Introduction
The formation and maturation of neuronal circuits in development and consequently processes of memory formation critically depend on the dynamic regulation of synaptic transmission and excitability. Multi-domain scaffolding proteins play essential roles in these processes by assembling and (re-)organizing proteinaceous cytomatrices on both sides of chemical synapses. In the presynaptic cytomatrix at the active zone (CAZ), which organizes the synaptic vesicle cycle underlying regulated neurotransmitter exocytosis, the two closely related proteins Bassoon and Piccolo/Aczonin fulfill such scaffolding functions together with other multi-domain proteins, such as the Rab3-interacting molecules (RIMs), RIM-binding proteins (RBPs), ELKS/CAST proteins, and Munc13s (Fejtova and Gundelfinger
2006; Schoch and Gundelfinger
2006; Südhof
2012; Gundelfinger and Fejtova
2012; Ackermann et al.
2015). Bassoon is a very large CAZ protein (420 kDa) involved in the developmental assembly of the active zones (Shapira et al.
2003; Ziv and Garner
2004; Maas et al.
2012) as well as in processes of presynaptic plasticity (for review see: Gundelfinger and Fejtova
2012; Ivanova et al.
2016). The
Bsn gene has been well-conserved among mammals including humans (Winter et al.
1999). Bassoon expression is significantly up-regulated in the prefrontal cortex of schizophrenic patients (Martins-de-Souza et al.
2009) and the content of the protein in the synaptic proteome is rapidly down-regulated after frequency-modulated tone discrimination learning in mice (Kähne et al.
2012).
The analysis of constitutive
Bsn mutant mice and of primary neuronal cultures derived from these mice has implied multiple critical synaptic functions for the protein. To date, two different
Bsn mutants have been studied:
BsnΔEx4/5, a hypomorphic mutant lacking exons 4 and 5 encoding a large central part of the Bassoon (Altrock et al.
2003), and
Bsngt, a gene trap mutant lacking Bassoon at most synapses in the brain (Hallermann et al.
2010; Frank et al.
2010). Both mutants display similar phenotypes with sensory disturbances and seizures. Bassoon is importantly involved in anchoring synaptic ribbons, i.e., highly specialized forms of the presynaptic cytomatrix, to the active zone. This has been observed at photoreceptor ribbon synapses (Dick et al.
2003; tom Dieck et al.
2005) as well as ribbon synapses of inner ear hair cells (Khimich et al.
2005; Frank et al.
2010) and likely accounts for deficits in visual and auditory sensory processing of Bassoon-mutant animals. Moreover, Bassoon is involved in the localization of presynaptic voltage-gated Ca
2+ channels. This has been observed at inner ear hair cell synapses (Frank et al.
2010; Jing et al.
2013) as well as at excitatory hippocampal synapses, where specifically P/Q-type Ca
2+ channels are mislocalized (Davydova et al.
2014). Bassoon-deficient synapses show a deficit in the replenishment of synaptic vesicles, most obviously at synapses with particularly high firing rates such as the cerebellar mossy fiber synapse or the endbulb of Held in the auditory system (Hallermann et al.
2010; Mendoza Schulz et al.
2014). Bassoon and its paralog Piccolo are involved in the maintenance of synaptic integrity by regulating presynaptic ubiquitination and proteostasis (Waites et al.
2013). The loss of these two proteins triggers degradation of synaptic vesicles and enhances autophagocytic processes in presynaptic terminals. Essentially, the interplay of Bassoon with Atg5, an E3-like ubiquitin ligase crucial for autophagy, is critical for the control of the formation of autophagic structures in presynaptic boutons (Okerlund et al.
2017). Another important function that is shared by Bassoon and Piccolo is the activity-dependent recruitment of the transcriptional suppressor protein CtBP1 to presynapses (Ivanova et al.
2015). In this way they control the distribution of CtBP1 between synapses and nuclei within neurons and in turn the expression of CtBP1-dependent genes (Ivanova et al.
2015,
2016; Gundelfinger et al.
2016). For example, Bassoon-deficient synapses recruit 30–40% less CtBP1 and display an altered synapto-nuclear distribution of the transcriptional repressor (Ivanova et al.
2015).
Another peculiar phenotype of
BsnΔEx4/5 mutant mice is their increase in forebrain volume (Angenstein et al.
2007) that starts to become significant one month after birth and is accompanied by increased levels of brain-derived growth factor (BDNF; Heyden et al.
2011).
BsnΔEx4/5 mice moreover display reduced synaptic fatigue during induction of long-term depression (LTD), spontaneous epileptic seizures (Altrock et al.
2003), impaired long-term potentiation (LTP) at CA1 synapses (Sgobio et al.
2010) and an abnormal synaptic plasticity at various striatal synapses (Ghiglieri et al.
2009). In the hippocampus, these animals further display an increased neurogenesis, reduced apoptosis (Heyden et al.
2011) and disturbance in the development of mossy fiber synapses (Lanore et al.
2010).
These data demonstrate that Bassoon is of critical importance for a variety of synaptic and network functions throughout the central nervous system. However, an in-depth behavioral assessment of the consequences of their disturbance in
BsnΔEx4/5 mice is hampered by their visual and auditory impairment (Dick et al.
2003; Khimich et al.
2005) and the development of epilepsy (Altrock et al.
2003). In initial experiments, an altered performance in a socially transmitted food preference task (Sgobio et al.
2010) and an improved performance in a two-way active avoidance task that could be normalized by a TrkB antagonist (Ghiglieri et al.
2010) were observed. However, the underlying cellular processes are difficult to address due to above-mentioned sensory impairments and potential gain of function effects by the residual Bassoon fragment lacking its central part, i.e., about two-thirds of the entire protein (Altrock et al.
2003). Furthermore, it has to be considered that Bassoon is expressed at both excitatory and inhibitory neurons (Richter et al.
1999).
To address Bassoon functions in different types of neurons and their role for behavioral control, we have begun to dissect its functions genetically. Here, we report the generation and characterization of mice conditionally lacking Bassoon at glutamatergic synapses of the forebrain (hippocampus and neocortex). We characterized these animals behaviorally and addressed the putative physiological and morphological correlates of the observed disturbances. Based on the above-mentioned disturbances of hippocampal functions in constitutive Bsn mutants, we focused our analysis mainly on hippocampus-dependent behavior and memory formation. In fact, our data thus reveal an altered performance of Bsn cKO mice in contextual and spatial discrimination/pattern separation task, associated with increased excitability at the medial perforant path (MPP) to dentate gyrus (DG) synapse and morphological and physiological changes in DG granule cells that are indicative of a reduced maturation of the DG and increased adult neurogenesis in these animals.
Discussion
To genetically dissect Bassoon functions at different types of synapses and to investigate its role in behavioral and network functions, we generated Bsn cKO mice lacking Bassoon at excitatory forebrain synapses. These mice show an increase in contextual fear memory and increased preference for the novel object location in a spatial discrimination/pattern separation task. These behavioral changes are associated with enhanced neurogenesis, an increased expression of markers for immature granule cells in the DG and a concomitant preservation of juvenile neuronal excitability at MPP-DG synapses. As one of the central organizers of the presynaptic CAZ, Bassoon thus seems to control synaptic properties related to specific forms of memory formation in the adult central nervous system, and to be involved in the structural and functional maturation of the DG.
Constitutive
Bsn mutant mice have been generated previously, but the analysis of genuine Bassoon functions in control of behavior has been hampered by occurrence of epileptic seizures and sensory deficits in these animals (Altrock et al.
2003; Dick et al.
2003; Khimich et al.
2005). This caveat has been overcome with the generation of
Bsn cKO mice, which show a loss of Bassoon expression selectively from the VGLUT1-positive terminals of excitatory forebrain neurons. By contrast, the activation of the conditional allele with an
Emx1-Cre driver mouse spared VGAT-positive terminals of inhibitory interneurons and putative neuromodulatory afferences (Gasbarri et al.
1997; Picciotto et al.
2012) bearing residual Bassoon labeling in VGAT- and VGLUT-negative terminals in the molecular layer of the DG (Fig.
2). Although we cannot exclude that susceptibility for seizure generation might be elevated under certain conditions (chemically induced epilepsy models, etc.),
Bsn cKO mice did not show any overt spontaneous seizures, survived well and without any difference from their WT littermates (Online Resource 1e). Furthermore, they performed equally or even superior to their WT littermates in several visual and auditory tasks (Fig.
3,
4 and Online Resource 2,4).
The behavioral analysis of
Bsn cKO mice showed an enhanced contextual fear memory without alterations in auditory cued fear conditioning suggesting a change of hippocampus-dependent, but not amygdala-dependent fear memory processing (Maren et al.
2013). We could not observe any change in shock sensitivity, exploratory or anxiety-related behavior or in response learning in an active avoidance task. On the other hand, preliminary data on improved performance during initial training sessions of a frequency-modulated tone discrimination task (Tischmeyer, Annamneedi et al. unpublished) point to a possible change in auditory cortex-dependent learning of
Bsn cKO mice. In contrast, when subjecting
Bsn cKO mice to the Morris water maze task, we found no evidence for a change in spatial learning or re-learning. We further employed a novel object location task to assess intrinsically motivated one-trial spatial learning. Since both WT and
Bsn cKO performed well in the Morris water maze task we decided to render the novel object location task rather difficult by changing object location by 90 degrees from the original position and testing memory after 24 h. In fact, WT showed relatively poor performance, whereas
Bsn cKO mice showed a tendency to prefer the novel object location in this paradigm. More importantly,
Bsn cKO mice displayed a significantly increased preference than WT mice for the novel location in a related spatial discrimination/pattern separation task. Metric processing of object location can be utilized to examine pattern separation in mice (Bekinschtein et al.
2013; van Hagen et al.
2015). Again, we employed a rather difficult paradigm with a small distance between the targeted objects and WT mice indeed displayed avoidance rather than approach as has previously been observed when animals failed to learn such tasks (Bekinschtein et al.
2013).
Bsn cKO mice on the other hand showed a strong selective approach to the novel location in this learning paradigm. We cannot exclude that the different responding (approach in
Bsn cKO, avoidance in WT) in this task may involve functions other than the ability for spatial discrimination and pattern separation. However, we found no evidence for altered novelty responding or anxiety-related behavior in the
Bsn cKO mice. The converging evidence from contextual fear conditioning and spatial discrimination/pattern separation tasks thus led us to further investigate potential changes in hippocampal physiology in
Bsn cKO mice.
In fact, in hippocampal slice preparations we observed higher baseline excitability and increased fEPSP-to-FV amplitude ratio at MPP-DG synapses that may explain the behavioral observations. These physiological changes are likely to increase granule cell responsiveness to stimulation during the acquisition and/or retrieval of contextual and spatial tasks (Saxe et al.
2006; Deng et al.
2009) and are also in good agreement with a role of the DG in contextual fear memory processing (Lee and Kesner
2004; Kheirbek et al.
2013; Liu et al.
2012). They are also in agreement with the critical role of the DG in spatial discrimination (Hunsaker et al.
2008; Clelland et al.
2009) and pattern separation (Gilbert et al.
1998; Leutgeb et al.
2007) as well as the responsiveness of this function to stimulation of DG function (Bekinschtein et al.
2013).
The observed increases of FV and EPSP at the SC-CA1 synapse, on the other hand, were not accompanied by any change in fEPSP-to-FV ratios. This is indicative of a normalized baseline transmission at this synapse, which might be due to postsynaptic adaptive mechanisms rescuing this effect in SC-CA1 pathway. Nevertheless, increased excitability of the Schaffer collateral pathway may also affect hippocampal information processing during memory tasks.
Strikingly, we did not find any change in LTP in DG or CA1 under any of the applied stimulation protocols, indicating that expression of Bassoon per se is not required for the induction of this form of neural plasticity. These data contrast previous findings of altered LTP at CA1 synapses in
BsnΔEx4/5 mice (Sgobio et al.
2010). Considering the role of Bassoon in regulated neurotransmitter release (Altrock et al.
2003; Gundelfinger et al.
2016) and the importance of the GABAergic system in epilepsy (Treiman
2001), we suggest that absence of Bassoon from inhibitory synapses during network development and/or epileptic seizure activity may contribute to impaired long-term plasticity in constitutive
Bsn mutants. Epilepsy-induced changes, sensory impairments or disturbance of GABAergic interneuron function may also have been responsible for deficits of
BsnΔEx4/5 mice in active avoidance learning (Ghiglieri et al.
2010) that we could not recapitulate with
Bsn cKO mice in the current study. On the other hand, we cannot rule out that blockade of GABAergic transmission by addition of picrotoxin, which is a prerequisite for reliable LTP induction in the MPP-DG (Arima-Yoshida et al.
2011), could have masked a potential genotype difference (Sahay et al.
2011). Clearly, it will be interesting to address the specific role of Bassoon in GABAergic interneuron functions in the future.
High excitability as well as expanded dendrites with reduced spine density are the typical properties of young immature DG granule cells (Ge et al.
2007; Schmidt-Hieber et al.
2004; Spampanato et al.
2012). Similarly, we observed an about twofold increase in dendrite branching and accompanying decrease of spine density (~ twofold reduction) in the DG of
Bsn cKO mice. These might reflect a compensatory response to the increased cellular excitability. However, as these changes occur in a dendritic region more than 60 µm away from the soma it can be expected that the relative weight of inputs to the DG, i.e., perforant path vs. commissural path, may be altered. The processing of information regarding object context or location involves the MPP (Eichenbaum et al.
2007) and DG granule cells are critical components of contextual memory engrams (Liu et al.
2012; Ramirez et al.
2013) with the size of activated DG granule cells ensembles correlating with context memory strength (Stefanelli et al.
2016). Thus, an enhanced excitability and associated structural changes in the medial molecular layer of DG might well account for the altered performance of
Bsn cKO mice in spatial discrimination/pattern separation and contextual fear conditioning tasks. Importantly, comparable morphological changes were not observed in the CA1, indicating that altered neuronal development in the DG of cKO mice might be of particular relevance for their physiological and behavioral phenotype.
In fact, immature granule cells in the rodent DG are involved in the formation of contextual fear memories and in performance in pattern separation tasks (Saxe et al.
2006; Clelland et al.
2009) and Nakashiba et al. (
2012) have suggested that different populations of granule cells in the DG, i.e., young immature and older granule cells mediate different aspects of pattern analysis tasks. Accordingly, mice with elevated adult hippocampal neurogenesis are improved in differentiating the overlapping context representations (Sahay et al.
2011).
Our physiological data suggest that the increased excitability of the adult DG may reflect similar changes in the perforant pathway-to-dentate gyrus circuit in
Bsn cKO mice. In fact, we also observed an altered expression of maturation markers calbindin, calretinin and doublecortin in DG granule cell layer of these animals (Hagihara et al.
2013; Ming and Song
2011; Spampanato et al.
2012). Reduced calbindin staining intensity in granule cells has been associated with increased excitability of DG (Magloczky et al.
1997). The reduced calbindin staining, which was seen in both
Bsn cKO mice (this study) and in
BsnΔEx4/5 mice (Dieni et al.
2015), can occur as a consequence of
Bsn gene ablation independently of epileptiform activity. We also demonstrate a profound increase in the density of calretinin and doublecortin labeled cells in the granule cell layer of
Bsn cKO mice, which may indicate an increased number of DG granule cells in an early postmitotic differentiation stage (von Bohlen und Halbach
2007). Finally, we show an increase of Ki67 positive cells in
Bsn cKO mice, suggestive of an increase in neurogenesis, which has previously also been observed in
BsnΔEx4/5 mice (Heyden et al.
2011).
An induction of neurogenesis through epileptiform activity in
Bsn cKO mice is highly unlikely, since in contrast to
BsnΔEx4/5 mice, they did not display any overt seizure activity or increased lethality. Then, how can the lack of Bassoon, a protein that is expressed in post migratory neurons, enhance neurogenesis and reduce maturation markers in the DG? In the wild-type brain, Bassoon transcripts occur in late embryonic development and highest levels are seen at postnatal day 21, particularly in the pre-granule cell layer and differentiating granule cells of the DG (Zhai et al.
2000). Given the expression onset of
Emx1 at embryonic day 10.5 (Gorski et al.
2002), we can expect that Bassoon is not expressed in excitatory synapses of cKO mice at any stage of development. This might lead to a deficit in axonal maturation and hyperexcitability, as indicated by the increase in FV in the MPP. Noteworthy, stimulation of the entorhinal cortex has previously been shown to stimulate hippocampal neurogenesis (Stone et al.
2011).
Bassoon supports the assembly of presynaptic boutons from preassembled Piccolo-Bassoon transport vesicles (PTVs) (Zhai et al.
2001; Shapira et al.
2003; Dresbach et al.
2006). By interacting with dynein light chain DLC-1, it serves as one cargo adaptor to microtubules and disturbance of these interactions attenuates transport of PTVs in young axons (Fejtova et al.
2009). Thus, Bassoon knockout may lead to a shortage of material for synapse assembly and maturation. In line with this, an interaction of Bassoon with the autophagosome factor Atg5 and requirement for synaptic autophagy has been reported recently (Okerlund et al.
2017). Autophagy is critical for presynaptic development and homeostasis (Vijayan and Verstreken
2017) as well as axonal pruning during development (Song et al.
2008), and the lack of decrease in FV amplitude and DG excitability during the postnatal development of
Bsn cKO mice may well result from a disturbance of these functions. Another, not mutually exclusive scenario considers that the developmental gene expression program is affected in Bassoon-deficient neurons. Bassoon can bind the chromatin-modifying transcriptional co-repressor CtBP1, a well-known neurodevelopmental regulator (Chinnadurai
2007), and thereby control synapto-nuclear shuttling of the repressor (Ivanova et al.
2015). Lack of Bassoon can shift the equilibrium towards a higher CtBP1 concentration in the nucleus and may thus repress transcriptional programs required for controlling axonal excitability. In fact, the increased FV was not restricted to the MPP, but also observed in the SC pathway. However, while SC-CA1 transmission was balanced by synaptic adaptation, an increase in baseline transmission was only observed in the DG, maybe due to the existence of a neurogenic niche and activity-induced enhancement of adult neurogenesis.
Together, our data document the importance of Bassoon in the postnatal development of the perforant path-to-DG circuit and in contextual and spatial learning in the adult. The altered performance of
Bsn cKO mice in contextual fear conditioning and a spatial discrimination/pattern separation task appears to be related to an increased excitability of the DG granule cells, but apparently not to any change in synaptic plasticity. It is important to note that this finding does not merely reflect an overall delay of development as adolescent mice have been reported with normal or even reduced context fear memory (Pattwell et al.
2011; Akers et al.
2012). We observed morphological and physiological alterations as well as changes of expression of differentiation markers in the DG of adult
Bsn cKO mice that resemble characteristics of immature granule cells. Disturbed maturation of the DG has been identified as a critical process in schizophrenia and depression (Hagihara et al.
2013) and future studies will have to address the potential involvement of Bassoon-mediated cellular processes in these psychopathologies.