Lipoprotein receptor loss in forebrain radial glia results in neurological deficits and severe seizures

The Alzheimer disease‐associated multifunctional low‐density lipoprotein receptor‐related protein‐1 is expressed in the brain. Recent studies uncovered a role of this receptor for the appropriate functioning of neural stem cells, oligodendrocytes, and neurons. The constitutive knock‐out (KO) of the receptor is embryonically lethal. To unravel the receptors' role in the developing brain we generated a mouse mutant by specifically targeting radial glia stem cells of the dorsal telencephalon. The low‐density lipoprotein receptor‐related protein‐1 lineage‐restricted KO female and male mice, in contrast to available models, developed a severe neurological phenotype with generalized seizures during early postnatal development. The mechanism leading to a buildup of hyperexcitability and emergence of seizures was traced to a failure in adequate astrocyte development and deteriorated postsynaptic density integrity. The detected impairments in the astrocytic lineage: precocious maturation, reactive gliosis, abolished tissue plasminogen activator uptake, and loss of functionality emphasize the importance of this glial cell type for synaptic signaling in the developing brain. Together, the obtained results highlight the relevance of astrocytic low‐density lipoprotein receptor‐related protein‐1 for glutamatergic signaling in the context of neuron–glia interactions and stage this receptor as a contributing factor for epilepsy.

context of neuron-glia interactions and stage this receptor as a contributing factor for epilepsy.
In the central nervous system, the role of Lrp1 in neurons is well studied in vitro and in the adult brain. Here, Lrp1 is crucial for amyloidbeta clearance and catabolism: altered interactions of Lrp1 with neuronal amyloid-beta contribute to Alzheimer's disease pathogenesis (Spuch, Ortolano, & Navarro, 2012). Neuronal Lrp1 is also vital for the establishment of proper synaptic responses, including long-term potentiation (LTP) (May et al., 2004;Zhuo et al., 2000) and the regulation of calcium influx through the N-Methyl-D-aspartate receptor (NMDAR) (Qiu, Strickland, Hyman, & Rebeck, 2002). Loss of neuronal Lrp1 furthermore impacts the integrity of postsynaptic densities and leads to hyperexcitability in neuronal conditional Lrp1 mutants (Maier et al., 2013;May et al., 2004;Nakajima et al., 2013;Qiu et al., 2002).
Recent research suggests that glial Lrp1 is of equal importance for effective brain functioning: astrocytic Lrp1 mediates the uptake and metabolism of amyloid-beta, thereby becoming an essential target for Alzheimer's disease research (B. Liu, Teschemacher, & Kasparov, 2017). Astrocyte Lrp1-mediated uptake and recycling of tissue plasminogen activator (tPA) is vital for preventing NMDARmediated neurotoxicity (Cassé et al., 2012). Furthermore, Lrp1 is involved in myelin phagocytosis and oligodendrocytic precursor cell differentiation and maturation (Hennen et al., 2013;Lin, Mironova, Shrager, & Giger, 2017;Safina et al., 2016). Despite progress in understanding the function of neuronal and glial Lrp1, knowledge about the role of Lrp1 in the developing brain is scarce. A hint that Lrp1 may be equally crucial in early brain development, as it is in mature circuitry, came from a study showing that the constitutive knock-out (KO) of the Lrp1 gene in mice is lethal for the embryo (Herz, Clouthier, & Hammer, 1992).
In our recent work, we unraveled that the elimination of Lrp1 in cortical neural stem precursor cells (NSPCs) in vitro leads to their altered differentiation properties including an impaired neurogenesis and reduced oligodendrocyte numbers, whereas differentiation toward astrocytes is enhanced (Hennen et al., 2013;Safina et al., 2016;Schafer et al., 2019). This raises the until now unanswered key question concerning the specific impact of Lrp1 deletion from the radial glia stem cell compartment in the living animal.
In the current study, we present a conditional KO mouse line that lacks the Lrp1gene in radial glia cells and their progeny in the dorsal telencephalon. In order to generate these mice, we crossed Lrp1-floxed mice with a reporter line, where the Cre expression is under the control of Empty Spiracles Homebox 1 (Emx1) promoter (Iwasato et al., 2000). The expression of Emx1 in the mouse starts to become visible at embryonic day 9.5 (E9.5) and is confined to the dorsal telencephalon that gives rise to the cerebral cortex, olfactory bulbs and hippocampi. During further brain development, the Emx1 gene expression is found in radial glia and their progeny: postmitotic projection neurons of the dorsal, lateral and medium pallium that form the neocortex, piriform cortex, and hippocampi respectively as well as astrocytes and oligodendrocytes of the pallial corpus callosum and fimbria (Cecchi & Boncinelli, 2000;Gorski et al., 2002;Guo et al., 2000). Emx1 lineage cells do not give rise to GABAergic (inhibitory) interneurons and subpallial oligodendrocytes, but can contribute to the generation of excitatory cells of the amygdala and medium spiny neurons (Cocas et al., 2009;Gorski et al., 2002;Guo et al., 2000).
The Lrp1 KO mice, as described in our current work, exhibit pronounced changes in brain function, including epileptic seizures, ataxia, and an increased lateral ventricular volume. With our study, we underline a vital biological function of Lrp1 expression in the progeny of radial glia cells during early postnatal development and highlight the importance of astrocytic functionality in the establishment of appropriate synaptic connectivity. Taken together, we propose Lrp1 as a gene of interest for deciphering the mechanisms underlying the cause of seizures.

| Ethics statement
The present study was carried out in accordance with the European Communities Council Directive of September 22nd, 2010 (2010/63/ EU) for care of laboratory animals and approved by a local ethics committee (Bezirksamt Arnsberg) and the animal care committee of North Rhine-Westphalia, Germany, based at the LANUV (Landesamt für Umweltschutz, Naturschutz und Verbraucherschutz, Nordrhein-Westfalen, D-45659 Recklinghausen, Germany). The study was supervised by the animal welfare commissioner of Ruhr-University. All efforts were made to minimize the number of mice used for this study.
F1 Lrp1 flox/wt Emx1 Cre/wt mice were subsequently crossed to Lrp1 flox/ flox mice, F2 Lrp1 flox/flox Emx1 Cre/wt (25% of the offspring) exhibited telencephalic Lrp1 KO (see Figure S1 for mating scheme). Cre-positive Lrp1 homozygous conditional KO mice were compared to Crenegative homozygous and heterozygous littermates in all the experiments. Mouse weight was measured starting from P12 and ending with P70 for mice that were still vital. The day of detection of a vaginal plug was taken as E0.5. For experiments, both male and female mice were used. Loss of Lrp1 expression was confirmed for all the analyzed animals.
For PCR products, detecting transgenes and the transgene orientation please refer to Supporting Information Figure S2A,B.

| Kaplan-Meier curve
For the generation of the Kaplan-Meier curve, survival of 50 KO and 48 wildtype (WT) mice, both male and female, was monitored. The plot was generated using the lifelines library (version 0.23.9), pandas library (version 1.0.0), matplotlib (version 3.1.2), and the Python programming language (version 3.6.8).

| Video monitoring
Three P15 KO mice (1 female, 2 males) and six P15 WT mice (3 females and 3 males) were monitored in a test cage for the occurrence of spontaneous epileptic seizures. After 2 hr of acclimation, the frequency and duration of epileptic attacks over a 1-2 week long period were video recorded with Ethovision software (Noldus, Kerpen, Germany).

| Magnetic resonance imaging
For magnetic resonance imaging (MRI), the animals were housed in groups in a temperature and humidity controlled vivarium (Scantainer Ventilated Cabinets, Scanbur A/S, Karlslunde, Denmark) with constant 12 hr light/dark cycle (lights on from 7 a.m. to 7 p.m.). The animals had ad libitum access to food and water.
Six Lrp1 KO and nine WT mice (2 months old, both male and female) were used for the experiments. MRI images were acquired by means of a 7T horizontal bore scanner (BioSpec, 70/30 USR, Bruker, Ettlingen, Germany), using a 8.5 cm inner diameter transmit volume resonator and a planar single-loop 20 mm receiver coil. MR was conducted using the ParaVision 5.1 software (Bruker). Animals were anaesthetized using isoflurane during recordings. The respiration rate and body temperature of the animals were continuously monitored during MR scanning. 3D structural measurements were acquired at an isotropic resolution of 100 μm using a T2 weighted 3D RARE pulse sequence with the following imaging parameters: repetition time (TR) = 2,300 ms; effective echo time (TE) = 62.5 ms; RARE factor = 10; field of view (FOV) = 20 × 14 × 8 cm.
Analysis of ventricular volume was conducted using the semiautomatic segmentation function of the software ITK-SNAP 3.6 (www. itksnap.org; Yushkevich et al., 2006). Hippocampal volume was analyzed using manual segmentation with ITK-SNAP 3.6. Eleven continu-

| Electroencephalogram analysis
The EEG telemetry and video recordings were conducted as previously described (Bedner et al., 2015).
n being the normalization factor and hc(t) c(0)i the autocorrelation function of the EEG data series c(t). The analysis utilized zero-padding and a window function of the Hann-type.  -clamp mode using an amplifier (EPC10 USB, HEKA Electronic, Lambrecht/Pfalz, Germany). The data were subjected to low-pass filtering at 2.9 kHz and digitized at 10 kHz. The FITMASTER software (HEKA Electronic) was used for offline analysis.
During the experiments, the cells were allowed to stay at their own resting potential. Resting membrane potential was determined as the mean value recorded during a continuous period of 60 sec. Input resistance was calculated from the slope of the linear fit of the relationship between the change in membrane potential (ΔV) and the intensity of the applied current (between −60 pA and +60 pA) duration of 600 ms. Membrane time constant was measured during the application of a square current −60 pA, duration 600 ms. To analyze the action potential threshold, a square current, duration 600 ms, was applied from the resting potential levels to 300 pA with 5 pA steps.
The current amplitude necessary to evoke an action potential from the resting membrane potential was determined as a threshold current (in pA). Firing frequency rate was calculated as the number of spikes triggered by application of square current pulses (in 50 pA steps, duration 1 sec) in the range of 50-300 pA. Voltage sag ratio was analyzed after injection of a hyperpolarizing current from −300 pA to the resting potential (50 pA steps, duration of 1 sec). RRID:AB_10013383: this antibody labels S100B strongly, S100A1

| tPA in vitro uptake
The experiments were performed as described, with minor changes (Cassé et al., 2012). Briefly, after 14-16 days in culture, plated WT and KO cortical astrocytes were washed with DMEM (Life Technolo- On the next day, the beads were washed 3 times with RIPA buffer, incubated at 95 C for 10 min with protein loading buffer and used for further Western blot analysis.

| Image acquisition and data processing
Unless otherwise stated, before quantifications, the brightness/contrast ratio was enhanced for each image and the number of cells was calculated manually using the "cell counter" plugin in Fiji (Schindelin et al., 2012).
For the assessment of GFAP-and S100-positive cell numbers in For the assessment of cerebral cortex thickness in P28 mice, tissue sections were stained with cresyl violet. Images were taken with the Axiophot (Zeiss). Ten sections were analyzed per mouse and cortical thickness was measured at six positions using the free hand selection tool in Fiji (Schindelin et al., 2012). The cortical thickness measurements were averaged for each animal.
For the preliminary assessment of lateral ventricle size in P28 mice, tissue sections were stained with cresyl violet. Images were taken with the Axiophot (Zeiss). Three sections were analyzed per mouse and ventricular size was measured with the free hand selection tool in Fiji (Schindelin et al., 2012). The lateral ventricle size was averaged for each mouse.
For the assessment of tPA levels in the hippocampus and cortex of P7, P14 and P28 mice, the images were obtained using the LSM 510 Meta (Axiovert 200M, Zeiss) and a 63× objective (Plan Apochromat, 1.4 Oil DIC). Two sections per animal (one with the dorsal hippocampus and one with the ventral hippocampus) were imaged.
Regions of interest (ROIs) included cortical layers 1-3 and 4-6, hippocampal CA1, CA3 and DG. Images were subjected to background subtraction (rolling = 50) and the fluorescence intensity for each image was calculated using Fiji (Schindelin et al., 2012). The measured intensities were summed up for each ROI and averaged for each animal.
For the tPA uptake quantifications, 6-8 images per astrocyte culture were obtained randomly with the help of a fluorescent microscope (Axiophot, Zeiss). Before analysis, channels were split for each image in Fiji (Schindelin et al., 2012). where the Cre expression is under the control of Empty Spiracles Homebox 1 (Emx1) promoter (Iwasato et al., 2000). Emx1 expression is found as early as embryonic day 9.5 (E9.5) in the mouse and is restricted to radial glia cells of the dorsal telencephalon (Cecchi & Boncinelli, 2000). The telencephalic specificity of the Emx1-Cre line

| Emx1Cre-Lrp1 KO animals exhibit severe epileptic seizures
To determine the cause behind the early lethality of the Lrp1 KO mice, we monitored the mice daily, starting from P7. Upon handling of the KO mice, we observed seizure episodes, as early as P19 that corresponded in their severity to stage 4 and 5 of Racine's classification (Racine, 1972). Video monitoring of P15-28 mice confirmed that seizures indeed contributed to the premature mortality of the gross number of Emx1Cre-Lrp1 −/− mice. For example, the P19 KO mouse shown in Supporting Information Videos S1 and S2 experienced a severe seizure episode (Supporting Information Video S1) that  Table 1). The detected impairments included decreased action potential thresholds and increased action potential firing rates (Figure 2f,h), all hallmarks of a hyperexcitable phenotype.
In support of our findings, subsequent histology revealed that the upregulation of proliferating cells in the hippocampal dentate gyrus (DG), which is typical for acute phase epilepsy (Gu, Li, Shang, Hou, & Zhao, 2010;Sankar, Shin, Liu, Katsumori, & Wasterlain, 2000), only occurred in P28 KO mice ( Figure S4A,B). To further confirm increased excitability and determine affected regions, a P21 KO mouse that suffered from a seizure episode was transcardially perfused 2 hr after seizure onset together with a WT littermate. In order to detect active cells an antibody against the proto-oncogene c-fos was used. C-fos is an established marker for the detection of generalized seizure episodes as its basal expression is low in neurons and becomes only transiently expressed after synaptic activation occurs (Dragunow & Robertson, 1987;Sagar, Sharp, & Curran, 1988 Figure S2C F I G U R E 2 Legend on next page.
Emx1Cre-Lrp1 −/− mice, contribute to the shortened life span and weight loss and persist in surviving mice.

| Emx1Cre-Lrp1 KO animals exhibit ataxia and an enlarged ventricular volume
As Lrp1 mouse models show motor impairments (Q. Liu et al., 2010;May et al., 2004), we expanded the examination of our mutants by determining whether, in addition to epilepsy, also motor deficits occurred in Emx1Cre-Lrp1 KO mice. In the tail suspension test, young, P20 Emx1Cre-Lrp1 −/− mice promptly clasped their hindlimbs ( Figure 3a), a sign of either motor excitation, or reduction in motoneuron inhibition (Q. Liu et al., 2010;May et al., 2004). To elucidate whether Lrp1 KO mice indeed suffered from motor impairments, we subjected surviving P40 mice to a battery of motor tests. The motor tests specified that Emx1Cre-Lrp1 −/− mice displayed an ataxic phenotype, given their impaired performance in the rotarod, pole and hangwire tests (Figure 3b-d). A major characteristic of ataxic mice is the reduction in stride length, which is compensated by an increase in stride width. Subsequent footprint analysis of our mice unraveled altered gait including a decrease in stride length and, surprisingly, stride width (Figure 3e,f). In accordance, as seen in Figure 3f, KO mice require more steps to cover the same distance.
To delineate the cellular basis of the ataxic phenotype, brain tissue sections of the dorsal telencephalon of P28 WT and KO mice were analyzed. In comparison to age-matched WT mice, we found a striking increase in lateral ventricle size ( Figure S7A,B,D) and compressed appearing hippocampi in KO mice ( Figure S7D). To evaluate our observations quantitatively, we subjected P56 Lrp1 KO and WT animals to a magnetic resonance imaging (MRI) analysis. We   Figure S3-6 as well as in Table 1 and Supporting Information Videos S1, S2 and S3 Note: This is a summary table depicting data shown in Figure 2e-i. Data are presented as mean ± SD. For statistics, one-way analysis of variance (ANOVA) was applied, for analyzing firing frequency and voltage sag ratio two-way ANOVA repeated measures was applied. n is the number of neurons, p the probability, and P the postnatal day.
hippocampi were unchanged (Figure 4g,h, respectively; Figure 4i shows additionally a tridimensional overview of the analyzed hip-  ( Figure S9E). The levels of GluN1 detectable via WB represent the bulk levels of GluN1 that correspond to combined cell surface and intracellular levels. It remained therefore possible that despite stable bulk levels, the distribution of GluN1 within the neuronal cell membrane could be changed in our mutants, as shown before for the Lrp1 knock-in mouse model (Q. Liu et al., 2010;Maier et al., 2013;Martin et al., 2008). To enable a separate detection of surface and intracellular proteins in our model, we cultured E14.5 cortically derived Lrp1 WT and KO neurons and subjected them to cell surface biotinylation (Q. Liu et al., 2010;Maier et al., 2013;Martin et al., 2008). As shown in Figure 5g, GluN1 levels were increased on the surface of Lrp1 KO neurons, leading to the conclusion that Lrp1 deletion caused an accumulation of NMDARs on the neuronal surface.
Taken together, the upregulation of GluN1 surface expression and the developmental, tissue-dependent downregulation of PSD-95 levels contribute to a functional deficit in synaptic transmission, similar to the one observed in the SynapsinCre-Lrp1 −/− mouse (May et al., 2004). Yet, as neuronal Lrp1 mice exhibit rather a mild phenotype with increased hyperexcitability, but no seizures (Q. Liu et al., 2010;Maier et al., 2013;May et al., 2004;Nakajima et al., 2013), neuronal changes alone are not sufficient to fully explain the emergence of epileptic activity that we detected in our mice.
According to recent advances in the field of epilepsy, glial cells, and in particular astrocytes, appear vital for the establishment of adequate circuitry in the developing brain. Given the remarkable characteristic of the Emx1Cre-Lrp1 mouse, namely that Lrp1 loss occurs in telencephalic radial glia cells and their neuronal, and glial progeny, we proceeded to determine whether cortical and hippocampal astrocytes contributed to the severity of the neurological phenotype also in our case.

| Emx1Cre-Lrp1 KO mice show prominent changes in the astrocytic population
Considering that the in vitro KO of Lrp1 in cortical NSPCs leads to an increased astrocyte progeny number (Safina et al., 2016), we asked whether neuronal deficits present in Emx1Cre-Lrp1 −/− mice were accompanied by alterations in radial glia and astroglia during embryonic and postnatal development.  (Figure 6f-h), but a mild increase in their number was apparent in the ventricular zone ( Figure 6i). Given that we found no prominent alterations in the astroglia at E18.5, we focused on studying postnatal development of astrocytes. We examined P0 animals but found, similar to E18.5, no changes in Sox9-positive cell number in the cortices or the corpus callosum of KO mice ( Figure S10A). In the next step we therefore looked at the protein levels of glial fibrillary acidic protein (GFAP) in P7, P14 and P28 mice. During neonatal development, GFAP levels remained stable ( Figure S11A,B), but became significantly increased in cortices and hippocampi of P28 mice in comparison to a broad astrocyte marker aldehyde dehydrogenase 1 family member L1 (ALDH1L1) (Cahoy et al., 2008) (Figures 7a and S12A,B). This increase persisted in P56 mice (Figure 7b).
In support of the developmental upregulation of GFAP protein levels in KO animals, we found an increased number of GFAP-positive cells in the cortex and hippocampus of P28 KO mice (Figure 7f,g,j) but not in P7 ( Figure S13A,C,D) and P14 KO mice (Figure 7c,d,i).
To further analyze astrocyte development in neonatal and juvenile mice we used S100 calcium binding protein b (S100b), a broader astrocyte marker, the expression onset of which has been shown to occur when neocortical GFAP-positive cells lose their neural stem cell potential (Raponi et al., 2007). S100b levels were constant at P7 ( Figure S13B,C,E) and at P28 (Figures 7f,h and   S13G). At P14, however a significant increase in cortical S100bpositive cell number (Figure 7e), but not hippocampal ( Figure S13F) was observed. To confirm whether such transient S100b upregulation indicated increased astrocyte numbers and to clarify whether the upregulation of GFAP pointed to the presence of hypertrophic astrocytes (Pekny & Nilsson, 2005;Sofroniew & Vinters, 2010), we investigated in more detail astrocytic development at P14 and P28-stages before and after seizure onset, respectively.
We examined the levels of Vimentin, a type III intermediate filament protein that is both expressed in neural stem cells and considered a reactivity marker for astrocytes alongside Nestin (Liddelow & Barres, 2017). Vimentin expression appeared to be most visible in the hippocampal DG and at the level of the piriform cortex in P28 mice, supporting the presence of reactive astrocytes in seizing animals ( Figure S10B). Analysis of Sox9-positive cells in P28 mice revealed that although the somatosensory cortex did not exhibit altered astrocyte numbers (Figure 8a), at the level of the entorhinal/piriform cortex, a significant increase was found (Figure 8b). Both Vimentin and GFAP-positive cell number was significantly increased in all analyzed cortical regions, however the increase at the level of the piriform cortex was most prominent (compare Figure 8d with Figure 8e). The Reactive gliosis can be accompanied by an inflammatory response as shown for various injury models (Liddelow & Barres, 2017). The cojoint presence of reactive astrocytes and an increased number of Iba-1 microglial cells has been found for a different Lrp1 mutant (Q. Liu et al., 2010). To clarify whether in our model this was also the case, the Iba-1-positive cell number in the tissue of P28 mice was counted. No increase in the number of cortical or hippocampal Iba-1-positive cells was detectable (Figure S11C,E-G, respectively), however Iba-1-positive cells with an ameboid (active) morphology were encountered in cortical areas where the glial response was most prominent ( Figure S11D). This in turn suggests that astrocyte reactivity at least in the cortex can be accompanied by mild activation of microglial cells at this stage.
Taken together, the above results indicate that Emx1Cre-Lrp1 −/− mice exhibit developmental alterations in the astrocytic progeny, seen both prior to the appearance of neuronal deficits as well as coinciding with hyperexcitability and seizure emergence. Both characteristics impact on proper circuit formation and maturation in our model (Volterra & Meldolesi, 2005). As the above results suggest that glial changes are also the cause and not merely the consequence of epilepsy, we next elucidated whether astrocytic functionality was affected by Lrp1 loss. In the latter case, astrocytes would emerge as the key player explaining the different severity of phenotypes between our Lrp1 mutant and others.

| Emx1Cre-Lrp1 KO mice exhibit developmental variations in glutamate transporter levels
Glutamate clearance is one of the major functions of astrocytes that can be altered during epilepsy (Seifert, Schilling, & Steinhauser, 2006).
Proteins that are critical for astrocytic glutamate uptake are glutamate transporters Eaat1 (GLT-1) and Eaat2 (GLAST). In the neonatal cortex and hippocampus GLT-1 levels are low, while GLAST expression is high. GLT-1 controls 90% of total glutamate uptake, its expression increases significantly during synapse formation, and is heterogeneously controlled in the cortex and hippocampus (Danbolt, 2001;Hanson et al., 2015). GFAP, glial fibrillary acidic protein; kDa, kilo Dalton; KO, knock-out; MW, Mann-Whitney U test; N, number of animals; P, postnatal day; S100, calcium binding protein b; W, statistics value for the MW test; WT, wildtype; μm, micrometer; #, number of. Supporting Information is presented in Figures S11A,B, S12, and S13 Both GLAST and GLT-1 levels have been shown to be lower in the cortex than in the hippocampus. In consequence, glutamate uptake in the neonatal cortex is slower in comparison to the neonatal hippocampus (Hanson et al., 2015). Although the lack of GLT-1 leads to spontaneous seizures, while its overexpression mitigates epilepsy (Kong et al., 2012;Petr et al., 2015;Tanaka et al., 1997), more complex GLT-1 expression changes during the time course of epilepsy were discovered recently (Hubbard, Szu, Yonan, & Binder, 2016). As reduced efficiency of astrocytic glutamate uptake leads to increased glutamate presence in the extracellular space and can cause excessive neuronal activity, we were interested whether changes in glutamate transporter levels were present in our KO mice before seizure onset.
As visible in Figure 12a,b, at P7 GLT-1 levels remained unchanged. A prominent increase was found in the cortex (Figure 12c) but not in the hippocampus (Figure 12d) of P14 KO mice, providing support for our S100b results (see Figure 7c,e). GLAST levels were significantly Together, the above indicates that although glutamate uptake appears not to be disrupted in P28 and P56 mice, the detected transient neonatal changes in the ratio of glutamate transporters support the presence of precocious astrocyte maturation and inadequate astroglial development in the current mutant.
3.7 | In vitro Emx1Cre-Lrp1 −/− derived cortical astrocytes show reduced tPA uptake capabilities Healthy astrocytes possess the ability to influence the efficacy of synaptic responses not only by means of the clearance of extracellular glutamate, but also via tPA (Fernandez-Monreal, Lopez-Atalaya, Benchenane, Leveille, et al., 2004). The serine protease, tPA, is both a potent NMDAR interaction partner, a prominent Lrp1 ligand, as well as an immediateearly gene activated by neuronal activity during long-term depression (LTD), LTP and seizures (Nicole et al., 2001;Tsirka, Gualandris, Amaral, & Strickland, 1995;Zhuo et al., 2000). In vitro astrocytes with reduced levels of Lrp1 do not internalize neuron derived tPA efficiently, resulting in elevated levels of tPA in the synaptic cleft, that affect NMDARmediated signaling (Cassé et al., 2012;Fernandez-Monreal, Lopez-Atalaya, Benchenane, Leveille, et al., 2004;Makarova et al., 2003). We hypothesized therefore that in our Lrp1 mutants the resultant absence of    Liu et al., 2010;May et al., 2004). The emergence of seizures in our case can be explained through differences in cell populations affected by the KO. Emx1, the promoter guiding the KO to radial glia, is expressed in the frontal brain, but not in the ganglionic eminence. Therefore glutamatergic, but not GABAergic neurons lack Lrp1 in our model (Gorski et al., 2002;Kummer, Kirmse, Witte, & Holthoff, 2012). The development of seizures early in postnatal life, as seen in our study, may in consequence be favored by an imbalance between excitation and inhibition. This is in contrast to SynapsinCre-Lrp1 −/− mice where the KO affects both glutamatergic and GABAergic neurons, and the excitation-to-inhibition ratio is thereby reportedly unaltered (Chiappalone et al., 2009;May et al., 2004). In agreement with this interpretation, our hippocampal ex vivo patch-clamp recordings revealed enhanced excitability of hippocampal pyramidal neurons in P28 Emx1Cre-Lrp1 −/− animals. Which inhibitory cell types are affected, as well as the regions in which they are affected, in our model, shall be the focus of future studies.
Although the circuitry resulting in ataxia in the current model remains to be elucidated, Emx1-expressing medium spiny neurons can potentially contribute to its development, as they are involved in both movement facilitation and inhibition (Cocas et al., 2009).
A striking phenotype characteristic found in Lrp1 KO mice is the enlargement of the lateral ventricular volume. Dysfunctional aquaporin 4-containing perivascular astrocytic endfeet can offer a possible explanation behind this feature. Aquaporin, as a transmembrane water channel, is especially crucial for extracellular space volume regulation and waste clearance via the glymphatic system (Eidsvaag, Enger, Hansson, Eide, & Nagelhus, 2017;Haj-Yasein et al., 2011). The depletion of aquaporin specifically in astrocytes impairs water influx and efflux from the brain parenchyma (Haj-Yasein et al., 2011).
Although no deficits in cortical thickness, neuronal or hippocampal nuclei numbers were detected in the current study, further research is needed to determine whether neuronal apoptosis also contributes to the lateral ventricular enlargement. A detailed analysis of the hippocampal volume, for example, using light-sheet microscopy, could furthermore help clarify the detected volume differences between the dorsal, intermediate and ventral hippocampi.
The current study demonstrates that Emx1Cre-Lrp1 −/− mice display a neuronal phenotype that contributes to seizure generation. The scaffolding protein PSD-95 is found in the postsynapse of excitatory synapses and plays a crucial role in the stability of dendritic spines (Chen et al., 2011;El-Husseini, Schnell, Chetkovich, Nicoll, & Bredt, 2000).  (Wyneken et al., 2001(Wyneken et al., , 2003, but the mechanisms are not yet fully understood. Altered PSD-95 levels upon Lrp1 loss emerge thereby as a potential contributor to the epileptic phenotype also in humans. In addition to changes in the postsynaptic compartment upon Lrp1 loss, another Lrp1 mutant indicates that NMDAR surface expression is dependent on Lrp1 NPXY2 motif mutations that are associated with hyperactivity and cognitive deficits (Maier et al., 2013). Our Emx1Cre-Lrp1 mouse model corroborates this, as Lrp1 KO neurons displayed unaltered total NMDAR GluN1 levels but increased GluN1 surface levels. This is traceable to the decreased endocytosis of GluN1-containing NMDARs following Lrp1 loss and the resulting retention of the NMDAR on the neuronal surface (Maier et al., 2013).
Considering that the elevation of extrasynaptic NMDARs is suggested to be involved in the pathophysiology of epilepsy (Frasca et al., 2011;Parsons & Raymond, 2014), we propose that NMDAR upregulation in Emx1Cre-Lrp1 −/− mice contributes to seizure generation. Whether it is confined to the extrasynapse or not remains an interesting question for future studies.
It is conceivable that the described changes in the neuronal lineage do not suffice to unravel the current complex phenotype, as none of the Lrp1 mutants that target neurons display epilepsy. Given that   (Safina et al., 2016), as well as in NG2-positive glia in vivo (Schafer et al., 2019) and should be therefore studied further.
In postnatal Emx1Cre-Lrp1 −/− mice, prominent astroglial number and property changes have been observed. A developmental upregulation of GFAP was found starting from P28 in KO cortices and hippocampi. S100b numbers were unaltered and astrocyte proliferation was also not influenced at this stage. Given that GFAP levels in physiological conditions can largely differ not only between various brain regions and developmental stages, but also between neighboring astrocytes, astrocytic reactivity, visualized by a higher density of GFAP positive cells seen in this study is due to the presence of higher GFAP protein levels (Liddelow & Barres, 2017). In line with this, no changes in ALDH1L1 protein levels were detected in P28 mice, a broad, highly specific antigen for astrocytes (Cahoy et al., 2008). Corroborating this interpretation, a previous study by Q. Liu et al. (2010) reported that the elimination of Lrp1 in neurons causes GFAP upregulation, albeit in the absence of seizures (Q. Liu et al., 2010).
Our further analysis confirmed that reactive gliosis is present in our model: not only GFAP but also Vimentin levels were increased in the cortex and reactive stem cell fibers were found in the DG of P28 mice. The combined analysis of Sox9 and GFAP with the proliferation marker Ki67 further revealed that, astrocytes located especially at the level of the piriform cortex, seem to be the first to respond to altered excitability. This is an interesting finding as the piriform cortex, stud-  Figure S12C It has been already established that alterations in the expression and function of astrocytes cause disturbances in neuronal functioning (Steward, Torre, Tomasulo, & Lothman, 1992). However, whether reactive astrocytes are the cause, or rather the consequence of neuronal dysfunction is still debated (Robel & Sontheimer, 2016;Verkhratsky et al., 2012). Supporting the first possibility and in line with our results, a recent study revealed that a conditional deletion of β1-integrin in radial glia cells leads to a chronic reactive astrogliosis, Scale bars represent 20 μm. Data are expressed as median with interquartile ranges (* indicates p < .05, ** indicates p < .01). For all, representative images are shown. CA1, Cornu Ammonis 1; CA3, Cornu Ammonis 3; cx layers 1-3, cortical layers 1-3; cx layers 4-6, cortical layers 4-6; DG, dentate gyrus; FOV, field of view; KO, knock-out; MW, Mann-Whitney U test; N, number of animals; P, postnatal day; tPA, tissue plasminogen activator; W, statistics value for the MW test; WT, wildtype; μm, micrometer. Supporting Information is presented in Figure S14 sufficient to induce spontaneous seizures, without gross brain abnormalities, or pronounced inflammation (Robel et al., 2009(Robel et al., , 2015. Interestingly, and analogous to our current in vivo and previous in vitro findings, ablation of β1-integrin from neural stem cells of the hippocampal DG leads to an increase of astrocyte progeny, in conjunction with a decrease of the number of radial neural stem cells (Brooker, Bond, Peng, & Kessler, 2016;Safina et al., 2016). Given Lrp1 is tightly involved in β1-integrin processing (Salicioni, Gaultier, Brownlee, Cheezum, & Gonias, 2004;Wujak et al., 2018), future studies regarding the β1-integrin role in astrocytic maturity and functionality, in the current model, are of value and can help elucidate why Lrp1 loss results in gliosis and how neuronal and astrocytic crosstalk is impacted in epilepsy.
Next to astrocyte reactivity, we detected a transient increase in S100b-positive cell numbers and GLT-1 levels in P14 KO cortices.
The increase in P14 S100b-positive cells, but not in Sox9-positive astrocyte numbers, indicate an earlier maturation of cortical astrocytes (Raponi et al., 2007). In line, the increase in GLT-1 levels suggests a possible, tighter NMDAR control in the mutant neonatal cortex (Hanson et al., 2015). Alternatively, as basal c-fos levels were found higher in P14 KO cortices in the current study and given that GLT-1 expression varies depending on seizure occurrence and or their altered functionality that favors glutamatergic signaling (Danbolt, 2001;Hanson et al., 2015). In combination with the precocious maturation of astrocytes circuit formation becomes altered, leading to erroneous connectivity between neurons later in life (Robel & Sontheimer, 2016).
Alongside altered glutamate transporter levels, we found differential tPA protein levels in neonatal and juvenile KO mice. In particular, the decrease in tPA levels detected at P7, its increase at P14 and decrease at P28 was found more prominent in the hippocampus than in the cortex. This indicates that tPA levels can be impacted by tPAs' cellular origin and regional environment (Louessard et al., 2016;Ste-venson & Lawrence, 2018). As tPA is an immediate-early gene expressed early after seizures (Qian, Gilbert, Colicos, Kandel, & Kuhl, 1993), its transient level changes are also traceable to the varying onset and severity of the seizures in our model.
Upon glutamate application to astrocytes cultured in vitro, tPA is promptly endocytosed via Lrp1 in a clathrin-and dynamin-dependent manner, preventing NMDAR-mediated neurotoxicity (Cassé et al., 2012;Nicole et al., 2001). In our model, cultivated postnatal Emx1Cre-Lrp1 −/− astrocytes displayed impaired functionality visualized by reduced tPA uptake capabilities, in line with previous studies (Cassé et al., 2012;Fernandez-Monreal, Lopez-Atalaya, Benchenane, Leveille, et al., 2004). Increasing evidence suggests that astrocyte dysfunction plays a pivotal role in the pathophysiology of epilepsy (Bedner et al., 2015;B. Liu et al., 2017;Robinson & Jackson, 2016). It is appreciated that the disruption of astrocytic tPA uptake leads to an increase in the levels of tPA present in the synaptic F I G U R E 1 5 Early astrocytic dysfunction upon loss of Lrp1 in radial glia cells contributes to an impairment of astrocyte and neuronal functioning and the emergence of a severe neurological phenotype. In wildtype mice, the presence of Lrp1 in radial glia cells and their progeny ensures proper maturation and function of astrocytes and neurons that in turn supports stable and adequate glutamatergic signaling. Upon Lrp1 deletion from radial glia cells early in development, the animals exhibit severe alterations in cell signaling that are exemplified by changes in brain morphology, epilepsy and ataxia. Loss of Lrp1 in the astrocytic progeny of radial glia cells disturbs astrocytic maturation, results in GFAP upregulation and impairs astrocytic tPA uptake. Altered astrocytic functionality on top of neuronal deficits: reduced PSD-95 and upregulated NMDAR impacts astrocyte-neuronal interactions and thereby synaptic plasticity and predisposes the animal to seizures. GFAP, glial fibrillary acidic protein; KO, knock-out; Lrp1, low-density lipoprotein receptor-related protein-1; NMDAR, N-methyl-D-aspartate receptor; PSD-95, postsynaptic density protein 95; tPA, tissue plasminogen activator; WT, wildtype cleft. This, in turn causes NMDAR-mediated neurotoxicity that interferes with proper synaptic transmission (Cassé et al., 2012; Fernandez-Monreal, Lopez-Atalaya, Fernandez-Monreal, Lopez-Atalaya, Benchenane, Leveille, et al., 2004;Nicole et al., 2001). Given the importance of tPA for glutamatergic signaling, its biased uptake by astrocytes exerts detrimental effects on cell signaling in developing Emx1Cre-Lrp1 −/− mice: tPA can potentially remain in the synaptic cleft longer, diffuse further and activate both synaptic and extrasynaptic NMDARs (Bertrand et al., 2015;Nicole et al., 2001;Parcq et al., 2012;Samson et al., 2008). In conclusion, based on our results and available literature that highlights tPAs' pivotal role in NMDAR dependent Ca 2+ -influx and NMDAR surface diffusion , we put forward that tPA imbalance caused by astrocytic deficits contributes to the development of the severe epileptic phenotype in the current mouse line.
To sum up the findings of our study, we propose that Lrp1 deficiency early in development-in both neurons and astrocytes-results in the deterioration of NMDAR-Lrp1-tPA signaling of these cells that profoundly impacts synaptic plasticity and glutamatergic signaling in the developing brain ( Figure 15). Our study provides evidence that early astrocytic dysfunction is an important factor contributing to morphological changes during mouse brain development that lead to the buildup of hyperexcitability, and seizures. Our results highlight