Thomas Deller

While the accumulation of Aβ peptides, derived by proteolytic cleavage from the amyloid precursor protein APP, constitutes a hallmark of Alzheimer’s disease, the question of whether endogenous, physiological APP processing serves an essential role for normal brain function(s) has remained elusive. Here, we modified the mouse APP locus to delete the endogenous α- and β-secretase sites (APP residues S592-G621) while preserving the transmembrane topology and the APP C-terminus. These novel knockin mice (termed APPΔS622) lack APPsα, APPβ, and Aβ production and show only minimal soluble APPs shedding. These APPΔS622 mice provide a unique model to distinguish between functions mediated by unprocessed cell surface APP involved in synaptic adhesion or by proteolytic fragments. Strikingly, APPΔS622 mice exhibit pronounced impairments in both long-term potentiation (LTP) and homeostatic forms of synaptic plasticity. Extracellular field recordings revealed significant deficits in hippocampal LTP, associated with a reduced density of mature spines and reduced spine head size. LTP deficits could be restored by APPsα application, suggesting a requirement for non-amyloidogenic APP processing for LTP. In addition, APPΔS622 mice lacking endogenous Aβ displayed a lack of homeostatic synaptic scaling in response to network impairment that could be specifically rescued by exogenous Aβ. Importantly, these deficits in different forms of synaptic plasticity also led to impairments in cognitive behavior. In summary, our findings reveal an essential role of physiological APP processing products to differentially regulate complementary forms of synaptic plasticity that are critical for network stability, normal brain function and cognition.

Polyglutamine expansion in Ataxin-2 (ATXN2) is responsible for rare, dominantly inherited Spinocerebellar Ataxia type 2 (SCA2). Together with its paralog Ataxin-2-like (ATXN2L), both proteins have received much interest, since the deletion of their yeast and fly orthologs alleviates TDP-43-triggered neurotoxicity in Amyotrophic Lateral Sclerosis models. Their typical structure across evolution combines LSm with LSm-Associated Domains and a PAM2 motif. To understand the physiological regulation and functions of Ataxin-2 homologs, the phylogenesis of sequences was analyzed. Human ATXN2 harbors multiple alternative start codons, e.g., from an intrinsically disordered sequence (IDR) present since armadillo, or from the polyQ sequence that arose since amphibians, or from the LSm domain since primitive eukaryotes. Multiple smaller isoforms also exist across the C-terminus. Therapeutic knockdown of polyQ expansions in human ATXN2 should selectively target exon 1B. PolyQ repeats developed repeatedly, usually framed and often interrupted by (poly)Pro, originally near PAM2. The LSmAD sequence appeared in algae as the characteristic Ataxin-2 feature with strong conservation. Frequently, Ataxin-2 has added domains, likely due to transcriptional readthrough of neighbor genes during cell stress. These chimerisms show enrichment of rRNA processing; nutrient store mobilization; membrane strengthening via lipid, protein, and glycosylated components; and cell protrusions. Thus, any mutation of Ataxin-2 has complex effects, also affecting membrane resilience.

Polyglutamine expansion in Ataxin-2 (ATXN2) is responsible for rare, dominantly in-herited Spinocerebellar Ataxia type 2 (SCA2). Together with its paralog Ataxin-2-like (ATXN2L), both proteins received much interest since deletion of their yeast and fly orthologs alleviates TDP-43-triggered neurotoxicity in Amyotrophic Lateral Sclerosis models. Their typical structure across evolution combines LSm with LSm-Associated Domains and a PAM2 motif. To understand the physiological regulation and functions of Ataxin-2 homologs, the phylogenesis of sequences was analyzed. Human ATXN2 harbors multiple alternative start codons, e.g. from an intrinsically disordered se-quence (IDR) present since armadillo, or from the polyQ sequence that arose since amphibians, or from the LSm domain since primitive eukaryotes. Multiple smaller isoforms also exist across the C-terminus. Therapeutic knockdown of polyQ expansions in human ATXN2 should selectively target exon 1B. PolyQ repeats developed repeat-edly, usually framed and often interrupted by (poly)Pro, originally near PAM2. The LSmAD sequence appeared in algae as the characteristic Ataxin-2 feature with strong conservation. Frequently, Ataxin-2 has added domains, likely due to transcriptional readthrough of neighbor genes during cell stress. These chimerisms show enrichment of rRNA processing; nutrient store mobilization; membrane strengthening via lipid, protein, and glycosylated components; and cell protrusions. Thus, any mutation of Ataxin-2 has complex effects, also affecting membrane resilience.

The Ataxin-2-like (ATXN2L) protein is required to survive embryonic development, as documented in mice with the constitutive absence of the ATXN2L Lsm, LsmAD, and PAM2 domains due to knock-out (KO) of exons 5–8 with a frameshift. Its less abundant paralog, Ataxin-2 (ATXN2), has an extended N-terminus, where a polyglutamine domain is prone to expansions, mediating vulnerability to the polygenic adult motor neuron disease ALS (Amyotrophic Lateral Sclerosis) or causing the monogenic neurodegenerative processes of Spinocerebellar Ataxia Type 2 (SCA2), depending on larger mutation sizes. Here, we elucidated the physiological function of ATXN2L by deleting the LsmAD and PAM2 motifs via loxP-mediated KO of exons 10–17 with a frameshift. Crossing heterozygous floxed mice with constitutive Cre-deleter animals confirmed embryonic lethality among offspring. Crossing with CamK2a-CreERT2 mice and injecting tamoxifen for conditional deletion achieved chimeric ATXN2L absence in CamK2a-positive frontal cortex neurons and reduced spontaneous horizontal movement. Global proteome profiling of frontal cortex homogenate showed ATXN2L levels decreased to 75% and dysregulations enriched in the alternative splicing pathway. Nuclear proteins with Sm domains are critical to performing splicing; therefore, our data suggest that the Like-Sm (Lsm, LsmAD) domains in ATXN2L serve a role in splice regulation, despite their perinuclear location.

Ataxin-2-like (ATXN2L) protein is required to survive embryonic development, as documented in mice with constitutive absence of ATXN2L Lsm, LsmAD and PAM2 domains, due to knockout (KO) of exons 5-8 with frameshift. Its less abundant paralog Ataxin-2 (ATXN2) has an extended N-terminus, where a polyglutamine domain is prone to expansions, which mediate vulnerability to the polygenic adult motor neuron disease ALS (Amyotrophic Lateral Sclerosis), or cause the monogenic neurodegenerative processes of Spinocerebellar Ataxia type 2 (SCA2), depending on larger mutation sizes. Here, we elucidated the physiological function of ATXN2L by deleting the LsmAD and PAM2 motif, via loxP-mediated KO of exons 10-17 with frameshift. Crossing heterozygous floxed mice with constitutive Cre-deleter animals confirmed embryonic lethality among offspring. Crossing with CamK2a-CreERT2 mice and injecting tamoxifen for conditional deletion achieved chimeric ATXN2L absence in CamK2a-positive frontal cortex neurons and resulted in reductions of spontaneous horizontal movement. Global proteome profiling of frontal cortex homogenate found ATXN2L levels decreased to 75%, and dysregulations enriched in the alternative splicing pathway. Nuclear proteins with Sm domains are critical to perform splicing, so our data suggest that the Like-Sm (Lsm, LsmAD) domains in ATXN2L serve a role in splice regulation, despite its perinuclear location.

Objective: Degeneration of hilar mossy cells in the dentate gyrus is an important hallmark of hippocampal sclerosis and is often observed in patients with temporal lobe epilepsy. To understand the pathogenesis of hippocampal sclerosis and develop novel neuroprotective treatments, it is critical to determine the mechanistic processes of mossy cell degeneration and factors that influence cell vulnerability or resilience. However, suitable in vitro approaches are currently lacking. We have developed and validated an organotypic slice culture‐based in vitro model that facilitates mechanistic studies of activity‐dependent mossy cell vulnerability and resilience. Methods: A model was developed using entorhino‐hippocampal slice cultures. Dentate gyrus granule cells were transduced with adeno‐associated viruses to express channelrhodopsin2. Transduced cultures were chronically stimulated by light, and resulting cell damage was assessed by propidium iodide staining. Spontaneous synaptic activity before and after optical stimulation was recorded using whole‐cell patch‐clamp. Results: Selective and dose‐dependent hilar neuron degeneration was observed following chronic optogenetic stimulation of organotypic slice cultures expressing channelrhodopsin‐2 in granule cells. Treatment with the anticonvulsant retigabine reduced stimulation‐induced hilar neuron loss in a dose‐dependent manner. This demonstrates the suitability of our optogenetic in vitro model for drug screening. Patch‐clamp recordings verified strong synaptic activation of mossy cells during optical stimulation and a reduction in spontaneous excitatory synaptic activity after stimulation. Significance: The role of mossy cells in the context of epileptic seizures has been a controversial topic of discussion. The presented in vitro model allows the study of mossy cell vulnerability on a single‐cell level and provides the first evidence for changes in synaptic activity after stimulation. This model will facilitate our mechanistic understanding of temporal lobe epilepsy, providing a foundation for novel therapeutic interventions aimed at preserving mossy cell function in epilepsy patients.

Organotypic slice cultures (OTCs) are versatile tools for studying long-term structure-function relationships of neurons within a defined network (e.g. hippocampus). We developed a method for repeated experimenter-controlled activation of hippocampal granule cells (GCs) in OTCs within the incubator. After several days of contact-free photonic stimulation, we were able to ameliorate entorhinal denervation-induced structural damage in GCs. To achieve this outcome, we had to calibrate the intensity and duration of optogenetic (light) pulses using whole-cell electrophysiological recordings and multi-cell calcium imaging. Our findings showed that ChR2-expressing cells generated action potentials (APs) or calcium transients in response to illumination but were otherwise functionally indistinguishable from non-transduced GCs within the same neural circuit. However, the threshold for AP firing in single GCs varied based on the stimulus light intensity and the expression levels of ChR2. This information allowed us to calibrate light intensity for chronic stimulations. Denervated GCs exhibited significant spine loss four days post-denervation, but this detrimental effect was mitigated when AP firing was induced at a physiological GC bursting rate. Phototoxic damage caused by chronic light exposure was significantly reduced if illuminated with longer wavelength and by adding antioxidants to the culture medium. Our study presents a versatile approach for concurrent non-invasive manipulation and observation of neural circuit activity and remodeling in vitro.

Denervation of neurons is a network consequence of brain injury. The effects of denervation on neurons can be readily studied in vitro using organotypic slice cultures of entorhinal cortex and hippocampus. Following transection of the entorhino-dentate projection, granule cells (GCs) are denervated and show on average a transient loss of spines on their denervated distal dendrites but not on their non-denervated proximal dendrites. In the present study, we addressed the question how single GCs and their denervated and non-denervated segments react to entorhinal denervation. Local adeno-associated virus (AAV)-injections were employed to transduce dentate GCs with tdTomato and entorhinal projection neurons with EGFP. This made it possible to visualize both innervating entorhinal fibers and their target neurons and to identify dendritic segments located in the “entorhinal” and the “hippocampal” zone of the dentate gyrus. Confocal time-lapse imaging was used to image distal and proximal segments of single GCs after entorhinal denervation. Time-matched non-denervated cultures served as controls. In line with previous reports, average dendritic spine loss was ~30% (2–4 days post-lesion) in the denervated zone. However, individual GCs showed considerable variability in their response to denervation in both layers, and both decreases as well as increases in spine density were observed at the single cell level. Based on the standard deviations and the effect sizes observed in this study, a computer simulation yielded recommendations for the minimum number of neurons that should be analyzed in future studies using the entorhinal in vitro denervation model.

Dendritic spines are sites of synaptic plasticity and their head size correlates with the strength of the corresponding synapse. We recently showed that the distribution of spine head sizes follows a lognormal‐like distribution even after blockage of activity or plasticity induction. As the cytokine tumor necrosis factor (TNF) influences synaptic transmission and constitutive TNF and receptor (TNF‐R)‐deficiencies cause changes in spine head size distributions, we tested whether these genetic alterations disrupt the lognormality of spine head sizes. Furthermore, we distinguished between spines containing the actin‐modulating protein synaptopodin (SP‐positive), which is present in large, strong and stable spines and those lacking it (SP‐negative). Our analysis revealed that neither TNF‐deficiency nor the absence of TNF‐R1, TNF‐R2 or TNF‐R 1 and 2 (TNF‐R1/R2) degrades the general lognormal‐like, skewed distribution of spine head sizes (all spines, SP‐positive spines, SP‐negative spines). However, TNF, TNF‐R1 and TNF‐R2‐deficiency affected the width of the lognormal distribution, and TNF‐R1/2‐deficiency shifted the distribution to the left. Our findings demonstrate the robustness of the lognormal‐like, skewed distribution, which is maintained even in the face of genetic manipulations that alter the distribution of spine head sizes. Our observations are in line with homeostatic adaptation mechanisms of neurons regulating the distribution of spines and their head sizes.

The cytokine tumor necrosis factor (TNF) is involved in the regulation of physiological and pathophysiological processes in the central nervous system. In previous work, we showed that mice lacking constitutive levels of TNF exhibit a reduction in spine density and changes in spine head size distribution of dentate granule cells. Here, we investigated which TNF‐receptor pathway is responsible for this phenotype and analyzed granule cell spine morphology in TNF‐R1‐, TNF‐R2‐, and TNF‐R1/R2‐deficient mice. Single granule cells were filled with Alexa568 in fixed hippocampal brain slices and immunostained for the actin‐modulating protein synaptopodin (SP), a marker for strong and stable spines. An investigator blind to genotype investigated dendritic spines using deconvolved confocal image stacks. Similar to TNF‐deficient mice, TNF‐R1 and TNF‐R2 mutants showed a decrease in the size of small spines (SP−negative) with TNF‐R1/R2‐KO mice exhibiting an additive effect. TNF‐R1 mutants also showed an increase in the size of large spines (SP−positive), mirroring the situation in TNF‐deficient mice. Unlike the TNF‐deficient mouse, none of the TNF‐R mutants exhibited a reduction in their granule cell spine densities. Since TNF tunes the excitability of networks, lack of constitutive TNF reduces network excitation. This may explain why we observed alterations in spine head size distributions in TNF‐ and TNF‐R‐deficient granule cells. The changes in spine density observed in the TNF‐deficient mouse could not be linked to canonical TNF‐R‐signaling. Instead, noncanonical pathways or unknown developmental functions of TNF may cause this phenomenon.