New mouse line overexpressing truncated β-cat
N-terminally truncated β-cat is reported to be produced endogenously by NMDA receptor stimulation activating calcium-dependent calpain protease cleavage of full-length β-cat38. Consistently, we show increases in truncated β-cat levels induced by NMDA stimulation of acutely isolated synaptosomes from wildtype mouse hippocampus (Fig. 1a). The increases are prevented by NMDA stimulation in the presence of a calpain inhibitor (MDL 28170). These findings suggest activity-dependent, dynamically regulated production of a β-cat isoform whose neural function is unknown.
We are gaining insights into its role in learning using our newly generated mouse line that expresses truncated β-cat in forebrain glutamatergic neurons. We used conditional deletion of the β-cat degradation domain, resulting in the accumulation of stable truncated β-cat, even in the presence of its negative regulator, APC. We crossed CTNNB1fl(ex3)/+ mice37 with CamKIIα-Cre-93 mice39 (Fig. 1b), that express Cre recombinase predominantly in forebrain excitatory neurons, starting at late embryonic ages and progressively increasing to complete activation by P2136. This is a disease relevant developmental age and cell type for intellectual disabilities and autism based on both mouse and human brain studies40–43. The loxP sequences flank exon 3 of β-cat which encodes the domain (amino acids 4–80) that contains the GSK3-β and CK1 phospho-sites necessary for targeting to the APC-containing degradation complex (Fig. 1c). This 77 kDa construct closely resembles the β-cat N-terminal truncations (85 and 75 kDa products) observed physiologically upon NMDAR stimulation activating calpain38 (Figs. 1a,c).
Our β-cat cOE mice exhibit significant increases in β-cat levels (combined expression of the full-length product from the wild-type (WT) allele and truncated product from the exon3 deficient allele), relative to control littermates (fl/ex3, Cre negative and +/+, Cre positive), as seen in quantitative immunoblots of hippocampal lysates (Fig. 1d). Our previous studies of APC cKO mice, using the same CamKIIα-Cre-93 driver, show increased levels of full-length β-cat and moderate cognitive disabilities36. Total β-cat levels are elevated comparably in β-cat cOEs and APC cKOs, enabling a comparison of phenotypes caused by overexpression of truncated versus full-length β-cat, in the presence versus absence of APC.
Importantly, the truncated β-cat peptide retains all known β-cat protein interaction domains (Fig. 1c). Using co-immunoprecipitation, we show that it interacts with partners central to β-cat’s two key functions, N-cadherin in synaptic adhesion complexes and LEF1, the nuclear transcription co-activator in the canonical Wnt signal transduction pathway (Fig. 1e). We confirm this latter function by showing increased expression of several canonical Wnt target genes, SP5, neurog1 and syn244–46 (Fig. 1f).
Severe Learning Impairments
CTNNB1 and APC human gene mutations link to intellectual disabilities, including disruptive mutations in CTNNB1 exon 3, encoding the degradation domain3, 5, 6, 18. Further, APC cKO mice, with elevated full-length β-cat levels and APC loss, exhibit moderate learning and memory deficits36. We therefore examined the cognitive abilities of β-cat cOE mice with elevated levels of truncated β-cat, in the presence of APC.
Compared to control littermates, β-cat cOE mice display severe cognitive impairments in two different spatial learning tasks. In the Barnes Maze task, they are unable to learn the goal hole location over the 8 trial training period (Fig. 2a). They could not proceed to probe trial testing for memory acquisition due to the lack of learning. In comparison, APC cKO mice show moderate deficits suggestive of delayed learning, they take significantly longer to learn the location of the goal hole, compared to control littermates (Fig. 2a). Upon probe trial testing, APC cKOs are unable to perform the task (recall the goal hole location) one week after training, whereas littermate controls do retain the memory.
In the contextual fear-conditioning task, β-cat cOE mice exhibit severe deficits (Fig. 2b), with little improvement over the course of training (7 days). They never reach the freezing level of control littermates, and thus could not proceed to probe trials. APC cKO mice take longer to associate the context with the foot shock, compared with littermate controls; they also show deficits in long-term retention of the fear memory (Fig. 2b).
Neither locomotion deficits nor anxiety are potential confounds for the mutant mouse lines. The distance travelled and time spent in the center, both measured in the open field task, show no difference between genotypes (Supplementary Fig. 1a). Further, pain threshold does not differ between the genotypes, as measured by latency to paw lick on a heated plate (Supplementary Fig. 1b), suggesting the absence of sensory defects such as reduced sensitivity to the foot shock.
Thus, although β-cat levels are increased to similar levels in β-cat cOEs and APC cKOs, they exhibit cognitive deficits of significantly different severities. These results strongly suggest critical roles for both truncated β-cat and APC in cognitive function.
Reduced Synaptic Plasticity
To begin to elucidate mechanisms underlying the differences in cognitive skills between β-cat cOEs and APC cKOs, we examined electrophysiological correlates of synaptic plasticity required for learning and memory, both long-term potentiation and depression (LTP and LTD). Compared with control littermates, β-cat cOEs exhibit severely reduced LTP induced by 100Hz theta-burst stimulation (TBS) of CA3-CA1 synapses in acute hippocampal slices (Fig. 3a). β-cat cOE TBS-LTP shows decreased induction and only slight, albeit significant, increases in fEPSP slope above baseline values. In contrast, APC cKOs show modest enhancement of LTP, relative to littermate controls (Fig. 3a).
β-cat cOEs also display reduced LTD (Fig. 3b). One Hz stimulation of CA3-CA1 synapses induces LTD, but the reduction in fEPSP slope is not maintained and rapidly returns to baseline levels, relative to control littermates. In contrast, APC cKOs show normal LTD (Fig. 3b).
Basal synaptic transmission is not altered at β-cat cOE CA3-CA1 synapses, based on extracellular recordings of field excitatory postsynaptic currents (Supplementary Fig. 2). There is no change in the ratio of stimulus intensity to the slope of field excitatory postsynaptic currents (input/output) in either β-cat cOEs or APC cKOs36.
Altered Glutamate Receptor Levels
Next, we tested for glutamatergic synapse molecular changes that may underlie the differences in learning and synaptic plasticity between β-cat cOE and APC cKO mice. LTP and LTD are elicited by NMDAR activation leading to rapid insertion or removal, respectively, of AMPA receptors (AMPARs) containing the GRIA1 subunit that is trafficked in an activity-dependent manner. GRIA1 levels are decreased in β-cat cOEs, compared to control littermates, as determined by immunoblotting of membrane preparations isolated from the hippocampus (Fig. 3c). In contrast, GRIA1 levels are increased in APC cKOs, relative to control littermates.
Because hippocampal AMPARs are composed of primarily GRIA1/GRIA2 or GRIA2/GRIA3 heterodimers47, we also measured the levels of GRIA2, the constitutively trafficked subunit, to determine whether overall AMPAR levels may be changed. GRIA2 levels are normal in both β-cat cOEs and APC cKOs (Fig. 3c), consistent with their normal levels of basal synaptic transmission. These findings suggest a selective reduction in surface membrane levels of activity-dependent trafficked GRIA1, rather than a reduction in total AMPAR surface levels.
Further, NMDAR GRIN2A and GRIN2B subunit levels are both decreased in β-cat cOE hippocampal membrane fractions, as measured by immunoblotting (Fig. 3c). In contrast, GRIN2A and GRIN2B levels are unchanged in APC cKOs, compared with littermate controls. The reduced surface levels of NMDAR GRIN2A and 2B and AMPAR GRIA1 are likely responsible for the dramatic decreases in LTP induction and lack of maintenance of both LTP and LTD at β-cat cOE hippocampal CA1-CA3 synapses (Fig. 3a,b).
Protein 4.1N, an actin cytoskeleton binding protein, traffics AMPAR GRIA1 to postsynaptic sites48, 49. We show that protein 4.1N levels are altered in opposite directions in β-cat cOEs and APC cKOs, decreased in cOEs and increased in cKOs, similar to GRIA1 surface membrane levels (Supplementary Fig. 3) Although the extent of change is modest, similarly small alterations in protein 4.1N levels have been shown to significantly alter both GRIA1 insertion and synaptic plasticity48, 49.
Truncated β-cat Alters APC Function
APC, a direct binding partner of β-cat, is a large scaffold protein with multiple binding partners and functions required for excitatory synapse maturation and normal learning and memory36, 50–52. We therefore tested whether truncated β-cat alters APC function.
Relative to full-length β-cat, the truncated isoform displays greater association with APC (Fig. 4a) in co-immunoprecipitations with β-cat cOE versus control littermate hippocampal lysates (data standardized to total abundance of the isoforms in inputs). As support for the physiological relevance of this interaction, we show greater association between APC with truncated β-cat isoforms produced by recombinant calpain cleavage of hippocampal lysate from wild-type mice (Fig. 4b). The calpain-cleaved 85 and 75 kDa β-cat products show approximately 2 and 4 times more binding to APC, relative to the more abundant full-length β-cat.
We noted that APC displays a size-based mobility shift in immunoblots of the β-cat cOE hippocampus, compared to control littermates (Fig. 4c), suggesting possible posttranslational modification. Treating the lysate with λ-phosphatase ablates the mobility shift. Total APC protein levels do not differ between β-cat cOEs and littermate controls, as measured after λ-phosphatase treatment. Neither do APC mRNA levels, as determined by RT-qPCR (Fig. 4d). Thus, elevated levels of stable truncated β-cat lead to increased association with APC and increased APC phosphorylation in hippocampal neurons in vivo. Similarly, in non-neuronal cells in vitro, truncated β-cat (lacking the degradation domain) exhibits ~ 10-fold greater affinity for APC and increased APC phosphorylation, compared with full length β-cat53, 54.
Because APC is hyperphosphorylated in the β-cat cOEs, we tested for changes in the levels and activation of GSK3β, a kinase that phosphorylates both APC and β-cat, leading to increases in their binding affinity for one another55. We found no significant changes in either GSK3β total levels, or the levels of phosphorylation at Ser 9 (deactivating) or Tyr 216 (activating) of GSK3β in the β-cat cOE hippocampus, relative to control littermates (Fig. 4e). Thus, APC hyperphosphorylation in β-cat cOEs may stem from changes in molecular interactions (localization, sequestering) of GSK3β without affecting its activity, or from dysregulation of another kinase or phosphatase. APC has more than 100 predicted phosphorylation sites with multiple kinases implicated.
To gain insights into potential changes in APC functions, we first tested for alterations in its molecular associations between β-cat cOEs and control littermates, using unbiased, quantitative mass spectrometry proteomic analysis of hippocampal immunoprecipitates. We identified several proteins, in addition to β-cat, that exhibit at least a +/- 0.7 log2 fold-change differential abundance in association with APC in the β-cat cOEs, including direct and indirect binding partners that co-function with APC to regulate the microtubule and actin cytoskeletons, and organize pre- and post-synaptic complexes (Supplementary Fig. 4a). In comparison, mass spectrometry proteomic analysis of changes in β-cat molecular interactions showed only increased association with APC in β-cat cOEs (predominantly truncated β-cat) versus control littermates (predominantly full-length β-cat) in hippocampal immunoprecipitates) (Supplementary Fig. 4b), consistent with the greater affinity of truncated β-cat for APC (Fig. 4a). These results suggest that malfunction of β-cat and APC likely contribute to the β-cat cOE phenotypes.
We then tested for changes in another APC function, its role as an mRNA binding protein; its interactome includes several mRNAs involved in learning29. We focused on two APC mRNA targets that encode proteins with known roles in glutamatergic synapse organization, plasticity and learning, SynCAM1, a synaptic adhesion molecule and direct binding partner of protein 4.1N (Supplementary Fig. 3), and CDC42, a synaptic component that links to and regulates the submembranous actin cytoskeleton48, 49, 56–58. mRNA levels for cadm1 (encoding SynCAM1) and cdc42 were not changed in either β-cat cOEs or APC cKOs, relative to control littermates, as measured by RT-qPCR (Fig. 5a). In contrast, SynCAM1 and CDC42 protein levels are both decreased in hippocampal lysates of β-cat cOEs, compared with littermate controls. APC cKOs show the opposite change, increases in protein levels of both (Fig. 5b). We therefore tested for alterations in mRNA translation in the two mutant mouse models.
Convergent Roles of APC and β-cat in Modulating Selected mRNA Translation
Initially, we assessed global translation at basal state in the hippocampus. We measured total levels of puromycin incorporation into nascent proteins in acute hippocampal slices by performing surface sensing of translation (SUnSET) assays. Compared with control littermates, β-cat cOEs exhibit no difference in overall puromycin incorporation levels, whereas APC cKOs display significant increases (Supplementary Fig. 5). Thus, loss of APC leads to overall elevated levels and dysregulation of hippocampal mRNA translation. We then tested for specifically altered translation of the APC target mRNA cadm1. We immunoprecipitated puromycin, followed by immunoblotting of SynCAM1, using a C-terminal antibody to capture the full-length protein products59. APC cKOs exhibit increased puromycin incorporation into SynCAM1 nascent protein, suggesting augmented translation of cadm1 mRNA in the absence of APC (Fig. 5c). β-cat cOEs display the opposite cadm1 translation change, decreased puromycin incorporation into SynCAM1 nascent protein, indicating reduced translation, relative to control littermates.
APC association with cadm1 mRNA is altered in β-cat cOEs. We immunoprecipitated APC protein in the presence of RNAse inhibitors, followed by RT-qPCR to measure the levels of co-precipitated cadm1 mRNA in hippocampal lysates. We found increased levels of cadm1 mRNA associate with APC in β-cat cOEs, compared with control littermates (Fig. 5d). Taken together, the data suggest that excessive levels of truncated β-cat increase β-cat association with APC and alter APC phosphorylation state, binding to its mRNA targets, and reduces their translation.
To test whether excessive truncated β-cat is sufficient, or APC is also required, to reduce cadm1 mRNA translation, we crossed β-cat cOE mice with APC cKO mice. This depletes APC in neurons overexpressing truncated β-cat and thereby prevents their increased association. In the APC cKO, β-cat cOE double mutant hippocampus, protein levels of SynCAM1 are increased, relative to β-cat cOEs, and are comparable to APC cKO levels (Fig. 6a). Protein levels of CDC42 (another APC target mRNA29) are also increased, relative to β-cat cOEs, and resemble that of APC cKOs. Further, AMPAR GRIA1 levels are higher in APC cKO, β-cat cOE double mutants, compared to β-cat cOEs. Taken together, the results suggest that the increased interaction between APC and truncated β-cat leads to convergent dysregulation that underlies the molecular synaptic changes in β-cat cOEs.
As further support for a role of β-catenin in regulating translation in neurons, we show both truncated and full-length isoforms associate with the pre-initiation complex, as indicated in pull-downs with m7 GTP-agarose beads from the hippocampus of the mutant mouse lines (Fig. 6b). The 7-methylguanylate cap (m7 G) of mRNAs interacts with the eIF4E complex and associated proteins that regulate the initiation of translation.
Aberrant activity-dependent changes in translation in β-cat cOE synaptosomes
Activity-dependent local mRNA translation directs molecular changes required for normal learning and memory. We therefore assessed the levels of activity-dependent local translation near synapses in β-cat cOEs, compared with control littermates. NMDA stimulation increases puromycin incorporation into nascent proteins in synaptosomes acutely isolated from wildtype hippocampus (puromycin incorporation expressed as the ratio of NMDA stimulated to basal levels) (Fig. 6c). The increases are prevented by APV pretreatment to block NMDA receptor activation. In comparison, NMDA stimulation does not increase new protein synthesis in β-cat cOE synaptosomes (Fig. 6c). This finding suggests that the aberrant continuously elevated levels of truncated β-cat disrupt the dynamic, tightly regulated processes that control activity-dependent local translation near synapses.