A novel role for MLC1 in regulating astrocyte–synapse interactions

Loss of function of the astrocyte membrane protein MLC1 is the primary genetic cause of the rare white matter disease Megalencephalic Leukoencephalopathy with subcortical Cysts (MLC), which is characterized by disrupted brain ion and water homeostasis. MLC1 is prominently present around fluid barriers in the brain, such as in astrocyte endfeet contacting blood vessels and in processes contacting the meninges. Whether the protein plays a role in other astrocyte domains is unknown. Here, we show that MLC1 is present in distal astrocyte processes, also known as perisynaptic astrocyte processes (PAPs) or astrocyte leaflets, which closely interact with excitatory synapses in the CA1 region of the hippocampus. We find that the PAP tip extending toward excitatory synapses is shortened in Mlc1‐null mice. This affects glutamatergic synaptic transmission, resulting in a reduced rate of spontaneous release events and slower glutamate re‐uptake under challenging conditions. Moreover, while PAPs in wildtype mice retract from the synapse upon fear conditioning, we reveal that this structural plasticity is disturbed in Mlc1‐null mice, where PAPs are already shorter. Finally, Mlc1‐null mice show reduced contextual fear memory. In conclusion, our study uncovers an unexpected role for the astrocyte protein MLC1 in regulating the structure of PAPs. Loss of MLC1 alters excitatory synaptic transmission, prevents normal PAP remodeling induced by fear conditioning and disrupts contextual fear memory expression. Thus, MLC1 is a new player in the regulation of astrocyte‐synapse interactions.

contextual fear memory expression. Thus, MLC1 is a new player in the regulation of astrocyte-synapse interactions.

K E Y W O R D S
astrocytes, fear conditioning, MLC1, perisynaptic astrocyte process, synaptic transmission 1 | INTRODUCTION Astrocytes are crucial for neuronal network functioning. They maintain an environment that enables neuronal activity through metabolic support (Belanger et al., 2011) and by orchestrating ion and water homeostasis (Simard & Nedergaard, 2004). In addition, they shape neuronal excitability, synaptic transmission and synaptic plasticity through bidirectional communication with neurons (Semyanov & Verkhratsky, 2021). Astrocytes form fine morphological specializations that contact and enwrap neuronal synapses. These processes are known as perisynaptic astrocyte process (PAPs) or astrocyte leaflets (Khakh & Sofroniew, 2015;Semyanov & Verkhratsky, 2021). Bidirectional signaling between PAPs and neuronal synapses is thought to expand the computational power of neuronal networks (Min et al., 2012). Because of their crucial physiological role, astrocyte dysfunction has severe consequences for brain functioning Sosunov et al., 2018).
Astrocyte function is disrupted in the rare neurological disease megalencephalic leukoencephalopathy with subcortical cysts (MLC) (van der Knaap et al., 2012). MLC is characterized by increased brain water content, progressive white matter vacuolization, slowly progressive motor deterioration, epilepsy and mild cognitive deficits . In most MLC patients, the disease is caused by bi-allelic pathogenic variants in MLC1. Pathological hallmarks of MLC, such as the increased brain water content and progressive myelin vacuolization, are present in Mlc1-null mice (Dubey et al., 2015).
Mlc1-null mice do not show motor deficits or spontaneous seizures, but they exhibit a subclinical seizure phenotype. These mice provide a valuable tool to understand MLC disease mechanisms.
Whether MLC1 contributes to astrocyte physiology in other morphological astrocyte domains is unclear. Here we investigated whether MLC1 has a role in regulating the structural and functional organization of the synapse. We focused on excitatory synapses in the hippocampal CA1 region, a crucial hub for synaptic plasticity and memory formation. Using biochemistry and electron microscopy (EM) we show that MLC1 is present in PAPs that are in apposition to excitatory synapses. We find that the PAP tip that extends toward the synaptic cleft is shortened in Mlc1-null mice. Basal excitatory synaptic transmission and extracellular glutamate dynamics under challenging conditions are altered in Mlc1-null mice. Contextual fear conditioning-induced structural PAP remodeling is absent in Mlc1-null mice, and contextual fear memory expression is impaired. Together, this study uncovers a new role for MLC1 in regulating astrocyte-synapse interactions and in processing of aversive memories.

| Animals
Experimental procedures involving mice were in strict compliance with animal welfare policies of the Dutch government and were approved by the Institutional Animal Care and Use Committee of the VU University, Amsterdam. All experiments were performed using 2-5 month old Mlc1-null mice or wildtype littermates, both on a C57Bl6/J background. The experimenter was blind to the genotype.
For EM analysis and behavioral experiments, only male mice were used. Electrophysiology and glutamate imaging were performed on mice of either sex. Generation and characterization of these mice was described earlier (Dubey et al., 2015). Mice were housed on a 12 h light/dark cycle with unlimited access to standard lab chow and water.

| Immunoblotting on subcellular brain fractions
Mice were sacrificed by cervical dislocation followed by decapitation.
The S2 fraction was ultracentrifuged at 100,000Âg for 2 h; the pellet was recovered as microsomal fraction. S1 was subjected to ultracentrifugation in a 0.85/1.2 M sucrose density gradient at 100,000 Â g for 2 h. Synaptosomes were recovered at the interface of 0.85/1.2 M sucrose. The hypotonic shock of synaptosomes in 5 mM HEPES with protease inhibitor for 15 min yielded the synaptic membrane fraction, which was subsequently isolated by sucrose gradient ultracentrifugation as stated above at the interface of 0.85/1.2 M fraction. To obtain the PSD, the synaptosome fraction was extracted in 1% Tx-100 for 30 min, layered on top of 1.2/1.5/2 M sucrose, centrifuged at 100,000Âg for 2 h, and recovered as PSD-I at the interface of 1.5/2 M sucrose. PSD-I was subjected to second extraction in 2% Tx-100 for 30 min, subjected to sucrose gradient ultracentrifugation as stated above, and recovered at the 1.5/2 M sucrose interface. The PSD-II fraction was then pelleted in 5 mM HEPES by centrifuging at 100,000Âg for 30 min.

| Electron microscopy
Transmission EM analysis of the hippocampus was performed on tissue of 3-5 month old male Mlc1-null and wildtype mice. After transcardial perfusion with 4% paraformaldehyde (PFA) in phosphate buffered saline (PBS, pH = 7.4), the whole brain was dissected and post-fixated for 24 h in 4% PFA, followed by cryopreservation in 30% sucrose. Tissue was stored at À80 C until further processing.
Immuno-EM analysis was performed as described (Bugiani et al., 2017;Dubey et al., 2015). In brief, 40 μm thick hippocampal slices were cut on a cryostat for immunolabeling on free floating sections. Sections were quenched with endogenous peroxidase, treated with one freeze-thaw cycle and incubated with rabbit-anti-  (Gray, 1969).
For each synapse, the perimeter of the presynaptic bouton and the postsynaptic spine were measured. Contiguous astrocytes were identified by a relatively clear cytoplasm, irregular shape and occasional bundles of glycogen appearing as dark spots. Synapses were examined and categorized according to the type of astrocytic contact: No astrocyte contact, astrocyte-bouton contact, astrocytespine contact or astrocyte in contact with both bouton and spine elements. For synapses in which the astrocyte contacted the synaptic cleft, the shortest distance between the PAP tip and the PSD was measured. The length of the PAP tip was measured using a standardized method . Firstly, the base of the PAP tip was determined and a straight line was drawn from edge to edge. Secondly, the most distal part of the PAP tip was identified by taking the end of the protrusion, at the point closest to the synaptic cleft. The shortest distance between the line drawn at the PAP base and the most distal end of the PAP tip was defined as length of the PAP tip. All measurements were performed by the same experimenter, blinded for experimental conditions, to ensure consistency and reduce variation. All analyses were performed in ImageJ.

| Patch-clamp recordings
Slices were placed in a recording chamber continuously perfused with standard ACSF at 34 C. CA1 pyramidal cells were recorded in voltage clamp using a Multiclamp 700B amplifier (Molecular Devices) and PClamp (Molecular Devices) or MIES software (Allen Brain Institute). Germany). The stimulation pipette was placed 300 μm from the recorded CA1 pyramidal neuron and responses were measured in the presence of 10 μM gabazine. The intensity of stimulation was set from 10 to 100 μA to create an input/output curve. Subsequently, the stimulation intensity was set to an intensity which evoked half maximal EPSCs. When responses were stable for at least 10 min, γ-DGG was washed into the slice for 5 min. Two pulses were given with 100 ms interval to determine the paired pulse ratio (PPR). The PPR was calculated by dividing the amplitude of the second peak by the amplitude of the first peak.

| LTP recordings
Acute hippocampal brain slices were placed on an 8 Â 8 microelectrode array (MED64 system, Alpha MED Scientific; Osaka, Japan) and held in place inside the probe with a platinum ring. The probe with the slice was then placed on a connector and superfused with carboxygenated ACSF containing 1 μM gabazine at a speed of 2-3 mL/min for the rest of the experiment. Additional carbogen gas was blown over the recording chamber from above through a tube to optimize carbox-

| Virus injection
For iGluSNFR expression, mice received 0.1 mg/kg Tamgesic (RB Pharmaceuticals, UK) 30 min prior to start of surgery. They were anesthetized with isoflurane and mounted onto a stereotactic frame.

| Glutamate imaging
Mlc1-null mice (n = 6) and wildtype littermates (n = 6) were injected unilaterally with AAV2/5-GfaABC 1 D::SF-iGluSnFR.A184V vector as described above (virus injection). Three weeks after injection, coronal brain slices (350 μm) were obtained as indicated previously (acute brain slice preparation). Slices were constantly carboxygenated with 5% CO 2 /95% O 2 and bath temperature was set at 32 C with constant flow (2.5 mL/min) of standard aCSF (in mM): NaCl 125, NaHCO 3 25, KCl 3, NaH 2 PO 4 1.2, CaCl 2 1.3, MgSO 4 1, MgCl 2 1, sodium pyruvate 3, sodium ascorbate 1 and D(+)-glucose 25, containing CNQX (10 μM), APV (50 μM) and Gabazine (2 μM Traces were then averaged and decay tau was calculated. Both a sin- exponential function were fitted to the decay phase of each trace using Igor Pro8 (WaveMetrics). Per experimental condition a choice between using the single or the double exponential decay fit was made based on the following criteria: the double exponential fit was chosen if both τ fast and τ slow had positive values, both were < 30 s, and if goodness of fit (χ 2 ) improved by >15% when compared to the single exponential fit. In other cases, the single exponential fit was chosen.
Since most traces were fit well with a single exponential decay, and since in all experimental conditions the average arithmetic contribution of τ fast to the decay (A fast /(A fast + A slow )) was >90%, only τ fast was used for presentation and statistics.

| Behavioral experiments
Contextual fear conditioning was performed in a fear conditioning system with integrated tracking software (TSE, Bad Homburg, Germany). The system consisted of two Plexiglas boxes (36 Â 21 Â 20 cm). The "conditioned context" box was fitted with a stainless steel floor grid (4 mm diameter, distance 9 mm) inside a dark gray housing unit, with light intensity set at 120-500 lux. This box was cleaned with 70% ethanol prior to testing each animal. The "novel context" box had a solid plastic floor, white walls and ambient white light from the experimental room. This box was cleaned with 1% acetic acid prior to each trial to further differentiate the context by smell. A white noise background (68 dB) was present during all conditions. Animals were placed in the conditioned context box and baseline activity was monitored for 180 s, after which they received a foot shock (2 s, 0.7 mA). This allowed animals time to process the environment and link this context to the aversive stimulus. Animals remained in the box for an additional 30 s after which they returned to their home cage. For electron microscopy, mice were sacrificed 4 h after they received a foot shock in the conditioned context. For behavioral experiments, after 24 h, retention of the fear memory was tested by returning mice to the conditioned context for 180 s, without foot shock. Two hours after the retention trial in the conditioned context, mice were placed into the novel context and freezing behavior was monitored for another 180 s to ensure that any observed freezing behavior was context specific. In all tests, activity (cm/s) was measured by a photo beam detection system (10 Hz detection rate, resolution of 1.3 Â 2.5 cm). Freezing was defined as lack of movement other than respiration and heartbeat during a 5 s period.

| Statistics
Data representation and statistical analysis were performed using

| RESULTS
3.1 | MLC1 is expressed in perisynaptic astrocyte processes MLC1 is an astrocyte membrane protein that is mainly present at astrocyte processes contacting brain-fluid barriers (Boor et al., 2005;Boor et al., 2007). To investigate whether MLC1 protein is also found associated with synapses, we performed immunoblotting on subcellular fractions prepared from mouse hippocampus. Immunoblotting for MLC1 revealed a specific band (41 kDa), both in whole brain homogenates and in synaptosomes, which was absent in , synaptic membranes (SYM) and postsynaptic density (PSD) reveals an expression pattern for MLC1 that resembles that of the astrocyte glutamate transporter GLT-1. As a control, expression of the presynaptic protein synaptophysin (SYP) and the postsynaptic protein PSD-95 were monitored. Trichloroethanol gel stain (bottom) was used to affirm equal loading conditions. (C) Immuno-EM reveals immunopositivity for MLC1 in perisynaptic astrocyte processes (marked with asterisks; a: astrocyte; b: presynaptic bouton, s: postsynaptic spine). Astrocyte elements are pseudo-colored in purple. Scale bars: 200 nm.
( Figure 1a; Figure S1 for full gel images). Given that MLC1 is not expressed in neurons (Dubey et al., 2015), and PAP proteins are a constituent of synaptosomes (Carney et al., 2014;Rao-Ruiz et al., 2015), these findings suggest that MLC1 is expressed in astrocyte membranes associated with synaptic elements. We then fractionated our synaptosome samples into different membrane and synaptic compartments (Pandya et al., 2017). MLC1 protein was found in microsomes, P2 membrane fraction, synaptosomes and particularly enriched in synaptic membranes, whereas it was absent from the postsynaptic density (PSD; Figure 1b). Our fractionation was validated by determining the levels of the presynaptic protein synaptophysin (SYP; 38 kDa), which was enriched in synaptosome and synaptic F I G U R E 2 Legend on next page. membrane fractions while depleted from the PSD fraction, and of the postsynaptic protein PSD-95 (95 kDa), which was highly enriched in the PSD fraction. The expression pattern of MLC1 also resembled that of the glutamate transporter GLT-1, which is known to be highly expressed in astrocyte membranes surrounding excitatory synapses ( Figure 1b) (Danbolt, 2001). To confirm the localization of MLC1 in PAP membranes in the hippocampus we performed immuno-EM. The specificity of the MLC1 antibody for immuno-EM was confirmed by inspecting astrocyte endfeet, which are well known to have high expression of MLC1 (Teijido et al., 2004). DAB-positive endfoot staining, indicating MLC1 expression, was observed in wildtype, but not in Mlc1-null tissue ( Figure S2). When investigating excitatory hippocampal synapses, clear DAB-positive staining was observed in PAPs that were in close contact with excitatory synapses ( Figure 1C). Therefore, both biochemical experiments and EM analysis reveal that MLC1 is not solely localized to astrocyte endfeet at brain-fluid barriers, but is also found in PAPs associated with excitatory hippocampal synapses.

| Loss of MLC1 disrupts structural astrocytesynapse interactions
To investigate how loss of MLC1 affects the ultrastructure of synaptic elements and astrocyte-synapse interactions, we performed EM analysis of hippocampal synapses in wildtype and Mlc1-null mice. We identified excitatory synapses and classified them according to astrocyte contact category (no contact, only bouton in contact, only spine in contact, or both bouton and spine in contact; Figure 2a). In Mlc1-null mice we observed a significant alteration in synapse distribution over the different categories (multi-level χ 2 test; P = 0.049). This was due to a significant reduction in the fraction of synapses in which only the postsynaptic spine was contacted by a PAP in Mlc1-null mice (wildtype: 13.0%, Mlc1-null: 8.5%; multi-level χ 2 test; P = 0.0025), since distribution over the other contact categories was not significantly altered. The size of pre-and postsynaptic elements was not affected in Mlc1-null mice (Figure 2b). Similarly, no change was observed for the contact area with pre-or postsynaptic elements that were in contact with an astrocyte (Figure 2c).
However, we found that the tip of the PAP, which is the closest point to the synaptic cleft, was shorter in Mlc1-null mice (wildtype: 54.7 ± 1.6 nm, n = 184/6, Mlc1-null: 39.6 ± 1.4 nm, n = 152/6; nested 1-way ANOVA; P = 0.0085). In line with this, there was a trend towards an increased distance between PAP tip and PSD, although this did not reach significance (wildtype: 58.9 ± 5.7 nm, n = 185/6, Mlc1-null: 79.1 ± 3.8 nm, n = 152/6; nested 1-way ANOVA; P = 0.104). In conclusion, the interaction of astrocytes with excitatory hippocampal synapses is altered upon loss of MLC1, illustrated by a reduced number of spines contacted by an astrocyte and a reduced PAP tip length in Mlc1-null mice.
It is well established that astrocyte-synapse interactions play a crucial role in the ability of synapses to undergo long-term plasticity (Henneberger et al., 2010). Therefore, we investigated whether hippocampal long-term potentiation (LTP) was altered in Mlc1-null mice.
LTP in CA1 upon stimulation of the Schaeffer collateral pathway was assessed in acute hippocampal brain slices recorded in a multielectrode array setup (Figure 4a). We found that LTP was not affected

| iGluSNFR imaging reveals altered glutamate dynamics in Mlc1-null mice
PAPs are enriched in high-affinity glutamate transporters (GLT-1) and their localization close to the synaptic cleft is thought to decrease glutamate spill-over by increasing the efficacy of glutamate uptake (Danbolt, 2001;Henneberger et al., 2020;Pannasch et al., 2014).
Therefore, we tested whether the temporal dynamics of extracellular glutamate were altered by the structural alterations of PAPs caused  Therefore, we investigated whether hippocampus-dependent memory was affected in Mlc1-null mice using a contextual fear conditioning paradigm. Animals were placed in a fear conditioning chamber (conditioned context). In the first 3 min after being placed in the chamber, Mlc1-null mice did not differ from wildtypes in percentage of the chamber explored, percentage inactivity or total distance moved ( Figure S3). This is in line with our earlier findings that these mice have no gross motor abnormalities and do not show alterations in spontaneous activity levels (Dubey et al., 2015). In addition, both wildtypes and Mlc1-null mice showed low levels of baseline freezing, indicating no pre-existing fear of the conditioning box (baseline freezing, wildtype: 4.7 ± 2.1%, n = 15, Mlc1-null: 2.1 ± 1.3%, n = 11; P > 0.99; Figure 6a). Mice then received a single foot shock to establish a contextual fear memory and were returned to the conditioning chamber 24 h later to assess fear memory. At this point Mlc1-null animals showed significantly reduced levels of freezing (freezing in the conditioned context, wildtype: 44.4 ± 3.8%, n = 15, Mlc1-null: 28.9 ± 5.1%, n = 11; P = 0.022; Figure 6a). Specificity of the fear memory to the conditioned context was validated by placing the mice in a novel context. Both wildtype and Mlc1-null mice froze significantly less in the novel context when compared to the conditioned context (freezing in the novel context, wildtype: 20.4 ± 4.0%, n = 15, Mlc1-null: 12.4 ± 2.9%, n = 11; conditioned versus novel context, wildtype: P < 0.0001; Mlc1-null: P < 0.026; novel context, wildtype versus Mlc1-null: P = 0.65 Figure 6a). In conclusion, Mlc1-null animals show a deficit in contextual fear memory.
Previous studies suggest that increased neuronal activity and synaptic plasticity are associated with an alteration in structural astrocyte-synapse interactions (Bernardinelli et al., 2014;Genoud et al., 2006;Henneberger et al., 2020;Perez-Alvarez et al., 2014). In addition, 4 h after fear conditioning there are clear changes in PAP protein levels (Rao-Ruiz et al., 2015). Therefore, we used EM to compare structural properties of astrocyte-synapse interaction between naïve animals that did not leave their homecage and animals that underwent fear conditioning and were sacrificed 4 h later (delayed shock). In wildtype animals we observed that fear conditioning induced a shortening of the PAP tip directed towards the synapse (wildtype PAP tip length, home-cage: 54.7 ± 1.6 nm, n = 184/6, delayed shock: 39.4 ± 4.6 nm, n = 156/6; P = 0.0087; Figure 6b, c).

| DISCUSSION
In this study we assessed whether the astrocyte protein MLC1 is expressed in PAPs associated with hippocampal excitatory synapses, and whether it plays a role in regulating astrocyte-synapse interactions. MLC1 is almost exclusively expressed in astrocytes (Dubey et al., 2015), where it is highly enriched in astrocyte endfeet at brainfluid barriers (Boor et al., 2005;Teijido et al., 2004). Here it colocalizes and interacts with numerous other proteins involved in brain ion and water homeostasis (Boor et al., 2007; Together, our results reveal an unexpected novel role for MLC1 in astrocyte processes at the synapse.

| A synaptic role for MLC1
The importance of astrocytes in regulating synaptic transmission is well-established. Astrocytes can sense neurotransmitter release through expression of a wide array of metabotropic receptors and modulate synaptic transmission through the release of gliotransmitters. In this way, astrocytes can modulate basal synaptic transmission and alter both short-term and long-term synaptic plasticity (Perea et al., 2009). Furthermore, astrocytes provide metabolic support to meet the synaptic energy demand (Rouach et al., 2008). Finally, the close proximity of PAPs to the synapse limits spill-over of glutamate from the synapse (Bergles & Jahr, 1997).

EM analysis of astrocyte-synapse interactions showed that
PAPs were structurally altered in Mlc1-null mice. It is important to mention that this conclusion is based on 2D-analysis. Therefore, we may have missed synapses that were in contact with an astrocyte if the astrocyte process was not present in the cutting plane used for analysis (see Kater et al., 2023 for a more extensive discussion of this issue). Nevertheless, the number of synapses contacted by an astrocyte in wildtype mice in our study (49%) was close to the 57% observed using 3D EM analysis (Ventura & Harris, 1999). In addition, by sampling a large number of synapses the average distance of the PAP toward the synaptic cleft and the length of the PAP tip can be reliably compared between genotypes or conditions.
A number of synaptic properties were affected in Mlc1-null mice.
We observed an increase in mEPSC interevent interval. This could indicate either a decrease in number of excitatory synapses or a decrease in synaptic release probability. Surprisingly, amplitude and dynamics of evoked excitatory transmission were not altered in Mlc1null mice, which is in line with an earlier study where we assessed hippocampal excitatory transmission in Mlc1-null mice using field EPSP recordings (Dubey et al., 2018). Therefore, spontaneous and evoked excitatory neurotransmission are differentially affected in Mlc1-null mice. It is well established that evoked and spontaneous transmission rely on divergent (but partially overlapping) synaptic populations and/or different release mechanisms (Peled et al., 2014;Ramirez & Kavalali, 2011). But why loss of MLC1 specifically affects spontaneous transmission remains unresolved. It is important to emphasize that MLC1 is not present pre-or postsynaptically, but only on astrocytes.
Therefore, the impact of MLC1 loss on mEPSC properties most likely occurs through altered astrocyte-neuron interactions at the tripartite synapse. We also studied mIPSC properties, and found no alterations.
Therefore, basal inhibitory synaptic transmission is intact upon loss of MLC1 function.
In this study, we focused on the impact of loss of MLC1 on synaptic transmission in the hippocampus. However, MLC1 is expressed by astrocytes throughout the brain, therefore, similar effects on astrocyte-synapse interactions are likely also occurring in other brain regions. A recent study investigated the role of GlialCAM (also called HepaCAM) on synaptic transmission in neocortex (Baldwin et al., 2021). GlialCAM is a chaperone for MLC1 (Capdevila-Nortes et al., 2013). In patients, loss of GlialCAM function leads to classical MLC, which is indistinguishable from what is seen in patients with MLC1 mutations . The proteins share their localization to astrocyte endfeet, and loss of GlialCAM leads to mislocalization of endfoot MLC1 (Bugiani et al., 2017). Therefore, an important question is whether the roles of GlialCAM and MLC1 at the synapse also overlap. In the study by Baldwin and colleagues it was shown that GlialCAM regulates competition for territory between astrocytes during development (Baldwin et al., 2021). In addition, loss of GlialCAM from astrocytes was found to increase mEPSC amplitude and decrease mIPSC amplitude in the juvenile mouse neocortex. Frequency for both was unaltered. These observations deviate from our findings on hippocampal synaptic transmission upon loss of MLC1.
Therefore, either the synaptic role of MLC1 differs from that of Glial-CAM, or there are region-or age-specific differences in how these proteins alter synaptic transmission.

| Contextual fear memory and MLC1
The retraction of PAPs from excitatory synapses observed after contextual fear conditioning did not take place in Mlc1-null mice, where PAPs are already shorter under basal conditions. PAPs are highly motile (Hirrlinger et al., 2004) and can undergo activity dependent and behaviorally induced structural rearrangements (Genoud et al., 2006;Oliet et al., 2001;Ostroff et al., 2014 We previously showed that an immediate shock (IS) control, in which mice received a shock directly after entering the conditioning box without the chance to form an associative memory, is not sufficient to cause a substantial increase in PAP-PSD distance in WT mice (Badia-Soteras et al., 2022). This suggests that it is mainly the association between foot shock and a novel context, rather than these components in isolation, which causes PAP retraction.
We recently investigated the effect of loss of the astrocyte protein  (Dubey et al., 2015). They have an increased brain water content and show progressive vacuolization of the white matter.
Perivascular astrocyte processes in Mlc1-null mice appear swollen, and extracellular potassium dynamics are disrupted (Dubey et al., 2018). MLC mice show increased interictal spiking activity and a reduced seizure threshold (Dubey et al., 2018). Therefore, while the short PAP tip and the absence of fear conditioning-induced PAP remodeling in Mlc1-null mice could be directly related to disrupted memory, other neurological abnormalities in Mlc1-null mice should also be considered. In addition, similar to its role in endfeet, MLC1 might also control extracellular ion and water homeostasis at the synapse. Disruption of this process in Mlc1-null mice might affect synapse physiology via other ways next to the effect on PAP morphology described here. Investigating the relative contribution of perivascular and perisynaptic MLC1 roles would require a way to specifically interfere with these two subcellular compartments.
Unfortunately, tools to achieve this are not available yet.

| Molecular mechanism by which MLC1 may modulate PAP structure
An open question is how MLC1 modulates PAP structure. In endfeet, MLC1 directly interacts with several proteins involved in brain ion and water homeostasis, but the exact function of MLC1 in astrocyte processes at the synapse is unclear. Biochemically identified MLC1 interactors include GlialCAM, the beta1 subunit of the Na + /K + -ATPase, TRPV4, the members of the dystrophin-associated glycoprotein complex, Kir4.1, Caveolin-1 and the chloride channel ClC-2 (Brignone et al., 2015). In addition, MLC1 indirectly modulates volume regulated anion currents, and thereby affects the astrocyte response to cell swelling (Elorza-Vidal et al., 2018;Ridder et al., 2011). Furthermore, several studies indicate interactions of MLC1 and GlialCAM with the cytoskeleton. MLC1 overexpression in COS-7 cells leads to morphological changes and decreases cell motility, by interacting with the Actin capping proteins of the ARP complex (Hwang et al., 2019). Glial-CAM also directly interacts with the cytoskeleton and affects cellular motility and cell adhesion (Moh et al., 2009). Since MLC1 and Glial-CAM both play a role in astrocyte volume regulation and interact with the cytoskeleton, an exciting possibility is that in addition to MLC1 also GlialCAM has a role in determining PAP structure and motility through a concerted regulation of the cytoskeleton and ion and water fluxes. Cellular motility and filopodia formation require both precisely localized cytoskeleton alterations and associated directional water influx (Loitto et al., 2009). Future studies should unravel the way in which MLC1 regulates the structure and motility of PAPs.

| Linking MLC1 to cognitive dysfunction
Loss of MLC1 function is the primary genetic cause of the rare leukodystrophy MLC. MLC is mainly characterized by slowly progressive motor dysfunction and ataxia, neurological phenotypes that are not directly linked to synaptic dysfunction. However, behavioral problems, cognitive impairment and delayed onset cognitive decline are also common in MLC patients . A recent case report suggested that MLC can be accompanied by catatonia and bipolar disorder (Ishikawa et al., 2020). In line with this, some specific MLC1 variants have been associated with these disorders in the general population (Meyer et al., 2001;Selch et al., 2007;Verma et al., 2005).

ACKNOWLEDGMENTS
The authors thank Adrian Negrean and Sven Kerst for design and assembly of the custom-built two-photon microscope, and Yvonne Gouwenberg and Rien Dekker for excellent technical assistance. This study was supported by a ZonMW VIDI grant (91718392 to RM) and ZonMW Memorable (7330508160 to MHGV and ABS). MvdK and RM are members of the European reference network for rare neurological disorders (ERN-RND), project ID 739510.

FUNDING INFORMATION
This study was supported by a ZonMW VIDI grant (91718392 to RM) and a ZonMW Memorable grant (7330508160 to MSJK).