To identify MLC1 and GLIALCAM variants described in the literature, we performed a PUBMED search using the search words ‘MLC1’, ‘GLIALCAM’, ‘HEPACAM’ ‘AQP4 MLC’ and ‘GPRC5B’. Articles published until July 2022 were included. We included papers that were written in English, Dutch or French. Only papers discussing patient data were included (e.g., descriptions of cloned plasmid variants without patient relevance were excluded). Variants were reported only when the coding sequence position was reported and when a conclusive MLC diagnosis (based on MRI) was reported in the study.

Patients from the database of the ALC were included in this study upon conclusive MLC diagnosis by clinical features, MRI and genetic confirmation of variants in MLC1, GLIALCAM, GPRC5B or AQP4. Written informed consent was obtained from families for phenotyping.

The diagnostic workflow for MLC patients in the ALC is as follows: First, the diagnosis of MLC is established based on the presence of macrocephaly and characteristic MRI abnormalities now or in the past (Figure 1). Genetic testing starts with Sanger sequencing of MLC1. If no potentially pathogenic variants are found, this is followed by Sanger sequencing of GLIALCAM. When both are negative, and the MRI diagnosis is unambiguous, multiplex ligation-dependent probe amplification (MLPA) and cDNA analysis using lymphoblasts are performed for MLC1. If these do not uncover potentially pathogenic variants, next-generation sequencing (NGS, preferably whole genome sequencing (WGS)) is performed to identify potential rare (non-coding) variants.

Three reporter constructs were generated with the pNL1.1 vector (Promega), in which MLC1-expression regulating DNA sequences c.-2,645 to c.-1 (wild-type or with the c.-190A>G or c.-195T>C variant) were cloned directly upstream of the nanoluciferase open reading frame similarly as previously described (Hamilton et al., 2017). This DNA sequence includes the MLC1 core promoter and encodes the full 5′ untranslated region (5′UTR). Sanger sequencing was performed to confirm the MLC1 sequence with or without either of the two variants in the three p.NL1.1-MLC1 plasmids. Subsequently, U373 cells were cultured in DMEMF12 + 10% FBS. 24 h before transfection 3,000 cells were plated in white half area 96 well plates. The next day cells were co-transfected with a wild-type or mutant pNL1.1-MLC1 plasmid and the pGL3 plasmid (Promega) as internal standard using Fugene6 according to manufacturer’s instructions. The pGL3 plasmid encodes the firefly luciferase open reading frame under regulation of the SV40 promoter. Approximately 40 h after transfection, nanoluciferase and firefly luciferase activities were measured with a plate reader (Victor2; Perkin-Elmer Life Sciences, Waltham, MA), as described (Hamilton et al., 2017). The nanoluciferase signal was normalized to the firefly luciferase signal to obtain the relative expression driven by the wildtype and mutant MLC1 sequences. Statistical analysis was performed with Brown-Forsythe ANOVA followed by Dunnett’s T3 multiple comparisons test using GraphPad Prism 9 (GraphPad, USA). Statistically significant differences were defined as p ≤ 0.05. Data are represented as mean ± SEM.

All variants described in this study have been submitted to the LOVD database (www.lovd.nl): MLC1: https://databases.lovd.nl/shared/transcripts/00013671; HEPACAM: https://databases.lovd.nl/shared/transcripts/00009260; AQP4: https://databases.lovd.nl/shared/transcripts/00002726; GPRC5B: https://databases.lovd.nl/shared/transcripts/00008881.

To estimate allele frequency for some variants, data was obtained from approximately 500,000 participants from the UK Biobank (Bycroft et al., 2018), a population-based sample of adults in the UK with self-report surveys, linked electronic health records, and genotypic data. The National Research Ethics Service Committee North West–Haydock ethically approved this initiative (reference 11/NW/0382) and participants provided informed written consent. Data were accessed under application #16406.

The MLC1 gene is located on chromosome 22q13. The gene contains 12 exons and 11 introns (Figure 2A). The 5′UTR consists of exon 1 and part of exon 2. Predictions and experimental studies show that the MLC1 protein has 8 transmembrane regions with both the amino- and the carboxy-terminus residing in the cytoplasm (Figure 2B). MLC1 most likely forms a trimeric structure in the membrane (Hwang et al., 2021). The protein has very low homology with other proteins, with highest similarity (less than 20%) with the shaker-related voltage gated potassium channel Kv1.1 α-subunit (Teijido et al., 2004; Brignone et al., 2015). The exact function of MLC1 is not known. However, experimental studies have implicated an indirect role for MLC1 in cell ion and water homeostasis.

Our search yielded a total of 151 unique MLC1 variants in patients with a confirmed MLC diagnosis (Figure 2; Supplementary Table S1). Biochemical studies have been performed for some variants (see Supplementary Table S1) (Teijido et al., 2004; Duarri et al., 2008; Lopez-Hernandez et al., 2011b; Lanciotti et al., 2012; Capdevila-Nortes et al., 2013a; Sirisi et al., 2014; Xu et al., 2021). For most tested variants these studies reveal reduced plasma membrane levels, with retention of the protein in intracellular compartments (possibly the endoplasmic reticulum). In addition, protein stability is reduced for several variants. This might be a consequence of misfolding or defective oligomerization, leading to disrupted protein structure.

Several (possible) founder variants in MLC1 have been described. c.135dup; p.(Cys46Leufs*34) is a founder variant in East Indian individuals from the Agrawal community (Leegwater et al., 2002; Singhal et al., 2003; Gorospe et al., 2004). c.176G>A; p.(Gly59Glu) is a possible founder variant in Libyan Jews (Ben-Zeev et al., 2002). c.278C>T; p.(Ser93Leu) is common in Japanese individuals (Shimada et al., 2014), while c.824C>A; p.(Ala275Asp) is a founder variant accounting for the majority of MLC patients of Korean ancestry (Choi et al., 2017). c.908_918delinsGCA; p.(Val303Glyfs*96) is a founder variant in individuals with Egyptian ancestry.

The GLIALCAM gene is located on chromosome 11q24. It contains 7 exons and 6 introns (Figure 4A). The protein GlialCAM, encoded by GLIALCAM, is an immunoglobulin-like transmembrane adhesion protein of 417 amino acids. It contains two N-terminal immunoglobulin domains (IgV and IgC2), a transmembrane domain and an intracellular C-terminal domain (Capdevila-Nortes et al., 2015) (Figure 4B). GlialCAM tightly interacts with MLC1 (Capdevila-Nortes et al., 2013b), acts as an auxiliary subunit for ClC-2 chloride channels (Jeworutzki et al., 2012) and regulates Connexin-43 mediated gap-junctional coupling (Wu et al., 2016). Before GLIALCAM was linked to MLC, it was mainly known as a cancer gene (Moh et al., 2008; He et al., 2010; Zhang et al., 2011). A total of 29 unique variants in GLIALCAM were identified in patients with either classic or remitting MLC (Figure 4; Table 2).

The AQP4 gene is located on chromosome 18q11.2 and contains 5 exons and 4 introns (Figure 6A). It encodes the water channel aquaporin-4 (AQP4), the most abundant aquaporin in the brain (King et al., 2004). AQP4 has two main isoforms due to the use of two different translation initiation sites (Neely et al., 1999; Verkman et al., 2011): the longer M1 isoform and a shorter M23 isoform. Both isoforms, among other isoforms, are expressed in astrocytes and assemble as heterotetramers (Jorgacevski et al., 2020). AQP4 has 6 transmembrane domains and 2 highly conserved NPA domains that form the pore in the membrane (Ho et al., 2009) (Figure 6B). Both the C- and the N-terminus reside intracellularly. AQP4 is of vital importance for brain ion and water homeostasis (Nagelhus and Ottersen, 2013). It shares its location in astrocyte endfeet with MLC1 and GlialCAM (Nagelhus et al., 2004).

Recently, a recessive variant in AQP4 was found to lead to remitting MLC (Passchier et al., 2023). This is the first variant found in AQP4 to be disease linked and offers new insight into MLC disease mechanisms. The variant affects the key pore forming NPA motif in the AQP4 protein, exchanging an alanine, which is a hydrophobic amino acid, for a threonine, a hydrophilic amino acid (c.643G>A; p.(Ala215Thr) Figure 6A; Table 4). This leads to a loss of AQP4 at the cell membrane and therefore, a loss of function of the protein. Why an AQP4 defect would lead to remitting MLC is unknown. Passchier et al. describe that under certain conditions (e.g., massive overexpression) mutant AQP4 retains some function (Passchier et al., 2023). The physiological relevance of this is unclear, but this might form a basis for the observed radiological improvement. An alternative explanation could be redundancy in brain ion and water homeostasis, potentially with other aquaporins taking over the role of AQP4.

The GPRC5B gene is located on chromosome 16p12.3 and contains 4 exons and 3 introns. Exon 1 encodes the 5′UTR (Figure 6C). GPRC5B encodes the orphan G protein-coupled receptor (GPCR) GPRC5B. It is a class C GPCR and has 7 transmembrane regions and a relatively short extracellular N terminus compared to other GPCRs of the same class (Figure 6D). Multiple GPRC5B isoforms are known, including a long and a short isoform, as well as a brain specific isoform (Cool et al., 2010). GPRC5B is expressed in various tissues and cell types, including adipocytes, pancreatic islets, kidney podocytes and vascular smooth muscle cells. The receptor modulates inflammatory signalling through the NF-κB pathway, and has been linked to various disease states including diabetes (Soni et al., 2013), kidney disease (Zambrano et al., 2019) and artherosclerosis (Carvalho et al., 2020). Few studies have looked at GPRC5B function in the brain (Kurabayashi et al., 2013; Sano et al., 2018), but recently GPRC5B was identified as an MLC1/GlialCAM interacting protein that is expressed in astrocyte endfeet (Alonso-Gardon et al., 2021). Two dominant de novo variants in the fourth transmembrane region of GPRC5B were recently described in patients with classic MLC that had no variants in MLC1, GLIALCAM or AQP4 (Passchier et al., 2023). These two variants are specific amino acid duplications of two neighbouring amino acids in the fourth transmembrane domain of the GPCR (c.526_528dup; p.(Ile176dup) and c. 528_530dup; p.(Ala177dup); Figure 6C; Table 4). How these variants affect the function of GPRC5B is not yet understood.

Abnormalities of the brain white matter have been consistently associated with psychiatric disorders (Fields, 2008). Psychiatric symptoms are therefore often seen in leukodystrophies, even sometimes as presenting symptoms (Costei et al., 2021). A clear example is the frequent occurrence of psychosis in Metachromatic Leukodystrophy (MLD OMIM# 250100) (Hyde et al., 1992).

Regarding neurodevelopmental disorders, autistic features are common within the population of MLC patients (Lopez-Hernandez et al., 2011a). Surprisingly, autistic features are seen in ∼25% of patients with remitting MLC caused by dominant GLIALCAM variants, while this is only 9% for patients with classic MLC (Hamilton et al., 2018). Therefore, while motor symptoms are invariably milder in remitting MLC, the occurrence of autistic features is higher. The basis for this difference is not understood. It suggests that dominant GLIALCAM variants disrupt a function of GlialCAM that is separate from its role in astrocyte endfeet. For example, a recent study showed that astroglial release of GlialCAM from exosomes regulates neuronal axon outgrowth and dendritic spine formation (Jin et al., 2023). Both of these processes have been implicated in the etiology of autism (Gilbert and Man, 2017). While speculative, such a specialized function of GlialCAM might explain why dominant GLIALCAM variants more often lead to autistic features.

Genetic screens in cohorts of patients with autism spectrum disorder (ASD) uncovered additional heterozygous missense variants in GLIALCAM. The c.274C>T; p.(Arg92Trp) variant, which was previously found in a remitting MLC patient (Lopez-Hernandez et al., 2011a), was independently discovered in a study on ASD patients (Iossifov et al., 2014). The c.437C>T; p.(Ser146Leu) variant was discovered in another study on ASD patients (Li et al., 2017). Another new GLIALCAM variant that has not been observed in MLC patients, c.505T>C; p.(Ser169Pro), was found in a study on patients with intellectual disability (Lelieveld et al., 2016). Finally, a splice site variant in GLIALCAM, c.803+1G>A; p.? was found in an ASD cohort. This variant was classified as variant of unknown significance (Zhou et al., 2019). Together, these studies further substantiate the link between specific dominant GLIALCAM variants, ASD and intellectual disability.

For MLC1 it has been suggested that specific variants in heterozygous state are linked to schizophrenia or bipolar affective disorder, two psychiatric disorders thought to share a common etiology. A rare missense variant in MLC1 (c.1121C>A; p.(Leu309Met)) was associated with periodic catatonia, a familial subtype of catatonic schizophrenia (#OMIM605419) in a large pedigree (Meyer et al., 2001), although this variant was not found in other cohorts of schizophrenia or bipolar affective disorder patients (Meyer et al., 2001). Additional MLC1 variants have been significantly associated with schizophrenia and bipolar affective disorder in an Indian cohort (Devaney et al., 2002; Ewald and Lundorf, 2002; Jorgensen et al., 2002; McQuillin et al., 2002; Rubie et al., 2003; Kaganovich et al., 2004), with confirmation of two intronic MLC1 variants in an independent study, where they were specifically associated with periodic catatonia and not with other types of schizophrenia (Selch et al., 2007). Interestingly, variants linked to bipolar disorder or schizophrenia have not been observed in MLC patients. In addition, the c.1121C>A; p.(Leu309Met) variant has no influence on MLC1 protein expression levels in cellular studies (Teijido et al., 2004). Therefore, while these findings suggest that being a heterozygous carrier of specific MLC1 variants is associated with schizophrenia and bipolar disorder, the mechanistic link between MLC1 and psychiatry requires further research.

Finally, gene expression levels for MLC1 show potential association with depression and suicidal behavior. MLC1 transcript levels were significantly downregulated in post-mortem brain tissue of suicide victims (Thalmeier et al., 2008). A later study found reduced transcript levels in blood of patients with major depressive disorder (Spijker et al., 2010). Future studies confirming and extending the link between MLC1 transcript levels and suicidal behavior are required.

All MLC linked proteins are membrane proteins that are mainly expressed by glial cells in the brain (Zhang et al., 2016). MLC1 and AQP4 expression in the brain is restricted to astrocytes. GLIALCAM and GPRC5B are predominantly expressed by astrocytes, but both can also be detected in oligodendrocytes (Figure 7). Immunohistochemical studies on human brain tissue show that MLC1 and GlialCAM are highly enriched in astrocyte endfeet contacting the vasculature. In addition, both are found in other glial membranes at brain-cerebrospinal fluid interfaces, such as in subpial astrocyte endfeet and in ependymal cells (Boor et al., 2007; Lopez-Hernandez et al., 2011a). This location is shared with AQP4 (Nagelhus and Ottersen, 2013) and numerous other proteins involved in brain ion and water homeostasis, such as members of the dystrophin-associated glycoprotein complex (Boor et al., 2007), cation channels such as Kir4.1 (Higashi et al., 2001) and TRPV4 (Benfenati et al., 2011; Jo et al., 2015), chloride channels such as the essential volume regulated anion channel (VRAC) subunit LRRC8A (Bugiani et al., 2017; Formaggio et al., 2019), the chloride channel ClC-2 (Depienne et al., 2013) and the gap-junction protein Connexin-43 (Wu et al., 2016).

Direct or indirect interactions of many proteins that are enriched in astrocyte endfeet with MLC1 and/or GlialCAM have been described (for review see (Brignone et al., 2015)). Of particular importance is the early finding that a direct interaction between GlialCAM and MLC1 is required for the correct localization of both proteins to astrocyte-astrocyte junctions (Lopez-Hernandez et al., 2011a; Capdevila-Nortes et al., 2013b; Sirisi et al., 2014). A pathogenic variant in either of these proteins can disrupt the junctional localization of both. In addition, GlialCAM forms both cis and trans interactions with other GlialCAM proteins. Understanding the structural nature of these interactions likely holds the key to unravelling why certain GLIALCAM variants act in a dominant fashion while others are recessive (Elorza-Vidal et al., 2020). GlialCAM also directly interacts with, and acts as an auxiliary subunit for the chloride channel ClC-2. It is involved in both correct localization of the channel and in changing its gating kinetics (Jeworutzki et al., 2012). Recently, additional interaction partners of GlialCAM and MLC1 were identified, including the GPCRs, GPRC5B and GPR37L1 (Alonso-Gardon et al., 2021). Together, this information suggests that MLC1, GlialCAM and GPRC5B form an astrocyte endfoot complex interacting with ion channels, AQP4 water channels, gap junction proteins and other players involved in ion and water homeostasis.

Studying healthy human brain tissue has provided insight into the normal distribution of MLC related proteins and their interaction partners in the brain. However, the scarcity of post-mortem MLC patient brain material limits knowledge on how protein levels and protein distribution are altered in the disease context. Rare biopsy tissue has confirmed the linked localization of MLC1, GlialCAM and ClC-2 (Sirisi et al., 2014). Qualitative studies in animal models confirm this, but show that many other endfoot proteins have normal localization in MLC (Dubey et al., 2015; Bugiani et al., 2017). A quantitative study of the endfoot proteome in the context of MLC, in animal models or ideally in patient tissue, would be valuable to confirm and expand on these findings. In addition, how other cell types alter their protein make-up in the MLC disease context has been largely unstudied.

Pathological examination of a brain biopsy of an MLC patient revealed the existence of numerous vacuoles in the white matter (van der Knaap et al., 1996). These vacuoles are located in between the myelin lamellae, suggesting that they are the consequence of myelin splitting due to intramyelinic fluid accumulation. In the same tissue, perivascular astrocyte endfeet appear swollen and show intracellular vacuolization (Duarri et al., 2011). Together with the obvious swelling of the white matter observed on MRI, this has led to the hypothesis that MLC is characterized by a disruption of brain ion and water homeostasis (van der Knaap et al., 2012), leading to chronic white matter edema. The unraveling of the network of proteins interacting with MLC1, GlialCAM, GPRC5B and AQP4 described in section 4.1 supports this hypothesis.

Important mechanistic insight into the function of MLC1 and GlialCAM came from the discovery that lymphoblasts from MLC patients and astrocytes from MLC mouse models show a defect in volume regulation. Upon hypotonicity induced cell swelling most mammalian cells, including astrocytes, recover from swelling through a process known as regulatory volume decrease (RVD) (Jentsch, 2016). Lymphoblasts from patients with bi-allelic pathogenic MLC1 variants, as well as astrocytes isolated from Mlc1-null mice show reduced RVD when compared to healthy cells (Ridder et al., 2011; Dubey et al., 2015). A similar disruption of RVD was recently observed in lymphoblasts from patients with either of the dominant GPRC5B variants (Passchier et al., 2023). In addition, loss of AQP4 in astrocytes also disrupts the RVD process (Benfenati et al., 2011).

A key question is how MLC proteins are involved in the process of astrocyte volume regulation. The RVD process involves concerted activity of channels and transporters to orchestrate the net efflux of osmolytes from the cells, leading to associated water efflux through osmosis. A key channel involved in RVD is VRAC, a channel permeable to anions and small osmolytes which opens upon cell swelling (Jentsch, 2016). Loss of MLC1 or GlialCAM from cells reduces VRAC activity (Ridder et al., 2011; Capdevila-Nortes et al., 2013b; Dubey et al., 2015), and overexpression of MLC1 in various cells increases VRAC activity (Ridder et al., 2011). GPRC5B was also recently shown to regulate VRAC activity. Overexpression of GPRC5B, either mutant or wild-type, leads to increased VRAC activity in astrocytoma cells (Passchier et al., 2023), while GPRC5B knockdown in primary astrocytes reduces VRAC activity (Alonso-Gardon et al., 2021).

The observation that MLC1 shows minor homology to ion channels initially led to the speculation that MLC1 might be the long sought protein that forms the VRAC channel. Subsequent identification of LRRC8A-E subunits as the subunits of VRAC (Qiu et al., 2014; Voss et al., 2014) put this hypothesis to rest, and allowed investigation of molecular interactions between VRAC subunits, MLC1 and GlialCAM. This revealed that a direct molecular interaction between these proteins is lacking (Elorza-Vidal et al., 2018; Alonso-Gardon et al., 2021). Instead, the interaction between MLC1, GlialCAM, GPRC5B and VRAC activity likely involves an intermediate cellular process that is not fully resolved. Many signaling cascades and cellular properties are known to modulate VRAC functioning in different cell types (for review see (Jentsch, 2016)). A recent study has suggested that intracellular calcium might be an intermediate (Brignone et al., 2022), through CaMKII-mediated phosphorylation of MLC1. AQP4 forms a complex with TRPV4 and together these proteins affect RVD through a functional interaction with VRAC (Benfenati et al., 2011). Upon osmotic changes, water entry or exit from astrocytes is accelerated by the presence of AQP4. Rapid volume changes greatly enhance the activation of VRAC channels (Benfenati et al., 2007). Additionally, hypo-osmotic challenges increase expression of AQP4 at the plasma membrane (Kitchen et al., 2015). The membrane expression of LRRC8A proteins and AQP4 show mutual dependence in astrocytes (Liu et al., 2023). Therefore, MLC linked proteins converge on regulation of VRAC activity in RVD. Further mechanistic insight into how VRAC function is disrupted in MLC is a crucial step in our understanding of the disease and may aid in directing therapy development.

In addition to the effect on VRAC, MLC1 and GlialCAM interact with various other ion channels and pumps involved in brain ion and water homeostasis. As mentioned earlier, a direct interaction of GlialCAM with the chloride channel ClC-2 was described, with GlialCAM acting as an auxiliary channel subunit, potentiating channel function and changing desensitization kinetics (Jeworutzki et al., 2012). However, since neurological disease caused by loss of ClC-2 function differs substantially from MLC, both clinically and on MRI (Depienne et al., 2013), the role of ClC-2 in the pathogenesis of MLC is not well understood. Furthermore, MLC1, as well as AQP4, directly interacts with the mechanosensitive cation channel TRPV4 (Benfenati et al., 2011; Lanciotti et al., 2012; Jo et al., 2015). The nature of this interaction and its relevance in MLC pathogenesis is not well understood. Alonso-Gardon et al. (2021) described direct interaction between MLC1 and GPRC5B and between GLIALCAM and GPRC5B. They also showed interaction of GlialCAM with another GPCR, GPR37L1. Research into these GPCRs could give new insight into signaling pathways that are dysregulated in MLC.

As mentioned, there is large variability in disease severity between patients and a lack of a clear phenotype/genotype correlation. Because of the many interactions of MLC proteins with other proteins involved in ion and water homeostasis, an important open question is whether there are genetic modifiers of disease severity. Variants in genes involved in brain ion and water homeostasis, which by themselves do not lead to disease, might modulate disease severity in MLC patients. To illustrate this, the presence of several common single-nucleotide polymorphisms in AQP4, which are linked to lower levels of AQP4, have been associated with altered EEG activity during sleep in the healthy population (Ulv Larsen et al., 2020). It will be interesting to see whether the presence of such common or less common haplotypes, which constitute a polygenic risk for brain ion and water homeostasis, correlates with disease severity in MLC patients.

In conclusion, MLC1, GlialCAM, GPRC5B and AQP4 appear to be central in the organization of an osmoregulatory complex in astrocyte endfeet. The disruption of this complex in MLC hampers astrocytes in keeping ion and water homeostasis, thereby causing fluid accumulation in myelin and astrocytes.

Several animal models for MLC have been generated. Mlc1-null mice and Glialcam-null mice both recapitulate key disease features: increased brain water content and progressive myelin vacuolization (Hoegg-Beiler et al., 2014; Dubey et al., 2015; Bugiani et al., 2017). However, while in patients vacuolization and swelling is most apparent in the cerebral white matter outside the corpus callosum and to a lesser degree in the cerebellar white matter (van der Knaap et al., 1995b), mice have very little cerebral white matter beyond the corpus callosum and white matter vacuolization in MLC mice is more readily observed in cerebellar white matter and corpus callosum (Dubey et al., 2015; Bugiani et al., 2017; Hoegg-Beiler et al., 2014). Patients develop white matter swelling with megalencephaly in the first year of life, replicated in MLC mice. MLC patients typically develop motor and cognitive decline with a delay of 4–6 years. The short life span of mice probably explains why MLC mice show no obvious behavioral phenotype (Dubey et al., 2015), although detailed investigation does reveal hind limb clasping when the animals are lifted by the tail (Dubey et al., 2018). While the mice show no overt seizure phenotype, recording electrical brain activity in freely moving mice uncovered increased interictal spike occurrence. Furthermore, the threshold for kainate-induced seizure induction is lowered (Dubey et al., 2018). Therefore, similar to MLC patients the mice have a seizure phenotype.

Thus, Mlc1-null mice and Glialcam-null mice replicate early stages of classic MLC (Bugiani et al., 2017). A mouse model based on a dominant Glialcam variant has also been generated (Hoegg-Beiler et al., 2014). As expected, the vacuolization phenotype observed in mice heterozygous for this dominant variant was mild.

A mouse overexpressing wild-type MLC1 has also been generated. Interestingly, this overexpression mouse shows white matter vacuolization similar to classic MLC mouse models, but with a more rapid onset (Sugio et al., 2017). As outlined in section 4.2, MLC linked proteins are part of an osmoregulatory complex in astrocyte endfeet. It is likely that ion channels, pumps and transporters involved in these processes need to be carefully tuned, with deviation in their activity in either direction leading to disease. The fact that both a decrease and an increase in expression levels of MLC1 can lead to a similar phenotype is in line with this idea.

Two zebrafish models for MLC have been generated and characterized. Knockout of the zebrafish orthologs of MLC1 (mlc1) or GLIALCAM (glialcama) led to macrocephaly, but vacuoles were not observed in the brain of adult MLC zebrafish (Sirisi et al., 2014; Perez-Rius et al., 2019).

Long before the clinical discovery that loss of AQP4 function leads to remitting MLC (Passchier et al., 2023), Aqp4-null mice had been generated and extensively characterized (Ma et al., 1997). Similar to Mlc1-null and Glialcam-null mice, Aqp4-null mice show increased brain water content (Vindedal et al., 2016) and an alteration in seizure threshold (Binder et al., 2004; Binder et al., 2006). The alteration in seizure threshold described for Aqp4-null mice differs from what is seen in other MLC mouse models and in patients: Aqp4-null mice have an increased threshold for evoked seizures, but when seizures occur they are longer in duration. Whether this is due to slight differences in experimental conditions remains to be explored.

In conclusion, multiple animal models for MLC have been generated and characterized. These models do not show an exact phenocopy of the human disease, mainly with respect to most prominently affected white matter structures and a reduced severity of overt neurological phenotypes. However, central MLC related phenotypes can be observed in these models, and they enable researchers to study MLC pathomechanisms in the intact brain. This has allowed confirmation of the hypothesis that disturbed astrocyte volume regulation leads to disrupted extracellular potassium regulation in MLC: extracellular potassium dynamics are altered in Mlc1-null and Glialcam-null mice (Dubey et al., 2018). Disrupted potassium homeostasis is also observed in Aqp4-null mice (Binder et al., 2006; Haj-Yasein et al., 2015). In addition, Mlc1-null mice were used to uncover delayed maturation of perivascular astrocyte processes, altered astrocyte morphology and polarity, reduced vascular smooth muscle cell contractility and disturbed neurovascular coupling and parenchymal flow in MLC (Gilbert et al., 2021). MLC mouse models also allow for preclinical testing of therapeutic interventions. To date, this has been only done in the context of gene therapy. Viral transduction with wild-type MLC1 targeted to cerebellum could revert cerebellar white matter vacuolization in Mlc1-null mice (Sanchez et al., 2020). The use of MLC mouse models for preclinical testing of other interventions holds great promise for the future search for an MLC therapy.

Curative therapy for MLC is still lacking. The fact that white matter abnormalities in some MLC patients show normalization holds promise for treatment, since it indicates that myelin vacuolization in MLC is in principle reversible.

For monogenic diseases like MLC restoring the causative genetic defect through gene editing or overexpression of a healthy gene copy could be considered. A first proof-of-concept gene therapy study showed that cerebellar white matter vacuolization can be prevented or reversed when MLC1 levels are virally restored in the cerebellum of Mlc1-null mice (Sanchez et al., 2020). However, while technological advances in gene therapy are fast, it will take time before the risks associated with gene therapy (immune reactions, cancers, tissue damage upon intracranial delivery) can properly be addressed. Since clinically MLC is a relatively mild disease, it is important to carefully weigh these risks with therapeutic benefit. Furthermore, for effective gene therapy it is likely necessary to target astrocytes throughout the brain, which is not feasible currently. Finally, overexpression of Mlc1 in an animal model causes an MLC-like phenotype, indicating that expression levels of MLC related proteins likely need to be carefully tuned. These factors pose an additional challenge for gene therapy approaches. However, the observation that low residual levels of MLC1 might lead to mild or remitting disease offer new leads for tuning expression levels.

Alternatively, traditional pharmacological targeting of disrupted cellular processes in the MLC brain might hold promise. This relies on detailed mechanistic understanding of the disease and identification of suitable therapeutic targets. Ion and water channels involved in brain fluid homeostasis (VRAC, TRPV4, AQP4) can be considered as targets. The recent identification of two GPCRs that interact with MLC1 and GlialCAM raises additional exciting possibilities (Alonso-Gardon et al., 2021; Passchier et al., 2023), especially since modulation of GPCRs is the mechanism of action for ∼1/3 of current FDA approved drugs (Chan et al., 2019). Unravelling the underlying signalling pathways, how they are disrupted in MLC, and whether pharmaceutical modulators can be identified should be the focus point of future mechanistic studies. Advances in gene therapy or pharmacological approaches can then be followed by preclinical tests in MLC mouse models, which will hopefully pave the way for much needed therapy for MLC.