Analysis of co-isogenic prion protein deficient mice reveals behavioral deficits, learning impairment, and enhanced hippocampal excitability

Background Cellular prion protein (PrPC) is a cell surface GPI-anchored protein, usually known for its role in the pathogenesis of human and animal prionopathies. However, increasing knowledge about the participation of PrPC in prion pathogenesis contrasts with puzzling data regarding its natural physiological role. PrPC is expressed in a number of tissues, including at high levels in the nervous system, especially in neurons and glial cells, and while previous studies have established a neuroprotective role, conflicting evidence for a synaptic function has revealed both reduced and enhanced long-term potentiation, and variable observations on memory, learning, and behavior. Such evidence has been confounded by the absence of an appropriate knock-out mouse model to dissect the biological relevance of PrPC, with some functions recently shown to be misattributed to PrPC due to the presence of genetic artifacts in mouse models. Here we elucidate the role of PrPC in the hippocampal circuitry and its related functions, such as learning and memory, using a recently available strictly co-isogenic Prnp0/0 mouse model (PrnpZH3/ZH3). Results We performed behavioral and operant conditioning tests to evaluate memory and learning capabilities, with results showing decreased motility, impaired operant conditioning learning, and anxiety-related behavior in PrnpZH3/ZH3 animals. We also carried in vivo electrophysiological recordings on CA3-CA1 synapses in living behaving mice and monitored spontaneous neuronal firing and network formation in primary neuronal cultures of PrnpZH3/ZH3 vs wildtype mice. PrPC absence enhanced susceptibility to high-intensity stimulations and kainate-induced seizures. However, long-term potentiation (LTP) was not enhanced in the PrnpZH3/ZH3 hippocampus. In addition, we observed a delay in neuronal maturation and network formation in PrnpZH3/ZH3 cultures. Conclusion Our results demonstrate that PrPC promotes neuronal network formation and connectivity. PrPC mediates synaptic function and protects the synapse from excitotoxic insults. Its deletion may underlie an epileptogenic-susceptible brain that fails to perform highly cognitive-demanding tasks such as associative learning and anxiety-like behaviors. Supplementary Information The online version contains supplementary material available at 10.1186/s12915-021-01203-0.

Conclusion: Our results demonstrate that PrP C promotes neuronal network formation and connectivity. PrP C mediates synaptic function and protects the synapse from excitotoxic insults. Its deletion may underlie an epileptogenic-susceptible brain that fails to perform highly cognitive-demanding tasks such as associative learning and anxiety-like behaviors.
Keywords: Cellular prion protein, Epilepsy, Hippocampus, Behavior, Anxiety, LTP Background Cellular prion protein (PrP C ) is a cell surface glycosylphosphatidylinositol (GPI) anchored protein expressed in several tissues with high levels in the nervous system. It is expressed ubiquitously in the brain, especially in neurons and glial cells [1][2][3][4][5]. PrP C is known for its crucial role in the pathogenesis of human and animal prionopathies [6,7]. In these diseases, PrP C is transformed into a misfolded β-sheet-rich isoform, the infectious prion protein (PrP Sc ) [6]. Increasing knowledge about the participation of PrP C in prion pathogenesis contrasts with puzzling data regarding its natural physiological role/s [8][9][10]. Indeed, this controversy was also strengthened by the absence, until a few years ago, of an appropriate Prnp 0/0 mouse model without PrP C protein, with high breeding capability to dissect biological relevance in specific processes [9,[11][12][13].
PrP C has been previously described as neuroprotective, mainly by using loss-of-function approaches [14][15][16], while in other studies, PrP C overexpression was associated with increased susceptibility to neurotoxicity and cell death [15,[17][18][19]. This might mean that Prnp levels should be constrained to a certain level to develop their natural functions [11,15]. This balance is altered in several injuries and neurodegenerative processes, presenting changes in mRNA and protein expression, for example, in Alzheimer's disease, dementia with Lewy bodies, some tauopathies [20,21], human prionopathies (e.g., sporadic Creutzfeldt Jakob Disease (sCJD) [22]), and multiple sclerosis [23].
As indicated, numerous studies have explored the physiological roles of PrP C in vivo using Prnp 0/0 mice. However, it was later demonstrated that some physiological functions were unfortunately misattributed to PrP C due to genetic artifacts generated during the production of the most commonly used knock-out model, still in use in some laboratories: the Zürich I (Prnp ZH1/ZH1 ) mouse [24]. This mouse displayed a mixed background (B6129: C57BL6/J + 129Sv) and was further backcrossed with C57BL/6 J mice to generate the B6.129 mouse line [19,25], and with FVB mice to generate the FVBN-Prnp ZH1/ZH1 model [25] in order to reduce the 129/Sv-associated genes. However, the generated lines were systematically confounded by the Prnp-linked loci polymorphic region containing numerous 129/Sv-associated "flanking genes" (FG) close to Prnp deletion [26,27]. In fact, after crossing Prnp ZH1/ZH1 mice with C57BL/6 J for more than ten generations to reduce FG, a remnant of ≈ 2-5% of the 129/Sv genome markers still persisted in B6.129 mice [26,28,29]. Genome analysis of these models revealed that the number of FG in the chromosome 2 regions where Prnp is localized is 62, related to different functions (i.e., cancer, depression, anxiety, among others [26]). Thus, we consider that most of the physiological phenotypes attributed to the Prnp absence or overexpression, using these mouse models carrying FG, need to be revaluated and confirmed in other FG-free models. As an example, in previous studies, we and others demonstrated that these FGs masked the real neuroprotective function of PrP C against kainate (KA) administration in vivo [19,[30][31][32]. Although a full description of the FG-associated effects in a null Prnp background is not available, one of these FGs is the signal regulatory protein alpha (Sirpα), an important regulator of several innate immune functions [33]. Although prion disease evolution is not modified in Sirpα 0/0 mice [34], it has been clearly demonstrated that Sirpα is mainly responsible for a phagocytic function previously attributed to PrP C [26,27]. The number of functions misattributed to PrP C increased when a recent study described a substrain-related dependence of Cu(I)-ATPase activity among Prnp 0/0 mice related to the 129/Sv FGs and not PrP C [35].
In the present study, we focused our attention on reexamining some PrP C functions associated with neurotransmission, learning, and behavior, taking advantage of a recently generated Prnp 0/0 mouse model: the Zürich 3 (Prnp ZH3/ZH3 ) [50]. This co-isogenic mouse was generated in a pure C57BL/6 J background using TALEN technology [50] and it is resistant to prion infection [51].
Here we performed a set of behavioral tests to analyze Prnp ZH3/ZH3 mouse activity, learning, and memory capabilities. In addition, basic synaptic functions, KAmediated excitability, and LTP induction were evaluated electrophysiologically in alert behaving mice. Finally, PrP C roles during neuronal differentiation and activity were also evaluated in primary cortical cultures. Results indicate that adult Prnp ZH3/ZH3 mice display reduced activity and anxiety-like behavior. They also fail to acquire different instrumental learning tasks. In addition, our experiments show that hippocampal CA3-CA1 Prnp ZH3/ZH3 synapse fails to induce LTP, most likely due to an exacerbated endogenous excitability, further corroborated in vivo after KA administration. Lastly, our results are sustained by the observed alteration in the expression patterns of several genes associated with neuronal system function and synaptic protein-protein interactions in the Prnp ZH3/ZH3 hippocampus by an RNAseq analysis and its RT-qPCR validation.

Results
The absence of PrP C has been related to deficiencies in behavior, learning, and memory in several mouse models with different results [24,42,52,53]. In order to evaluate the implication of PrP C in systemic behavioral tasks, we took advantage of the new knock-out model, the Prnp ZH3/ZH3 mouse. First, we analyzed the nest building capacity between Prnp ZH3/ZH3 vs. Prnp +/+ mice as an indicator of mouse welfare. In contrast to Schmitz et al., where they used Prnp ZH1/ZH1 mice (reviewed in [54]), results showed a slightly increased but not significantly nest-building capacity in Prnp ZH3/ZH3 compared to controls (Prnp ZH3/ZH3 = 3.81 ± 0.64, n = 7 vs. Prnp +/+ = 3.00 ± 0.41, n = 7; mean ± S.E.M., p = 0.15; Mann-Whitney U non-parametric test) (Additional file 1: Fig. S1), suggesting that the two genotypes had similar welfare conditions [55].
To test for possible deficiencies in motor coordination and balance that could also affect in the results from the Skinner box and the open field tests, we compared the performances of both Prnp +/+ and Prnp ZH3/ZH3 mice in the accelerating rotarod test. After a training session, the mice latency to fall from the rods was tested for five sessions during two consecutive days. In the first day, the Prnp ZH3/ZH3 mice displayed a significantly lower latency just in the first two session compared to the control mice (Additional file 3: Fig. S3a). However, from the third run, their performance was not significantly different (  These results indicate similar motor capacities in both groups, but Prnp ZH3/ZH3 mice needed more trials to reach steady measurements for the same task compared to Prnp +/+ mice, suggesting that the knock-out mice have deficits in acquiring instrumental learning as we also observe with the Skinner test (Fig. 2), but not any motor impairment.
Finally, episodic memory was evaluated with the object recognition test (Additional file 4: Fig. S4). In our experiments, most of the Prnp ZH3/ZH3 mice interacted with the objects for just a few seconds (Additional file 4: Fig. S4cd). Therefore, we ruled out this approach due to Prnp ZH3/ZH3 mouse inactivity, related to the anxiety-like behavior we observed in the open field test. This inactivity led to reduced interactions with the objects that rendered the learning results unreliable. To further support increased anxiety levels in the object recognition test, the fecal bodies left in the arena after the habituation session were counted by the observer once the test subject was removed. Prnp ZH3/ZH3 mice exhibited significant increase in fecal bodies present when compared to wildtype mice (Prnp ZH3/ZH3 = 4.76 ± 0.56 vs. Prnp +/+ = 1.29 ± 0.52. mean ± S.E.M., *** p < 0.001, Mann-Whitney U non-parametric test) (Additional file 4: Fig. S4b). These results also correlate with the thigmotaxis levels measured in the Prnp ZH3/ZH3 mice and indicate that the knock-out mice showed increased emotionality and anxiety compared to their wild-type counterparts. This was in contrast to what was observed using Prnp ZH1/ZH1 mice, where the test could be performed. In the habituation session, knock-out mice showed a significant decrease in rearing episodes (Prnp +/+ = 43.21 ± 10.01 vs. Prnp ZH1/ZH1 = 15.7 ± 2.8, mean ± S.E.M., ***p < 0.001, Mann-Whitney U non-parametric test) (Additional file 4: Fig. S4e). No changes were observed in the training phase (Additional file 4: Fig. S4f), although the Prnp ZH1/ZH1 showed a tendency (p = 0.053) to explore the second object less time ((second − first) / total) compared to the wild-type mice (Prnp +/+ = 0.3 ± 0.15 vs. Prnp ZH1/ZH1 = − 0.17 ± 0.15, mean ± S.E.M.) (Additional file 4: Fig. S4g).
Increased paired-pulse facilitation at high intensities in Prnp ZH3/ZH3 Schaffer collateral pathway PrP C has been described as a regulator of glutamatergic neurotransmission in the hippocampus [19,37]. As an example, PrP C has been shown to inhibit NMDAr containing the NR2D subunit [37], or the GluR6/7 receptor [38], see also [57] for review. Therefore, we analyzed the activation of the well-characterized hippocampal Schaffer collateral pathway (CA3-CA1 synapses). Stimulating and recording electrodes were permanently implanted in the CA3 and CA1 regions, allowing us to record and quantify the evoked fEPSPs in living behaving mice (Fig. 3a). First, we evaluated the putative synaptic facilitation or depression evoked at CA3-CA1 synapses by paired-pulse stimulation of the ipsilateral Schaffer collaterals (Fig. 3b). Paired-pulse stimuli were presented to Prnp +/+ (n = 27) and Prnp ZH3/ZH3 (n = 24) mice at different inter-stimulus intervals (from 10 to 500 ms). As already reported for CA3-CA1 synapses [58], this approach generates a higher fEPSP from the second stimulus (fEPSP2) than from the first (fEPSP1) at short intervals due to presynaptic facilitation. In our experiments, as can be observed in the representative examples of fEPSP1 and fEPSP2 (Fig. 3c), no differences were observed between Prnp +/+ and Prnp ZH3/ZH3 facilitation ( Fig. 3b), suggesting that PrP C does not participate in presynaptic mechanisms related to synaptic facilitation at the least at the selected intensities (2 × Threshold; ≈ 0.2 mA).
To gain insight into the gene expression patterns altered in the Prnp ZH3/ZH3 mice, RNA-seq was performed from the hippocampus region of 8 Prnp +/+ and 8 Prnp ZH3/ZH3 animals. The sequencing data have been deposited at the Gene Expression Omnibus (GEO) with accession code: GSE189691.  Table S1 and  Table S2). According to pathway analysis in Reactome v77, the main alterations related to brain functions were the downregulation of genes associated with the "MECP2 regulates neuronal receptors and channels" (10 genes; padj = 1.06 E-07), the "neuronal system" (28 genes; padj. = 0.002), and "protein-protein interaction at synapses" (10 genes; padj. = 8.65 E−04) (Additional file 7: Fig. S5), all in line with the previous behavioral and electrophysiological findings. Among the dysregulated genes, we validated some by RT-qPCR that could explain the phenotype shown by the Prnp ZH3/ZH3 mice: including the downregulation of glutamate ionotropic receptor NMDA type subunit 2B (Grin2b), the Gamma-Aminobutyric Acid Type A Receptor Subunit Rho2 (Gabrr2), the Potassium Voltage-Gated Channels: Subfamily J Member 2 (Kcnj2) and 6 (Kcnj6), Subfamily A Member 1 (Kcna1), and Subfamily Q member 3 (Kcnq3) (Additional file 7: Fig. S5).
Additionally, we evaluated neuronal damage after KAinduced epilepsy with Fluoro-Jade B (Fig. 5c-h). Prnp ZH3/ZH3 mice showed relevant numbers of labeled cells in the pyramidal layer of the CA1 and CA3 (Fig.  5c-e), while no signal was observed in Prnp +/+ sections ( Fig. 5f-h). Indeed, CTCF analysis of Fluoro-Jade B labeling in the pyramidal layer of the CA1 and CA3 (see "Methods" section for details) demonstrated statistical differences between Prnp ZH3/ZH3 vs wild-type mice: CTCF value for Prnp ZH3/ZH3 = 2893 ± 349.3 vs. wildtype = 380 ± 84.05; mean ± S.E.M., *** p < 0,001; Mann- Whitney U non-parametric test (Fig. 5i). These results corroborated the absence of Prnp to generate an exacerbated synaptic excitability in the hippocampal region that increases susceptibility to electrical and KA-induced seizures, causing neuronal death in the CA1 and CA3 regions of the pyramidal layer of the hippocampus proper.

Discussion
PrP C has been associated with several physiological functions using in vivo approaches; however, the consequences of PrP C deletion in behavior and cognition have not been extensively evaluated [9]. There are some studies about Prnp 0/0 mouse behavior, motor capabilities, and learning performance, but the results are not conclusive, especially after the description of the so-called FG in the background of the Prnp ZH1/ZH1 model that masks specific PrP C roles [26]. For example, concerning KA susceptibility, a clear decrease is observed in mice expressing a lower percentage of 129/Sv-associated polymorphisms (B6.129) compared to the B6129 original strain of the Prnp ZH1/ZH1 mice with a higher percentage of 129/Sv genome [28]. Here we assess the consequences of the absence of PrP C in behavior and neurotransmission using the new strictly co-isogenic mouse model Prnp ZH3/ZH3 [50]. However, another relevant aspect of these studies is the age of the analyzed mice, since physiological differences in the absence of Prnp have been described in association with age for Prnp ZH1/ZH1 [45] or FVB/N-Prnp ZH1/ZH1 [61]. Thus, in our study, we only used and compared results obtained from mice of 3-5 months of age. Concerning nest-building behavior, our results suggest similar behavior in Prnp ZH3/ZH3 and wild-type mice. This contrasts with previously reported results [42], but it has been largely demonstrated that this capacity is dependent on mouse background [62,63].
Our results also reveal that Prnp ZH3/ZH3 mice display reduced activity in the open field test. The reduced rearing exploration and peripheral preference in the arena, and the high defecation rate observed in Prnp ZH3/ZH3 mice, suggest an anxiety-like behavior. Indeed, Prnp ZH3/ZH3 showed higher thigmotaxis than wild-type mice (see [64] for technical details). Our data correlate with those reported by Schmitz et al. [45] and Lobao-Soares et al., both using 3-month-old Prnp ZH1/ZH1 mice, illustrating reduced mobility between the inner and outer regions of the open field and increased defecation in Prnp ZH1/ZH1 mice [65]. However, they are in contrast to Nico et al. [66], where they found no differences, and with Gadotti et al. [52], that showed 10-week-old Prnp ZH1/ZH1 mice displayed increased crossing in open-field test although thigmotaxis changes were not analyzed in the study. Although technical details could also play a role (i.e., handling of the mice, the initial position of the mice in the field, geometry of the field), we think that these discrepancies reinforce the relevance of the homogenous genetic background in our study vs. the others. In fact, the differences we observed also extend to the rotarod test. Our results showed no motor alternations in the Prnp ZH3/ZH3 mice, but they need more sessions to learn the task compared to controls. Previous publications also showed any motor alterations in two different knock-out strains [45,65,67]. However, Nazor et al. observed changes just in mice older than 95 days [61].
In our experiments, Prnp ZH3/ZH3 mice also failed to achieve instrumental learning in the Skinner box. These results are similar to those observed in Prnp ZH1/ZH1 mice (B6.129 background, 2-5% 129/Sv markers, [28]). Striking differences in motility between wild-type and knockout mice were identified using this approach. However, learning capacity based on the ON/OFF paradigm confirmed the operant conditioning deficiencies, a type of associative learning.
Following our results, alterations in locomotor activity and increased latency to initiate exploration were previously reported in other Prnp 0/0 mouse models [42,47]. Anxiety-related behavior [54], depressive tendencies [52], and alterations in spatial memory and learning have also been described [42,45]. In contrast, Bueler et al. reported no alterations in Prnp ZH1/ZH1 behavior [24]. This disparity in results might be explained by the age of the animals used in the studies. Bueler and collaborators performed the test with 2-and 3-month-old mice, which could potentially uncover the behavior impairment as it was reported to be an age-dependent decline in other publications using the Prnp ZH1/ZH1 model [45,68]. Another study also showed no deficits in Prnp 0/0 mouse behavior, where Prnp was conditionally deleted at 12 or 16 months, and the results included no alterations in the Morris water maze or object recognition test [53]. We reported behavioral deficits in Prnp 0/0 mouse models with some discrepancies with the previous works, most likely related to age-related sampling and the mice's background.
Glutamate neurotransmission is, in large part, responsible for cortical signaling, and its impairment has been related to behavioral deficits [69]. PrP C has been described as a regulator of glutamate synapses [13]. Even glutamate inhibition with an NMDA antagonist (MK-801) ameliorates depressive-like behavior in Prnp 0/0 mice [52]. Thus, our next step was to study glutamate connectivity to better understand behavioral alterations. The hippocampal Schaffer collaterals and their implication in operant conditioning, spatial learning, and anxietyrelated behavior were evaluated as a well-defined model of glutamate circuitry [70].
The paired-pulse facilitation test did not show differences between Prnp ZH3/ZH3 and Prnp +/+ animals. Therefore, PrP C deletion did not alter synaptic facilitation, at least in the living behaving mouse model, at the least at relatively low stimulation intensities (Fig. 3b). These results may be explained by the fact that synaptic facilitation is mainly a presynaptic phenomenon [71], and PrP C has been related to postsynaptic neurotransmission mechanisms [37,38]. These differences in pre-and postsynaptic mechanisms would explain the different results collected from Prnp ZH3/ZH3 mice and their littermate controls in paired-pulse ratio and LTP tests.
Nevertheless, Prnp ZH3/ZH3 mice displayed increased synaptic facilitation at high intensities, a fact not observed in controls. This could represent a sort of compensatory phenomenon for their evident LTP deficits. In addition, Prnp ZH3/ZH3 mice presented an increased susceptibility to KA-induced seizures. This epileptogenic phenotype may explain our results on anxiety behavior in Prnp ZH3/ZH3 mice. Comorbid anxiety disorders affect patients with epilepsy [72], and cognitive decline has been described in epileptic animal models [73]. Moreover, increased excitability was previously reported, especially in susceptibility to KA, NMDA, and pentylenetetrazol (PTZ) insults [30,31].
However, contradictory results were published by other groups, who described an elevated epileptic threshold in Prnp 0/0 hippocampal slices treated with bicuculline, zero-magnesium conditions, and PTZ [74], and also normal neurotransmission-associated parameters compared with wild-type mice [25]. Both studies recorded hippocampal slices of the FVB/N-Prnp 0/0 model, a mouse with a triple mixture background (FVB/129Sv/ C57BL6) which carried the FG [26]. Different susceptibility to KA-induced seizures among mice with different backgrounds has been described; even the FVB background has been shown as highly susceptible to epilepsy [75,76]. Our results using Prnp ZH3/ZH3 animals in KA susceptibility were similar to those previously obtained with the other available Prnp 0/0 co-isogenic mouse model, the Edinburgh 129/Ola [28]. Therefore, we postulate that these contradictory results published about the excitability of Prnp 0/0 synapse are likely associated with the mouse backgrounds, the FG effect, and the experimental approach. However, the ZH3 mice results demonstrate that PrP C indeed protects against KAinduced epilepsy.
Our results show that the presentation of HFS protocols causes epileptic seizures in most Prnp ZH3/ZH3 mice but fails to generate significant LTP at the hippocampus. Some controversial results have been published about the implication of PrP C in LTP generation. Different experimental approaches (hippocampal slices or in vivo experiments) and a mixture of mouse models with distinct backgrounds were used, generating non-comparable data (i.e., [19,25,40,42]). Here, using ZH3 mice, we hypothesize that the absence of PrP C ends in LTP induction failure due to exacerbated synaptic excitability, although we cannot rule out putative GABAergic disinhibition. It is convincingly demonstrated that severe epileptic seizures cause neuronal death, which hampers LTP generation. Moreover, non-severe epileptic seizures generate similar molecular and synaptic changes to LTP [77,78]. This suggests that non-severe HFS-induced seizures somehow saturate the postsynaptic terminal, over-activating LTP-induction mechanisms that reduce LTP production capacity by HFS. Additionally, PrP C has also been described as interacting with key elements required for LTP-related mechanisms AMPA or NMDA receptors [37,39,79,80], see also [57] for review.
The presented data show that Prnp 0/0 hippocampal synapse is highly excitable and epileptogenic. Alterations in brain connectivity due to developmental alterations, traumas, or infections contribute to this neuromodulation imbalance [59]. In order to assess whether the epileptic phenotype displayed by Prnp ZH3/ZH3 animals came from neuronal connectivity alterations, we studied bursting and network formation in Prnp ZH3/ZH3 -derived primary neuronal cultures. Relevant PrP C expression in vitro is observed from 4 to 5 DIV [81]. Our results indicate that Prnp ZH3/ZH3 cultures did not mature or connect properly; they displayed asynchronous and very low bursting compared to wild-type cultures. These results suggest that the absence of PrP C causes a delay in neuronal maturation, but more relevantly in neural network formation and function. In fact, the role of PrP C -mediated signals in neuritogenesis has been demonstrated [82][83][84]. However, to our knowledge, this is the first description of network alterations due to the absence of PrP C . However, and as indicated by Benvegnu and coworkers [85], there are gene expression changes during the development of FVB/N Prnp 0/0 and wild-type hippocampus. Hence, changes in ion conductance or channel receptor expression might be involved in this delay. In this line, electrophysiological experiments with biochemical characterization might confirm the basis of the delay in maturation in vitro.
Finally, the gene ontology analysis of the RNAseq of Prnp ZH3/ZH3 and Prnp +/+ mouse hippocampus showed downregulation of genes associated with the neuronal system and protein-protein interaction at synapses, fitting the phenotype we observed in the knock-out mice. The significant downregulation of Gabrr2 and Grin2b in Prnp ZH3/ZH3 mice would produce dysregulation in the excitatory/inhibitory balance, increasing the excitability of the system, as we describe with the KA and the HFS analyses. The alteration of the inhibitory neurotransmission was already shown in Prnp 0/0 models since they were susceptible to suffering from epileptic crises (see references above). PrP C has been widely described as a regulator of glutamatergic neurotransmission and its receptors, as we show here with the Grin2b. Moreover, mutations at Grin2b have recently been related to a rare brain disease, the GRIN2B-related neurodevelopmental disorder that causes intellectual disability, autismspectrum-like behavior, epilepsy, and, sometimes, locomotor deficiencies as well [86,87]. Therefore, the altered expression of Grin2b might contribute to the behavior and learning deficits observed in the Prnp ZH3/ZH3 mice.

Conclusions
In conclusion, our study points that the absence of PrP C impairs neuronal network formation and connectivity, producing enhanced susceptibility to excitotoxicity insults such as HFS and KA exposure. This epileptogenic circuitry seems to impair highly cognitive-demanding functions such as associative learning, and it produces anxiety-like behavior.

Animals
Adult C57BL/6 J mice (Prnp +/+ ) were purchased from Charles River Laboratories (Paris, France). Prnp ZH3/ZH3 mice line was provided by A. Aguzzi (Switzerland) (see [50] for details). Prnp ZH1/ZH1 mice [24] were purchased from the European Mouse Mutant Archive (EMMA, Monterotondo, Italy). A total of 185 adult (3-5 months old) male mice (ZH3: Prnp +/+ = 81 and Prnp ZH3/ZH3 = 84; ZH1: Prnp +/+ = 10 and Prnp ZH1/ZH1 = 10) were used in the present study. In ZH1 mouse experiments, null Prnp ZH1/ZH1 and control mice (Prnp +/+ ) were obtained by crossing heterozygous Prnp +/ZH1 mice to obtain a mixed background (B6.129). It is well described that behavior and neural physiology are different between male and female rodent models due to several hormone and non-hormone-derived reasons [88]. Thus, we used only males in order to establish an equivalent group comparable with previous publications. All experiments were performed following the protocols and guidelines of the Ethical Committee for Animal Experimentation (CEEA) of the University of Barcelona. CEEA of the University of Barcelona approved the protocol for using animals in this study (CEEA approval #276/16 and 141/15). Behavioral and electrophysiological studies were performed following the guidelines of the European Union Council

Immunoblotting
Proteins from brain tissue lysates or primary cortical neurons were extracted using RIPA buffer with protease and phosphatase inhibitor cocktails (Roche). Total lysates were obtained by 30 s of centrifugation at 4°C. The protein concentration of the lysates was quantified using Pierce BCA Protein Assay Kit (Thermo Scientific). Then, 10-50 μg of proteins were loaded to SDS-PAGE gels and transferred and transferred to nitrocellulose membranes for 1 h. Membranes were blocked with Tris-buffered solution with 0.1% tween, 5% skimmed milk, and 2% of FBS for 1 h at room temperature (RT) and incubated with PSD95 (1:1.000, MAB1598; Millipore), PrP (1:500; 6H4; Thermo) or Actin (1:20.000; MAB1501; Millipore) antibodies at 4°C O/N. Following HRP-linked secondary antibody (Dako) incubation for 1 h at RT, membranes were developed with ECL substrate (Thermo).

Nest building
For this test, a total of 14 mice (3 months old) were used (Prnp ZH3/ZH3 = 7 and Prnp +/+ = 7). On the first day of testing, one piece of tissue paper (36 × 12 cm) was placed in the cage to facilitate nest building (Additional file 1: Fig. S1). The presence and the quality of each nest were photo-documented and evaluated the following day according to a modified 5-point scale using the method described by Deacon [89]. Two different blinded researchers evaluated the nest generated by each mouse. Data are presented as the mean ± S.E.M. in (Additional file 1: Fig. S1). The statistical analysis was performed with the Mann-Whitney non-parametric test (GraphPad Prism 8 software).

Open field test
In this test, Prnp ZH1/ZH1 mice were not used since detailed studies were already developed using this model [45,54]. In our experiments, mice (Prnp +/+ = 49 and Prnp ZH3/ZH3 = 49) were placed in a square open field altimeter box (35 × 35 × 25 cm, Cibertec, Madrid, Spain). The field had a grid (16 × 16 cm) of infrared lasers on the XY-axis and one on the Z-axis. Locomotor activity was measured for 15 min in mice with the MUX-XYZ16L software. Mice were placed in the box's periphery for 15 min for two consecutive days, and their behavior was recorded. The first day was considered a training session to reduce mouse anxiety associated with manual handling, and the data analyzed and displayed in the manuscript corresponded to the second session. The system inferred mouse activity by counting laser intersections. For anxiety-related behavior measurement, the center (inner) square of the field (10 × 10 cm) was considered as the central zone and the rest of the square as the peripheral (outer) zone [56] (see Fig. 1a). For quantification and to distinguish motility from exploratory behavior, it was considered that a mouse spent time in one of the regions (center vs. periphery) if it remained in the region at least 3 s. Rearing episodes were considered when the animal stood up for at least 3 s, and immobility episodes if immobile for an additional 3 s. Obtained data were analyzed, and the sum of the crossed X-and the Y-axes are presented together to show total mouse mobility in the experiments. The time spent in the maze periphery zones measures thigmotaxis or wallhugging behavior and indicates anxiety-related behavior [56]. Data are presented as the mean ± S.E.M. The statistical analysis was performed with a T-test or Mann-Whitney U non-parametric test (GraphPad Prism 8 software). The asterisks indicate significant differences: **p < 0.01 and ***p < 0.001. The arena and the walls were cleaned with soap and ethanol between trials to remove olfactory cues between experiments.

Operant conditioning tests
The instrumental learning tests were performed as described in previous studies of our group [90]. Six Skinner boxes were used simultaneously (12.5 × 13.5 × 18.5 cm; MED Associates, St. Albans, VT, USA). Each Skinner box was housed in a sound-attenuating cubicle (90 × 55 × 60 cm) constantly exposed to white noise (± 45 dB) and dim light (Cibertec, S.A, Madrid, Spain). The boxes had a trough to receive food pellets (Noyes formula P; 45 mg; Sandown Scientific, Hampton, UK) by pressing a lever. Before the test, mouse food availability was monitored for 7 days to reduce initial mouse weight to 85%. First, mice (ZH3: Prnp +/+ = 49 and Prnp ZH3/ZH3 = 49; and ZH1: Prnp +/+ = 10 and Prnp ZH1/ZH1 = 10) were trained to press the lever to receive food pellets in a fixed-ratio (1:1) schedule. Seven daily sessions (20 min/ each) were held. The boxes were cleaned with soap and ethanol (30%) between trials. Obtaining ≥ 20 pellets for two consecutive sessions was defined as the criterion to assume the learning criteria achievement. Following this first operant conditioning test, we increased the paradigm complexity to test the mice in a more demanding cognitive task for an additional 10 days. Only animals that met the learning criterion were tested (ZH3: Prnp +/+ = 24 and Prnp ZH3/ZH3 = 20; and ZH1: Prnp +/+ = 8 and Prnp ZH1/ZH1 = 5). The paradigm consisted of light (ON period) and dark periods (OFF period) randomly distributed during the session. The light was provided by a small light bulb located over the lever. During the ON period (20 s), lever presses were reinforced with food pellets at a ratio of 1:1. During the OFF period, lever presses were not rewarded and were penalized by adding ten additional seconds (20 ± 10 s) to the next ON period. The number of lever presses during the different conditioning paradigms was monitored and recorded with the MED-PC program (MED Associates, St. Albans, VT, USA). Statistical analysis was carried out using twoway ANOVA with repeated measures and Bonferroni's multiple comparisons test (GraphPad Prism 8 software). Asterisks indicate significant differences: **p < 0.01; and ***p < 0.001. Data are presented as the mean ± S.E.M. or as a percentage (as indicated in each figure).

Rotarod test
For this test, a total of 15 mice of 4 months were used (Prnp ZH3/ZH3 = 8 and Prnp +/+ = 7). Motor performance was tested using an accelerating rotarod. Mice were pretrained to the task to reach a minimum of 30 s performance at 5 rpms on the 1st day of testing. In each training run, animals were placed on the rods at an initial speed of 5 rpm for 30 s. After that, the testing consisted of 5 consecutive trials with 15-min inter-trial intervals. Each trial consisted of 30 s at 5 rpm followed by 5 rpm increases every 15 s with a cut-off of 5 min. Results are expressed as the mean latency of animals to fall from the rod ± S.E.M. The statistical analysis was performed with the two-way ANOVA + Bonferroni's multiple comparisons test (GraphPad Prism 8 software).

Object recognition test
The object recognition test was performed in a homemade arena (30 × 25 × 20 cm), as described [91]. A total of 23 Prnp +/+ and 24 Prnp ZH3/ZH3 mice were analyzed. Additionally, 6 Prnp ZH1/ZH1 mice were also used with 7 Prnp +/+ counterparts. The test consisted of four phases of 10 min/each. First, animals were habituated to the field without any object (habituation session). One hour later, two identical plastic objects were placed in the center of the arena for the training session. A short-term memory test was performed 2-3 h later by changing one of the objects (see Additional File 4: Fig. S4). ZH1 mouse mobility was expressed as the number of rearing episodes during the habituation session. The arena and the objects were cleaned with soap and 30% ethanol between trials to remove olfactory cues. Mouse behavior was recorded with a video camera placed over the arena, and these recordings were used to measure the exploratory behavior blindly. Sniffing and gently touching the objects were counted as exploratory behavior. To further support increased anxiety levels in the Prnp ZH3/ZH3 mice, fecal bodies left in the maze during the habituation session were counted by the observer once the test subject was removed since it has been demonstrated that highly emotional animals exhibit increased defecation [56]. Statistical analysis was performed with the Mann-Whitney U non-parametric test (GraphPad Prism 8 software). Data are presented as the mean ± S.E.M. or as a percentage (indicated in each figure).

Mouse surgery
A total of 98 adult male (3-5 months) mice were implanted with stimulating and recording electrodes (Prnp +/+ = 49 and Prnp ZH3/ZH3 = 49). Four of them died during surgery, and 33 mice were excluded because of the inability to obtain reliable and clean recordings. Thus, the experiments were performed with 61 mice (Prnp +/+ = 31 and Prnp ZH3/ZH3 = 30). Surgery was performed as described in [19,92]. Mice were deeply anesthetized with ketamine (35 mg/kg) and xylazine (2 mg/ kg), and electrodes were aimed at the right dorsal hippocampus. Two recording electrodes were implanted in the stratum radiatum of the CA1 area (2.2 mm caudal to Bregma, 1.2 mm lateral, and 1.3 mm ventral), and two stimulating electrodes were implanted in the Schaffer collateral pathway of the CA3 region (1.5 mm posterior to Bregma, 2 mm lateral, and 1.3 mm ventral). Electrodes were made of 50 μm Teflon-coated tungsten wires (Advent Research, Eynsham, UK). Electrode localizations were checked according to the field excitatory postsynaptic potential (fEPSP) profile evoked by a single stimulation. A silver wire was fixed to the skull as ground. All the wires were soldered to a six-pin socket (RS Amidata, Madrid, Spain) fixed to the skull with dental cement. Recordings were started at a minimum of 1 week after the surgery.

Electrophysiology recordings
Animals were consecutively recorded in groups of six individuals since they reach the total number of animals used in each experiment. Each animal was placed in a small plastic cubicle (5 × 5 × 10 cm) inside a large Faraday box (30 × 30 × 20 cm). fEPSPs were recorded with a high impedance probe (2 × 10 12 Ω, 10 pF) using differential amplification at a bandwidth of 0.1 Hz-10 kHz (P511, Grass-Telefactor, West Warwick, RI, USA). For each experiment, artefactual recordings were discarded. The stimulation intensity threshold of each animal was set with paired-pulse stimulations at 40 ms of interstimulus interval. The stimulus intensity was set to 40-60% of the amount necessary to evoke a suturing response. These intensity values were used for all the experiments.

Paired-pulse stimulation
For synaptic facilitation experiments, 51 mice (Prnp +/+ = 27 and Prnp ZH3/ZH3 = 24) were stimulated at Schaffer collaterals with a pair of pulses at different interstimulus intervals (10,20,40,100,200, and 500 ms) at 2 × threshold intensities (≈ 0.2 mA). Threshold values were previously defined for each mouse. As classically defined, threshold values were determined as the intensity evoking fEPSP responses in 50% of the cases. For all the inter-pulse intervals, the stimulations were repeated ten times. Data are represented as the mean percentage increases of fEPSP2 from fEPSP1 recordings (fEPSP2 / fEPSP1 × 100) ± S.E.M.

Input/output curves
Schaffer collaterals of 29 mice (Prnp +/+ = 14 and Prnp ZH3/ZH3 = 15) were stimulated with paired pulses at 20 increasing intensities (from 0.02 to 0.4 mA, increased in steps of 0.02 mA) at 40 ms of inter-stimulus interval. For all the selected intensities, the stimulations were repeated ten times. Data are represented as the mean of fEPSP slopes (V/s) ± S.E.M. The same data are presented as the mean of paired-pulsed ratio (PP ratio) ± S.E.M. PP ratio is the percentage of the increase of the fEPSP2 from fEPSP1 recordings (fEPSP2 / fEPSP1 × 100). The area under the curve (AUC) was calculated from the PP ratio of all the animals using GraphPad Prism 8 software. Statistical analysis was performed using the Mann-Whitney-Wilcoxon non-parametric test or two-way ANOVA + Bonferroni's multiple comparisons test (GraphPad Prism 8 software). The asterisks indicate significant differences: *p < 0.05; **p < 0.01; and ***p < 0.001 in the figure.  Fig. 4a,b). Next, LTP was evoked with a highfrequency stimulation (HFS) protocol. HFS consisted of five trains of pulses at a rate of 1/s (200 Hz, 100 ms) with the same intensity as the baseline recording. The HFS was repeated six times at intervals of 1 min. After the HFS protocol, fEPSPs were recorded, as for baseline, for 1 h. The following 4 days, the recordings were repeated for 30 min. fEPSPs and 1 V rectangular pulses corresponding to stimulus presentations were saved on a PC using an analog/digital converter (CED 1401 Plus, Cambridge, England). Data were analyzed offline using Spike2 and Signal 5.04 software with homemade representation programs [58]. Collected recordings were represented and analysed offline with the help of commercial (Spike 2 and Signal 5.04) programs following procedures described elsewhere. The slope of collected fEPSPs was computed as its first derivative (volts/ s). Five successive fEPSs were averaged and the mean value of the slope was determined. Data are presented as the mean of the percentage compared to the baseline ± S.E.M. The statistical analysis was performed using twoway ANOVA + Bonferroni's multiple comparisons test (GraphPad Prism 8 software). The asterisks and symbols indicate significant differences: *p < 0.05; **p < 0.01; and ***p < 0.001; ## p < 0.01; and ### p < 0.001.

KA-induced epilepsy and seizure analysis
Adult (3-4 months old) male mice were used for these sets of experiments (Prnp +/+ = 18 and Prnp ZH3/ZH3 = 20) essentially as described in [28]. A KA (Sigma-Aldrich, Darmstadt, Germany) solution was freshly prepared for each experiment in 0.1 M phosphate buffer. Mice were injected with KA (10 mg/kg b.w.) three times: at 0 min, 30 min, and 60 min. After the first injection, mice were placed in clean boxes (1-3 mice/box). The presence of epileptic seizures was monitored in situ and recorded with a video camera for 3 h after drug administration. Seizure severity was scored in grades following the following criteria: grade I-II: hypoactivity and immobility; grade III-IV: hyperactivity and scratching; grade V: loss of balance control and intermittent convulsions; grade VI: continuous seizures and bouncing activity (also reported as blinking episodes or "pop-corn" behavior). Data are presented as the mean ± S.E.M. or as a percentage (as indicated in each figure). Statistical analysis was performed with the Mann-Whitney U non-parametric test (GraphPad Prism 8 software). The asterisk indicates significant differences: *p < 0.05 in the figure.

RNAseq
Hippocampi were extracted, flash frozen on dry ice, and RNA was harvested using RNAEasy Mini kit (Qiagen). Libraries were prepared using the TruSeq Stranded mRNA Sample Prep Kit v2 according to the manufacturer's protocol. Briefly, 500 ng of total RNA was used for poly(A)-mRNA selection using Oligo (dT) magnetic beads and subsequently fragmented to approximately 300 bp. cDNA was synthesized using reverse transcriptase (SuperScript II, Invitrogen) and random primers. The second strand of the cDNA incorporated dUTP in place of dTTP. Double-stranded DNA was further used for library preparation. dsDNA was subjected to Atailing and ligation of the barcoded Truseq adapters. All purification steps were performed using AMPure XP Beads. Library amplification was performed with PCR using the primer cocktail supplied in the kit. Final libraries were analyzed using Agilent DNA 1000 chip to estimate the quantity, check the size distribution, and then quantified by qPCR using the KAPA Library Quantification Kit (KapaBiosystems, Merck, Darmstadt, Germany) before amplification Illumina's cBot. Libraries were sequenced 1 × 50 bp on Illumina's HiSeq 2500.
The quality of the fast files was checked using the FastQC software (http://www.bioinformatics.babraham. ac.uk/projects/fastqc/). An estimation of ribosomal RNA in the raw data was obtained with riboPicker [93]. Reads were aligned with the STAR mapper [94] to release M14 of the Mus musculus Gencode version of the genome (GRMm38/mm10 assembly) (https://www.gencodegenes. org/mouse/release_M14.html). A raw count of reads per gene was also obtained with STAR (-quantMode Tran-scriptomeSAM GeneCounts option). The R/Bioconductor package DESeq2 [95,96] was used to assess differential expression between experimental groups (Wald statistical test + false discovery rate correction). Prior to processing the differential expression analysis, genes for which the sum of raw counts across all samples was less than two were discarded. Deregulated genes with a padj < 0.05 were used to disclose relevant pathway alterations in the REACTOME v77 pathway database gene expression. The analysis has been done just with the protein-coding genes. The gene difference was considered biologically relevant if they are upregulated or downregulated with a fold change of > 1.2 or < 0.85, respectively. A pathway was considered relevant if it was related to neuronal and/or cerebral functions, showed significance (padj < 0.05) and contained more than 10 deregulated genes. The sequencing data have been deposited at the Gene Expression Omnibus (GEO) with accession code: GSE189691 (Matamoros-Angles, A; Hervera, A; Soriano, J; Martí, E; Carulla, P, Llorens, F; Nuvolone, M; Aguzzi, A; Ferrer I; Gruart, A; Delgado-García, JM; Del Río, JA. RNA sequencing of hippocampus of Prnp+/+ and PrnpZH3/ZH3 animals. https:// identifiers.org/geo:GSE189691).

RT-qPCR
For RT-qPCR validations, cDNA was obtained with a High-Capacity cDNA Reverse Transcription kit (Applied Biosystems) following the supplier's instructions. RT-qPCR reactions contained 4.5 μL cDNA and mixed with 0.5 μL 20X TaqMan Gene Expression Assays and 5 μL of 2X TaqMan Universal PCR Master Mix (Applied Biosystems) for a final volume of 10 μL. TaqMan probes used were as follows: (Applied Biosystems). Actb Mm02619580_g1 and Aars Mm00507627_m1 were used as endogenous controls. The assay was performed using technical duplicates per sample in 384-well optical plates with ABI Prism 7900 Sequence Detection system (Applied Biosystems, Life Technologies) following the supplier's parameters: 50°C for 2 min, 95°C for 10 min, and 40 cycles of 95°C for 15 s and 60°C for 1 min. The Sequence Detection Software (SDS version 2.2.2, Applied Biosystems) was used for data processing. It was further analyzed with the ΔΔCt method, which consists of obtaining ΔCt by normalizing each target gene to its endogenous control, followed by subtracting the mean-ΔCt of the control group samples to each ΔCt to obtain ΔΔCt values, and finally using these ΔΔCt values as the negative exponent with base 2, thereby obtaining fold change per sample.

Fluoro-Jade B staining
Mice were perfused 7 days after the KA administration with 4% paraformaldehyde (PFA) dissolved in 0.1 M phosphate buffer (pH = 7.3-7.4). Brains were dissected and postfixed overnight with the same fixative solution. The following day, they were cryoprotected in 0.1 M phosphate buffer containing 30% sucrose (w/v). After freezing in dry ice, 50-μm-thick coronal sections were obtained with a freezing microtome (Leica, Wetzlar, Germany). Sections containing the dorsal hippocampus were selected and rinsed in 0.1 M phosphate buffer and mounted on gelatin-coated slides. The slides containing sections were dried at 37°C overnight. The following day, they were heated at 50°C for 45 min before staining to improve adhesion. The staining started with pretreatment for 3 min in absolute alcohol, followed by 1 min in 70% ethanol and 1 min in deionized water. After that, they were oxidized in a solution of 0.06% KMnO 4 for 15 min. Following three rinses (2 min/each) in deionized water, they were incubated in a solution of 0.001% Fluoro-Jade B (Chemicon, Temecula, CA, USA) containing 0.05% of DAPI in 0.1% acetic acid for 30 min. Finally, sections were rinsed in deionized water (3 min), dehydrated with ethanol, cleared with xylene, and coverslipped with Eukitt TM (Merck, Darmstadt, Germany). Considering that wild-type mice did not displayed Fluoro-Jade B labeled cells in the CA1-CA3 regions after KA treatments and only a very few background could be seen at high magnification and long exposures times (> 500 ms), the Fluoro-Jade B fluorescence in the pyramidal layer of the dorsal hippocampal region (4 sections of each mouse, n = 4 mice per genotype) was photodocumented using an Olympus BX61 epifluorescence microscope equipped with a cooled DP12L camera (Hamburg, Germany). Photomicrographs were obtained using a × 40 objective with identical time exposure between preparations from each wild-type and respective knock-out mouse. No modifications were applied to the obtained pictures. Changes in Fluoro-Jade B labeling were determined by analyzing the corrected total cell fluorescence (CTCF) values (see Matamoros-Angles et al. [97], for details) in the pyramidal layer of hippocampal CA1-3 regions of four mice of each genotype, taking a region of interest of 200 × 100 μm centered in the pyramidal layer, and 4-5 different sections per animals were analyzed and averaged. Data were expressed as mean ± S.E.M. The statistical analysis of the obtained data was performed using Mann-Whitney U nonparametric test in GraphPad Prism 8 software. A value of ***p < 0.001 was considered statistically significant in the CTCF analysis.
Primary cortical cultures of Prnp ZH3/ZH3 and wild-type mice Primary cortical cultures were fashioned from E16.5-E17.5 Prnp +/+ and Prnp ZH3/ZH3 mouse embryos, as explained elsewhere [98]. Brains were removed from the skull and rinsed in cold Hank's balanced salt solution (HBSS) containing glucose (6.5 mg/ml). The meninges were removed, and the cortical lobes were isolated. Tissue pieces were treated with trypsin for 15 min at 37°C. After the addition of horse serum followed by centrifugation, cells were isolated mechanically with a polished glass pipette after treatment with 0.025% DNAse for 10 min at 37°C. One million cells were plated on a 35-mm diameter glass-bottom gridded culture dish (Ibidi, Martinsried, Germany) previously coated with poly-D-lysine (Sigma-Aldrich). Neurobasal TM medium supplemented with 6.5 mg/ ml glucose, 2 mM glutamine, penicillin/streptomycin, 5% of horse serum, and B27 was used as a culture medium (all from Invitrogen-Thermo Fisher Scientific, MA, USA). As Prnp 0/0 -derived cells are sensitive to serum removal [99], after 24 h, the serum was reduced to 2.5%. The medium was changed every 2 days. Horse serum was entirely removed on the eighth day of culture.

Calcium imaging in neuronal culture
Primary cortical neurons were infected 24 h after seeding with AAV9-Synapsin-GCaMP6f [60] (Watertown, MA, USA). In our cultures, the genetically encoded calcium indicators started to express 3-4 days after infection. Calcium changes in GCaMP6f-expressing neurons were recorded at 8, 11, 13, and 15 days in vitro (DIV) using an Olympus IX71 inverted microscope (Olympus, Hamburg, Germany), equipped with an ORCA-Flash 4.0 camera (Hamamatsu Photonics, Japan). During recording, the cells were maintained in a microscope stage incubator at 5% CO 2 and 37°C (Okolab S.R.L., Italy). The same region of the culture was recorded throughout the days following the culture dish grid references. Images (1024 × 1024 pixels) were captured using a × 20 objective and 470 nm wavelength (CoolLED's pE-300 white , Delta Optics, Madrid, Spain) every 100 ms for 8-10 min using the CellSens TM software (Olympus) or the Micro-Manager Open Source Microscopy Software (https://micromanager.org). Exposure levels and frequency were maintained between cultures and evaluation days. GCaMP6f activity was measured in four different identified squares of each culture dish during these 4 days.

Neuronal activity traces, spike events, and network bursts
The recordings were analyzed offline using two MATLAB TM toolboxes: NETCAL (www.itsnetcal.com) [100,101] and NeuroCa [102]. In NeuroCa, an automatic analysis was performed afterward to corroborate obtained NETCAL results. Using NETCAL, a highly contrasted image of the recording's average fluorescence was created, and regions of interest (ROIs) were automatically detected as those objects with a circular shape whose brightness was over a preset threshold. NETCAL and NeuroCa software-rendered a similar number of ROIs and calcium traces. About 400 ROIs, uniformly covering the field of view, were typically identified per recording. The average fluorescence F i (t) in each ROI i along the recording was then extracted, corrected from global drifts and artifacts, and finally normalized as (F i (t) − F (0,i) ) / F (0,i) = f i (t), where F 0,i is the background fluorescence of the ROI. The time series of f i (t) was analyzed with NETCAL to infer neuronal activation timing using the Schmitt trigger method. Our analysis used + 2 S.E.M. of the baseline noise as the high threshold, + 1.5 S.E.M. as the low threshold, and 200 ms as the minimum event length. Calcium traces were calculated, and raster plots of network activity were then constructed by representing the trains of detected neuronal spikes over time. Next, network bursts were analyzed to quantify the ability of the neuronal networks to exhibit collective dynamics, i.e., the collective activation of a group of neurons in a short time window. Bursts were investigated using two approaches. In the first approach, raster plots of spike events were scanned to detect collective occurrences in which at least 5% of the neurons in the network fired synchronously within a 500-ms window. This threshold of 5% was set to disregard random activations that coincided in time. In the second approach, the fluorescence time series of all neurons in the network were averaged. The resulting trace was analyzed with the Schmitt trigger method to detect sufficiently strong fluorescence peaks associated with bursting episodes. Both approaches produced consistent results. Although the detected bursts contained a different number of participating neurons, this information was disregarded in the present analysis and treated later. The total number of detected network bursts divided by the recording duration reflected the culture's activity and was indicated as bursts/min.
The fraction of active neurons in the network was calculated as follows. All detected ROIs were assigned as neurons. After inferring the spike trains, those neurons exhibiting at least two spikes along the recording were considered active, and their number N A was set as a proxy of the healthy population in the neuronal network. The average fraction of active neurons in each condition was then determined as f = N A /N T , where N T is the total number of detected ROIs.
At least ten videos of each genotype/day from different culture plates were consecutively analyzed. Data are presented as the mean of network burst/min ± S.E.M. Statistical analysis was performed using two-way ANOVA + Bonferroni's multiple comparisons test (GraphPad Prism 8 software). Asterisks indicate significant differences between Prnp +/+ and Prnp ZH3/ZH3 cultures at a given DIV: *p < 0.05 and ***p < 0.001. The # symbols indicate significant differences between a given DIV with the initial value at 8 DIV: ### p < 0.001.

Statistical analysis
All statistical analysis was performed with GraphPad PRISM 8 (GraphPad Software, USA). Unless otherwise stated, data is plotted as the mean ± SEM. All experiments were performed three times unless specified. Normality of the distributions was checked via Shapiro-Wilk test. All tests performed were two-sided, and adjustments for multiple comparisons and/or significantly different variances (Fisher's F) applied were indicated. All data analysis was performed blind to the experimental group by two independent experimenters. Unless otherwise stated, sample size was chosen in order to ensure a power of at least 0.8, with a type I error threshold of 0.05, in view of the minimum effect size that was expected.