidance behaviors at 3 and 5 months of age, whereas at 4 months of age, they showed shorter distance traveled in the open field, suggestive of motor impairment (Samaco et al., 2013). At 5 weeks of age, Mecp2 HET mice spent more time in the open arm in the elevated plus maze test and greater time in the light side of the light-dark box assay (Samaco et al., 2013). These data are further recapitulated by Mecp2^1lox mice spending more time in the open arms of the open-field apparatus as well as the open-zero maze at 8 weeks of age (Stearns et al., 2007). All in all, the reduced anxiety phenotype of Mecp2 HET mice is consistent among multiple studies, whereas aggressive behaviors in Mecp2 KO males was also found to be a consistent behavioral phenotype, which was at large unexplored in female mice.

Collectively, our observations indicate that sex, genotype and possibly genetic background are highly determinant for the presentation of behavioral phenotypes in MeCP2-based mouse models of RTT, and, therefore, these criteria should be included when defining the outcome measures used for characterizing the longitudinal progression of the disorder and the efficacy of therapeutic approaches.

Breeding pairs of mice lacking exon 3 of the X chromosome-linked Mecp2 gene (B6.Cg-Mecp2^tm1.1Jae, ‘Jaenisch’ strain in a pure C57BL/6 background) (Chen et al., 2001) were purchased from the Mutant Mouse Regional Resource Center at the University of California, Davis (stock #000415). A colony was established at the University of Alabama at Birmingham by mating male WT C57BL/6 mice with female Mecp2 HET mice. Genotyping was performed by PCR of DNA samples from tail clips at weaning age P28. Male hemizygous Mecp2 mice (i.e. Mecp2 KO), develop typically until 5-6 weeks of age (P35-P42), when they begin to exhibit RTT-associated motor symptoms, such as hypoactivity, hind limb clasping and reflex impairments (Chen et al., 2001). Male transgenic mice expressing GFP-tagged MeCP2 (Lyst et al., 2013; The Jackson Laboratory stock #014610) were a gift from Dr Adrian Bird (University of Edinburgh, Scotland) and Dr Ben Philpot (University of North Carolina, USA), and were crossed with female Mecp2 HET mice, which allows identification of the cellular genotype in the female mosaic brains resulting from X-chromosome inactivation. Female Mecp2 HET mice develop similar RTT-associated symptoms as those observed in male Mecp2 KO mice, but with a delayed progression, most evidently starting around 2-3 months of life (Samaco et al., 2013). Animals were handled and housed according to the Committee on Laboratory Animal Resources of the National Institutes of Health (NIH). All experimental protocols were annually reviewed and approved by the Institutional Animals Care and Use Committee (IACUC) of the University of Alabama at Birmingham.

Organotypic hippocampal slice cultures were prepared from male postnatal day 5-7 (P5-P7) Mecp2 KO mice and WT littermates as previously described (Chapleau et al., 2009; Pozzo Miller et al., 1993). Briefly, 500-µm-thick hippocampal slices were cut with a custom-made wire-slicer (Katz, 1987), plated on tissue culture plate inserts (Millicell-CM, Millipore) inside six-well plates with culture medium containing 78% Neurobasal-A medium without phenol red (Invitrogen), 20% heat-inactivated equine serum (Invitrogen), 2% B27 supplement (Invitrogen) and 0.5 mM L-glutamine (Invitrogen), and placed in an incubator at 36°C, 5% CO₂ and 90% relative humidity. Serum was titrated out over 3 days in vitro (DIV), and treatments were performed in serum-free culture media, as previously described (Chapleau et al., 2008).

Seven-DIV slice cultures were transfected with a plasmid encoding eYFP by particle-mediated gene transfer using a Helios Gene Gun (Bio-Rad) as described previously (Alonso et al., 2004; Chapleau et al., 2009). Briefly, the eYFP-encoding cDNA plasmid (Clontech) was precipitated onto colloidal gold particles (1.6 µm; Bio-Rad) and coated onto Tefzel tubing. Slices were bombarded with gold particles accelerated by ∼586.25 kPa of helium gas from a distance of 2 cm using a modified gene-gun nozzle with a 2 µm filter. Biolistic transfections were performed 24 h after culture medium was changed to serum-free culture medium with a commercial antibiotic/antimycotic mixture (1:100; penicillin/streptomycin/amphotericin-B; also known as Fungizone, Invitrogen). 24 h after transfection, the culture medium was changed to serum-free medium without antibiotics or antimycotics.

Nine-DIV hippocampal slice cultures were randomly assigned to treatment groups with the following drugs dissolved in serum-free medium: recombinant human BDNF (250 ng/ml, Promega); BDNF+K-252a (200 nM, Calbiochem); K-252a alone; LM22A-4 (500 nM; synthesized and validated in the laboratory of F.M.L.); LM22A-4+K-252a; LM22A-4+ANA-12 (100 µM, Sigma-Aldrich); and ANA-12 alone. Culture medium was removed from culture wells and replaced with drug-containing medium; in addition, 50 μl of drug-containing medium was gently applied on top of each slice. All treatments lasted 48 h. Eleven-DIV slice cultures were fixed with 4% paraformaldehyde in 100 mM phosphate buffer overnight at 4°C, washed with 100 mM phosphate-buffered saline (PBS), trimmed from the filter membrane of the cultured inserts, and mounted on glass slides with Vectashield (Vector Laboratories).

Four- to six-month-old female Mecp2 HET mice and female MeCP2-GFP HET mice and their age-matched female WT littermates received intraperitoneal injections of either sterile LM22A-4 (50 mg/kg) or vehicle (0.9% NaCl) twice daily for 4 weeks, following an established dosing regime with efficacy in other studies using Mecp2 mutant mice (Kron et al., 2012; Li et al., 2017; Schmid et al., 2012). Mice were randomly assigned to each treatment. LM22A-4 was prepared daily from sterile stocks and dissolved in sterile saline.

Mice were deeply anesthetized with a mixture of 100 mg/kg ketamine and 10 mg/kg xylazine, and transcardially perfused with ice-cold ‘cutting’ artificial cerebrospinal fluid (aCSF), containing 87 mM NaCl, 2.5 mM KCl, 0.5 mM CaCl₂, 7 mM MgCl₂, 1.25 mM NaH₂PO₄, 25 mM NaHCO₃, 25 mM glucose and 75 mM sucrose, bubbled with 95% O₂ and 5% CO₂. The brain was rapidly removed and cut transversely at 300 µm sections using a vibrating blade microtome (VT1200S, Leica Microsystems) in the same ice-cold cutting aCSF. Slices were transferred to normal oxygenated aCSF containing 119 mM NaCl, 2.5 mM KCl, 2.5 mM CaCl₂, 1.3 mM MgCl₂, 1.3 mM NaH₂PO4, 26 mM NaHCO₃ and 20 mM glucose (with 95% O₂ and 5% CO₂) at 32°C for 30 min, and then allowed to recover for 1 h at room temperature before use. Individual slices were transferred to a submerged chamber mounted on a fixed-stage upright microscope (Axioskop FS or AxioExaminer D1, Zeiss) and continuously perfused at room temperature with normal oxygenated aCSF containing 119 mM NaCl, 2.5 mM KCl, 2.5 mM CaCl₂, 1.3 mM MgCl₂, 1.3 mM NaH₂PO4, 26 mM NaHCO₃ and 20 mM glucose (with 95% O₂ and 5% CO₂). Pyramidal neurons in CA1 stratum pyramidale were visualized by infrared differential interference contrast microscopy with water-immersion objectives (63×0.9 NA, or 20×1.0 NA plus 0.50-4× zoom, Zeiss). In MeCP2-GFP mice, GFP was imaged with 475 nm LED illumination (X-Cite Turbo, Excelitas), a GFP cube (Semrock) and a QuantEM:512SC cooled CCD (Photometrics). Whole-cell pipettes contained 120 mM Cs-gluconate, 17.5 mM CsCl, 10 mM Na-HEPES, 4 mM Mg-ATP, 0.4 mM Na-GTP, 10 mM Na₂-creatine phosphate, 0.2 mM Na-EGTA and 8 mM biocytin (Sigma-Aldrich), at 290-300 mOsm, pH 7.3 (final resistance, 3-4 MΩ). After ∼15 min of whole-cell access to allow biocytin loading, slices were fixed in 4% paraformaldehyde in 100 mM PBS and stained with streptavidin-conjugated Alexa Fluor 488 (Life Technologies) as described previously (Li et al., 2016).

eYFP-expressing CA1 and CA3 pyramidal neurons in hippocampal slice cultures from male Mecp2 KO mice and male WT littermates, and biocytin-filled, Alexa Fluor 488-labeled CA1 pyramidal neurons in hippocampal slices from female MeCP2-GFP HET mice and female WT littermates were selected for confocal imaging if they showed fluorescent labels throughout the entire dendritic tree and lacked signs of degeneration (e.g. dendritic ‘blebbing’). High-resolution z-stack images of apical secondary and tertiary dendrites were acquired with oil-immersion 60× (NA 1.42) objectives (PlanApo) plus 3× digital zoom in either a Fluoview FV-300 (Olympus) or a LSM-800 (Zeiss) confocal microscope. Image stacks were acquired at 0.1 µm intervals in the z-plane.

Z-stack image stacks were three-dimensionally reconstructed, surface rendered and analyzed semi-automatically with the ‘Filament Tracing’ module of Imaris (Bitplane) (Swanger et al., 2011). Image stacks were loaded into Imaris, a region of interest including a 40-to 80-µm-long dendritic shaft was selected, and ‘Filament Tracing’ was used to trace and render dendrites and their associated spines. Spines were defined as dendritic protrusions shorter than 3 μm. The numerical densities of spines were calculated for each dendritic segment and normalized to 10 μm of dendritic length. The relative intensity of individual spines and their parent dendritic shafts was used to estimate the volume of individual spines (Benavides-Piccione et al., 2013). An experimenter masked to genotype and treatment performed dendritic spine analyses.

All handling and testing were done in the dark phase of the standard 12 h light/12 h dark cycle (06:00 ON, 18:00 OFF), with the experimenter wearing a red headlamp and infrared illumination for digital videography.

We used the standard three-chamber social interaction test (Moy et al., 2004). Prior to testing, mice underwent a 3-day acclimation period during which they were familiarized to the experimenter with handling for 3 min each day at the same time as testing. The testing environment was a three-chambered box that contained two empty inverted pencil cups placed in the two side chambers. Mice were placed in the central chamber and allowed to freely explore for 5 min. After this initial acclimation period, mice were directed to return to the central chamber, and barriers were positioned over the side openings. At this time, a novel mouse was placed beneath one of the previously empty pencil cups in one of the two side chambers, alternating between trials. The barriers were then removed and the test mouse was allowed to freely explore the chambers for 10 min. Subsequently, the test mouse was again directed to return to the central chamber and barriers were positioned over the side openings. A second novel mouse was placed beneath the formerly empty pencil cup, the previously added mouse was now considered ‘familiar’. After removing the side barriers, the test mouse was allowed to explore the chambers for an additional 10 min. Following each test, the apparatus was meticulously cleaned with 70% isopropanol. The time that the test mouse spent actively sniffing each pencil cup, containing either a social conspecific or empty, was quantified. For the sociability test, the discrimination index was calculated as: [(time investigating cup with mouse)−(time investigating empty cup)]/[(time investigating cup with mouse)+(time investigating empty cup)]×100. For the social memory trial, the discrimination index was calculated as: [(time investigating cup with unfamiliar mouse)−(time investigating cup with familiar mouse)]/[(time investigating cup with unfamiliar mouse)+(time investigating cup with familiar mouse cup)]×100. Test mice that spent over 75% of the acclimation time in a single compartment were excluded from the study. Statistical analyses were performed on the discrimination indices for each test mouse.

One week prior to testing, the backs of sentinel novel and familiar mice were dyed with blond hair dye (Born Blonde Maxi, Clairol) with different patterns for computer vision tracking. ‘Sentinel’ mice were the familiar and unfamiliar mice used as social stimuli for the ‘test’ mouse in the unrestricted social assay. Mice were habituated to the testing arena, an open-field box (30×40 cm) containing clean bedding, for 3 days prior to the testing day, 10 min per day. During the testing day, test mice were habituated to the testing arena for 10 min. Sentinel mice, including one cage-mate and one unfamiliar mouse from a different cage, were placed in the testing arena and were allowed to interact freely for 10 min. After this time, sentinel mice were placed into a neutral cage and the test mouse was returned to their home cage. The arena was meticulously cleaned with 70% isopropanol and filled with new bedding between each test. Sentinel mice interacted with a maximum of five test mice and were no longer used within the study if they fought with other sentinels or displayed excessive grooming phenotypes. Test videos were loading into Motr (https://github.com/motr/motr) (Ohayon et al., 2013) to create tracks that were exported to the machine-learning model JAABA (https://github.com/kristinbranson/JAABA) (Kabra et al., 2013) for unbiased behavioral scoring. JAABA classifiers were first trained on pilot data sets. Behavioral scores for social memory and other behaviors were taken from the entire video, as different behaviors emerged at later times during the 10-min trial, and social behavior times were pooled between those for unfamiliar and cage-mate mice. Mice not interacting with sentinels for more than 3 s (out of 240 s) were excluded from analyses. Statistical tests were performed on the time per behavior for each test mouse.

All experiments were repeated in at least three independent slice cultures or ex vivo acute slice preparations. All statistical analyses were performed using Prism (GraphPad Software), P<0.05 was considered significant. Outliers were detected using the ROUT method. Data are presented as mean±standard error of the mean (s.e.m.) and were compared using unpaired two-tailed Student's t-tests for two groups or one-way ANOVA with Bonferroni's post hoc test for more than two groups. Data with more than three groups that did not fit the normal distribution were analyzed using the Kruskal–Wallis test, with Dunn's post hoc test for multiple comparisons; post hoc tests were only performed for ANOVA and Kruskal–Wallis tests that reported P<0.05. The Kolmogorov–Smirnov (K-S) test was used for comparisons of cumulative frequency distributions.

We thank Ms Lili Mao for mouse colony management and neuronal cultures, Dr Takafumi Inoue (Waseda University, Tokyo, Japan) for data acquisition and analysis software, and Dr Adrian Bird (University of Edinburgh, Scotland) and Dr Ben Philpot (University of North Carolina, USA) for MeCP2-GFP reporter mice. Confocal microscopy was performed in the Cellular Imaging facility of the Civitan International Research Center at the University of Alabama at Birmingham.