Golgipathies in Neurodevelopment: A New View of Old Defects

One hundred and twenty years ago, in 1898, Camillo Golgi [1] first published his observations of an “internal reticular complex” that surrounded the nucleus of Purkinje cells in the barn owl cerebellum, using his now eponymous silver reduction method. Despite Golgi’s demonstration of his reticular complex in a variety of mammalian cell types in addition to neurons, the very existence of this complex, which we now call the Golgi complex or Golgi apparatus (GA), as an independent and consistent intracellular entity was in doubt for much of the first half of its history, in large part because of its variable morphology and technological and methodological limitations. Also in doubt was what specific role, if any, this strange structure could possibly play. We have known for a few decades now that the GA not only exists but is ubiquitous in all eukaryotic cells, from the yeast upward. However, our views on what it does and how it does it continue to evolve.

The best known and most widely accepted roles of the GA are of course the biosynthesis of some lipids such as glycolipids and sphingomyelin, and the processing, sorting, packing, routing, and recycling of secretory cargo downstream of the endoplasmic reticulum (ER). These roles are important for cellular and organ function throughout the life span of a cell. However, the GA plays several other roles, many of which are of particular importance during development. It participates in cell division, migration, morphogenesis, and growth (all of which involve diverse aspects including intracellular organization, polarity, and compartmentalization) as well as apoptosis, autophagy, and the stress response. In keeping with these multiple roles, many “GA proteins” are also present in other intracellular compartments and involved in more than one function, including acting as signaling molecules, rather than acting at a purely structural/mechanical or enzymatic level.

The ubiquitous nature of the GA means that any pathogenic mutation perturbing the function of a GA protein could, and mostly does, result in clinical repercussions across multiple tissues and systems. Strikingly, of the mutated GA-associated genes identified in various monogenic disorders so far, > 40% affect the normal development or functioning of the central or peripheral nervous system, a far greater proportion than for any other tissue type [2], highlighting its crucial role in different types of neural cells. Indeed, the magnitude of this impact on the nervous system structure and physiology, and on neurodevelopment in particular, should not be surprising, when one considers the intensive use that neurons and glia presumably make of the GA. This use includes the intricately coordinated processes involved in progenitor cell organization, orientation and mitosis; the subsequent highly directional migration of young neurons to their target regions; the exponential increase in surface area and thus of the biosynthesis of both membrane lipids and proteins required by diverse processes such as dendritic arborization in neurons and myelin sheath production in oligodendrocytes; the heavy demands of synaptogenesis and synaptic and neurosecretory activity, etc. The existence of Golgi outposts in dendrites is a prime example of the adaptation of the GA to the highly compartmentalized nature of the nervous system and its specific trafficking needs (review [3, 4]).

Intriguingly, in the course of our identification of the causal gene and molecular mechanisms underlying Dyggve-Melchior-Clausen syndrome (DMC, MIM #223800), a disorder in which skeletal defects are associated with a characteristic neurological profile combining postnatal-onset microcephaly (POM), white-matter defects, and intellectual disability (ID) [5], we noted the presence of the same combination of symptoms in several other monogenic disorders, each of which was linked to a defect in a gene involved in the GA trafficking machinery. This common profile indicated a similar pathophysiology, i.e., defective neuronal or oligodendrocytic maturation at the postmitotic stage, which leads to symptoms later in neurodevelopment than in the primary or congenital microcephalies which result principally from proliferation deficits. Additionally, several of these newly or previously identified disorders appeared to involve a mutation in a GA-associated member of the small RAB GTPase family which is involved in a plethora of trafficking functions in the cell, or in molecules that interact with the RABs including their direct effectors and modulators such as the guanine nucleotide exchange factors (GEFs) which activate them by replacing GDP with GTP and allowing them to bind specific effectors, or the GTPase-activating proteins (GAPs) which dissociate the active complex.

These observations prompted us to propose the term “Golgipathies” or “Golgipathic microcephalies” to designate this emerging class of disorders that is characterized by a similar phenotype (POM, white-matter defects, notably an abnormal corpus callosum, and varying degrees of cognitive impairment), and is caused by mutations in RAB GTPases or their molecular partners [4]. However, an examination of the literature and recently published studies provides ample evidence that defective GA proteins and processes are linked to a wide spectrum of neurodevelopmental defects, in addition to the combination that first drew our attention. Not only are there a large number of isolated reports of POM related to GA- or RAB-associated mutations, but a number of other disorders, including several characterized by congenital microcephaly, could also justifiably be classified as Golgipathies, based on the fact that the primary pathogenic mechanism in all of them is a defective GA protein, regardless of its intracellular function (GA structure or intracellular organization, trafficking machinery in the anterograde or retrograde directions, posttranslational modifications, cell-cycle involvement, etc.). In the light of these new insights, we would like to refine the notion of Golgipathies, taking into account their major neurodevelopmental involvement, with a brief discussion of the relationship between these disorders and the possible GA functions affected.

Despite their great variety, Golgipathies with a neurodevelopmental component also show considerable overlap and a number of unifying features. Some of these putative neurodevelopmental Golgipathies are described below, grouped according to their principal clinical signs (the presence or absence of microcephaly and its onset), and the probable function or position in the trafficking pathway of the affected protein, in order to stimulate discussion as to their potential classification. We have excluded cancer-related GA genes/proteins or those primarily involved in neurodegenerative processes.

Perhaps due to the fact that the most important and “permanent” function of the GA is lipid and protein trafficking, crucial throughout the life span of the cell, the POMs constitute a disproportionately large group of Golgipathies affecting neurodevelopment. While many involve other organs and tissues and show remarkable differences in phenotype between one individual and the next, for the purpose of this review we concentrate on the neurological aspects (summarized in Table 1).

Dyggve-Melchior-Clausen syndrome (DMC; MIM #223800), which first prompted our speculation about Golgipathies as a class of disorders, is an autosomal recessive skeletal dysplasia associated with POM and ID and is caused by loss-of-function mutations in the DYM gene encoding DYMECLIN, a GA protein involved in intracellular trafficking [6-8]. Brain MRI reveals a marked thinning of the corpus callosum and brain stem, supported by significant deficits in the volume and structure of myelin in Dym–/– mutant mice [5] Dym-deficient neurons display a fragmented GA and impaired ER-to-GA trafficking [5], but Dym may also play a role in the retrograde transport of vesicles from the GA to the ER [7]. Several lines of evidence suggest that it has a tethering role during vesicle trafficking between the ER and the GA [6, 7] in conjunction with the golgin GIANTIN and interactions with RAB1 or RAB6, the most abundant type of RAB [9-11]. Interestingly, Smith McCort dysplasia, a clinical variant of DMC with identical skeletal defects but normal intelligence and no microcephaly, has been found to result either from specific missense mutations in DYM that could result in some residual activity of the protein (SMC1; MIM #607326) [6, 12] or from loss-of-function mutations in the RAB33B GTPase, also localized in the GA (SMC2; MIM #615222) [13, 14].

Warburg-Micro syndrome (WARBM1–4; MIM #600118, #614225, #614222, #615663) is an autosomal recessive disorder [15] characterized by neurodevelopmental defects including POM with profound ID, and progressive limb spasticity associated with progressive peripheral axonal neuropathy [16], severe visual impairment, and hypogonadism. Brain MRI shows bilateral frontal polymicrogyria and hypoplasia of the corpus callosum and cerebellar vermis [17, 18]. Loss-of-function mutations in 4 distinct genes, RAB3GAP1, RAB3GAP2, RAB18, and TBC1D20, have been implicated in WARBM [18-20]. RAB18 has been confirmed to localize in the ER and the cis-Golgi compartment [21], in addition to other compartments. While TBC1D20 acts as a GAP for RAB18, in addition to acting on ER-localized RAB1 [22] the -RAB3GAP complex in this context acts not as a GAP but as a GEF for RAB18 [23]. In the related Martsolf syndrome (MIM #212720), only RAB3GAP2 mutations have been identified, but most patients also display severe neurodevelopmental defects including POM and white-matter defects (where examined), ID, hypogonadism, and ocular defects [19, 24, 25].

Mutations in COL4A3BP (collagen 4A3-binding protein, also known as Goodpasture antigen-binding protein, GPBP, or ceramide transporter, CERT) lead to an autosomal dominant mental retardation (MRD34; MIM #616351). The COL4A3BP protein is thought to mediate ER-to-GA transport of ceramide, an essential lipid component. One of the first patients identified with a mutation in this gene presented with POM in addition to other neurological and skeletal defects [26].

Autosomal recessive mental retardation 13 (MRT13; MIM #613192) is linked to loss-of-function mutations in TRAPPC9. Initially considered a nonsyndromic autosomal recessive ID [27-29], with additional cases, MRT13 is beginning to show a fairly distinctive phenotype, including moderate-to-severe POM, a peculiar facial appearance, obesity, and hypotonia. MRI reveals a reduction in cerebral white matter, with a marked thinning of the corpus callosum [30-32].

TRAPPC9 is a subunit of the trafficking protein particle (TRAPP) complex that mediates the tethering of COPII-coated ER-derived vesicles to cis-Golgi membranes [33]. The TRAPP complex acts by activating RAB1, which in turn recruits specific cis-Golgi effectors such as the golgins P115 and GM130 to allow vesicle tethering [34]. The TRAPP complex and particularly TRAPPC9 are also involved in the interaction between COPII-coated vesicles and the microtubule motor protein dynactin 1 (DCTN1, equivalent to the Drosophila p150^Glued) [35].

Progressive childhood encephalopathy (PEBAS; MIM #617669), also linked to mutations in a TRAPP subunit, TRAPPC12, has been identified in 2 families as being linked to progressive microcephaly (including 1 case of POM), agenesis of the corpus callosum and other MRI abnormalities, and severe developmental delays [36]. Mutations in another TRAPP subunit, TRAPPC6B, have also been linked to a POM syndrome (NEDMEBA; MIM #617862) in several families in a recent report [37].

In Takenouchi-Kosaki syndrome (TKS; MIM #616737), caused by mutations in the GA GTPase CDC42 which regulates bidirectional Golgi transport by regulating the cargo-sorting and carrier-formation functions of COPI [38], patients show a range of phenotypes affecting different organs, including ID and either POM or congenital microcephaly (review [39]).

A neuromuscular syndrome with microcephaly noticeable from the 4th postnatal month has been described in an individual who also displayed developmental delays, seizures, hypotonia, and muscular dystrophy [40]. Brain MRI revealed delayed myelination and a thinning of the corpus callosum. The molecular cause was a loss of function of the GOLGA2 gene (MIM #602582), which encodes GM130, a multifunctional golgin involved in both the assembly/maintenance of GA structure and the regulation of the secretory pathway and vesicle tethering to the cis-Golgi compartment through interactions with RAB1B, P115, etc.; GM130 also binds to other RAB proteins involved in membrane traffic regulation at the ER/Golgi interface, such as RAB2 and RAB33B [41-43].

Autosomal recessive mental retardation 61 (MRT61; MIM #617773), also called Alwadei syndrome, is a POM syndrome that includes profound ID, epilepsy (in 2 cases), and a thin corpus callosum [44]. It has been found to be caused by mutations in IPORIN (MIM #617773), a protein encoded by the RUSC2 gene. Little is known about the RUSC2 protein except that it is ubiquitous and highly expressed in the brain, and it interacts with both RAB1B and, possibly directly, with GM130 [45].

Cohen syndrome (COH; #MIM 216550) is an autosomal recessive disorder characterized by motor delays, progressive retinal dystrophy, and severe myopia, hypotonia, joint hypermobility, and progressive POM associated with ID [46], and it is associated at times with nonneurological signs [47]. Brain MRI reveals a relatively thicker but more compact corpus callosum in some patients, associated with markedly smaller sagittal diameters of the brain stem [48]. COH1, the only gene associated with Cohen syndrome so far, encodes the vacuolar protein-sorting protein VPS13B, a large peripheral membrane protein that is actively recruited at the GA by RAB6 [49, 50]. VPS13B likely plays a specific role in the dynamics and function of the GA, in particular during neuronal maturation.

Progressive cerebellocerebral atrophy type 2 or pontocerebellar hypoplasia type 2E (PCCA2/PCH2E; MIM #615851) is an autosomal recessive neurodegenerative disorder characterized by normal development during the initial months of life, followed by motor delays, progressive POM, progressive spasticity, and epileptic seizures by the age of 2 years [51]. Brain MRI reveals a gradual decrease in cerebral white matter associated with delayed myelination and thinning of the corpus callosum [51]. The responsible gene encodes VPS53, another vacuolar protein-sorting protein that participates in the transport and recycling of endosome-derived transport vesicles [52]. VPS53 is part of 2 large multi-subunit complexes named Golgi-associated retrograde protein (GARP) and endosome-associated recycling protein (EARP), that ensure proper tethering between endosomes and their acceptor compartment [53, 54]. Both complexes cooperate with SNAREs for subsequent membrane fusion. GARP has been found to interact with RAB6A in the trans-Golgi network [55] and EARP associates with RAB4-containing vesicles [54]. A previously unknown neurodevelopmental disorder linked to a mutation in yet another GARP/EARP subunit, VPS51, also appears to exist, based on our own observations (case report in prep.) as well as a recently published description of a 6-year-old patient with severe global developmental delay, pontocerebellar abnormalities, microcephaly, hypotonia, epilepsy and several systemic and peripheral dysfunctions [216].

A pontocerebellar hypoplasia caused by TBC1D23 mutations (PCH11; MIM #617695) has been shown in 2 different reports to be associated with POM or microcephaly of unknown onset, along with various other neurodevelopmental deficits and neurological signs (including a hypoplastic corpus callosum, as seen in several Golgipathic POMs, severe cognitive delays, motor weakness or lack of motricity, behavioral problems, etc.) [56, 57]. TBC1D23, like TBC1D20 in WARBM, appears to be a RAB GAP (although neither its GTPase activity nor its RAB specificity has been clearly demonstrated so far). It is localized at the trans-Golgi and regulated by the small GTPases ARL1 and ARL8, and it appears to mediate the binding of endosomal vesicles to golgin-245 and golgin-97 on trans-Golgi membranes to mediate retrograde transport [58]. It has been implicated in the regulation of neuronal migration/positioning during corticogenesis, as well as neurite/axon differentiation [56].

Autosomal recessive periventricular heterotopia with microcephaly (ARPHM; MIM #608097) has been described in several patients to be caused by ARFGEF2 mutations. The microcephaly in these reports was progressive postnatally in 3 reports [59-61], and of unknown onset/progression in 1 [62]. The patients also displayed severe developmental delays and ID, epilepsy, brain atrophy, and myelination delays associated with a thin corpus callosum. ARFGEF2 encodes BIG2, a GA protein responsible for interior membrane trafficking in the trans-Golgi network and endosomes [63]. BIG2 interacts with RAB11, a trans-Golgi protein involved in diverse functions such as synaptic transmission, but also in neuronal migration and axonal growth, functions that could explain the phenotypes seen in the reports above [64, 65].

MEDNIK syndrome (MEDNIK; MIM #609313) is caused by mutations in AP1S1, a subunit of 1 of 5 adaptor protein complexes (AP1–5) located at the trans-Golgi and in endosomes that match cargo molecules, including neurotransmitters (and in the case of AP1S1, specifically, the copper pumps ATP7A and B), to their carriers (review [66]). Symptoms of MEDNIK, although predominantly cutaneous, include POM in addition to other typical neurological signs such as moderate-to-severe ID and deafness [67]. Montpetit et al. [68] also mention the occurrence of peripheral neuropathy and microcephaly in the original French-Canadian cohort, without indicating age of onset, when discussing their animal model, although the original article by Saba et al. [69] and subsequent discussions of this cohort only mention brain “atrophy” [67, 68]. In addition, mutations in AP1S2 caused Pettigrew syndrome (PGS, MIM #304340), an X-linked disorder characterized by ID, seizures, Dandy-Walker malformation and microcephaly, possibly postnatal, although head circumference was variable in the original family [70, 71].

COG-associated congenital disorders of glycosylation (CDGs) represent 1 subgroup of a very large family of multisystemic autosomal recessive pathologies involving dysfunctions in the processing of N- and O-linked glycans. While most of the mutations identified so far involve genes encoding glycosylation enzymes [72], the CDGs described in this paragraph are caused by defects in 1 of the 8 subunits of the conserved oligomeric Golgi (COG) complex, localized to the cis- and medial-Golgi as well as surrounding vesicles [73]. The COG complex is thought to act as a tethering factor, in particular during intra-GA and retrograde GA-to-ER trafficking, where it mediates the recycling of glycosyltransferases [74], suggesting that the incidence of CDG when COG subunits are mutated is likely due to the inability of these glycosylation enzymes to reach their target proteins [73]. COG-associated CDGs are associated with multiple neurological manifestations, including POM, in patients carrying mutations in COG1 (MIM #611209), COG2 (MIM #617395), COG4 (MIM #613489), COG5 (MIM #613612), COG7 (MIM #608779), and COG8 (MIM #611182) [75-81]. Brain MRI revealed hypoplasia of the corpus callosum in 4 patients [77, 78] and brainstem atrophy in 1 case [75]. COG6 mutations have also been associated with borderline POM in Shaheen syndrome (SHNS; MIM #615328) as well as a CDG with congenital microcephaly (MIM #614576) [82, 83]. Interestingly, the COG complex has been shown to interact with molecules at all levels of GA organization and trafficking as well as with a number of GA-associated RABs (overview [4]).

Another family of GA-associated proteins whose mutation gives rise to a wide range of CDGs with neurological phenotypes is the solute carrier family 35 (SLC35), whose members carry a variety of molecules, mostly sugars, to the GA for the glycosylation of proteins. Mutations in SLC35A3 have been associated with POM and ID [84] in a syndrome characterized by arthrogryposis, mental retardation, and seizures (AMRS; MIM #615553). Similarly, mutations in SLC35C1 give rise to CDG2C (MIM #266265), in which severe ID and POM have also been noted [85], while mutations in other family members give rise to various CDGs with ID, microcephaly of unknown onset, and other neurological manifestations (e.g., SLC35A1: CDG2F, MIM #603585; SLC35A2: CDG2M, MIM #300896).

Among CDGs caused by a defective trafficking mechanism, rather than an enzymatic or metabolic deficiency, and including POM, psychomotor delays, and white-matter abnormalities among their clinical signs, CDG2K should also be included; it is caused by mutations in the transmembrane GA protein TMEM165 (MIM #614727). While the exact function of this protein is not yet known, it is probably an ion transporter that plays a role in pH, calcium, or manganese homeostasis, and mutations also appear to affect GA morphology [86, 87].

A number of other reports of patients with microcephalies linked to mutations in GA-associated proteins exist in the literature. However, data on whether these microcephalies are congenital or postnatal (or congenital but progressive after birth) are not always available. Further studies might well turn up other syndromes where a POM or a progressive worsening of microcephaly would indicate a defect in maturation (including trafficking or biosynthetic defects), or other processes that take place later in development, such as gliogenesis, as opposed to a congenital microcephaly that would indicate a deficit in the production of new neurons earlier in development, even though these different microcephalies are likely to lie along the same continuum rather than being mutually exclusive. The other side of the coin is that mutations in several GA-related proteins appear to result in macrocephaly, possibly indicating an opposite effect on the regulation of neurogenesis or neuronal apoptosis during maturation. A few examples of these other microcephalies and macrocephalies linked to GA protein mutations are given below (and summarized in Table 2).

Mutations in the WDR62 gene, which codes for a WD40 repeat protein highly expressed in the developing brain, cause autosomal recessive primary microcephaly 2 (MCPH2; MIM #604317) with seizures and ID in addition to congenital microcephaly [88, 89]. Interestingly, even though it is best known for its presence at the centrosome during mitosis and appears to play a critical role in neuronal progenitor proliferation during corticogenesis, WDR62 also associates with GM130 and is localized at the GA during interphase, where its function is unknown [89]. Similarly, at least 2 other MCPH genes encode products that associate with the GA. Autosomal recessive primary microcephaly 3 (MCPH3; MIM #604804) is related to mutations in the CDK5 regulatory subunit-associated protein 2, CDK5RAP2 [90], which plays an important role in progenitor division during neurogenesis [91, 92]. Interestingly, CKD5RAP2 is necessary for microtubule nucleation and organization not only at the centrosome but also at the GA [93, 94]. Similarly, our group and others have found that autosomal recessive -primary microcephaly 17 (MCPH17; MIM #617090) is caused by mutations in the gene for the Citron Rho-interacting serine/threonine kinase, CIT [95, 96]. The longer CIT-K protein, which localizes at the cleavage furrow and midbody during cell division (review [97]), is known from animal models to play a role in cytokinesis and progenitor survival during neurogenesis [98], while the shorter brain-specific isoform, CIT-N, is known to be enriched at the somatic GA of certain neurons as well as in dendrites where it interacts with peripheral Golgi outposts and plays a key role in the maintenance of dendritic spines [99, 100].

Mutations in TRAPPC11, a subunit of the TRAPP complex that plays a role in ER-GA trafficking and activates several RABs, as described above, lead to limb girdle muscular dystrophy (LGMDR18; MIM #615356) with microcephaly of unknown onset, myopathy, infantile hyperkinetic movements, ataxia, and ID [101, 102].

A microcephalic dwarfism syndrome (SRMMD; MIM #617164) characterized by, in addition to microcephaly, facial dysmorphism, severe micrognathia, rhizomelic shortening, and mild developmental delay, has been associated with heterozygous mutations in ARCN1 [103]. ARCN1 (COPD) encodes ARCHAIN 1, the delta subunit of the vesicular coat protein COPI, necessary for both intra-GA and GA-ER retrograde transport, and ARCN1 haploinsufficiency leads to COPI-mediated transport defects. In addition, a hypomorphic missense mutation in the COPB2 gene, encoding the beta subunit of COPI, has also recently been found to be the cause of a primary autosomal recessive microcephaly (MCPH19; MIM #617800) in which the congenital microcephaly is associated with severe developmental delay, cortical blindness and spasticity [104]. COPB2 appears to be involved not only in retrograde GA-ER transport, possibly interacting with DYMECLIN [7], but in early neuronal proliferation and apoptosis [104].

Progressive encephalopathy with edema, hypsarrhythmia, and optic atrophy-like syndrome (PEHOL; MIM #617507) caused by a mutation in the CCDC88A gene coding for GIRDIN, a coiled-coil GA protein, shows, in addition to profound psychomotor delay, seizures and progressive brain atrophy, severe congenital microcephaly in all patients, and dysmyelination in 1 case [105]. GIRDIN, also known as KIAA1212, appears to be involved in the migration and positioning of new neurons, controlling soma size and dendritic branching [106].

AP4-deficiency syndrome is caused by a mutation in a subunit of another of the 5 adaptor protein complexes, AP1–5, mentioned above (review [66]). AP4-deficiency syndrome, which includes a very large range of phenotypes, may be included in the large number of hereditary spastic paraplegias [107]. Among them, the AP4E1 mutation (MIM #613744) leads to a form of hereditary spastic paraplegia with congenital microcephaly, ID, thin corpus callosum, psychomotor deficits and spasticity [108-110], which Moreno-de-Luca et al. [110] found to be similar to another hereditary spastic paraplegia caused by mutations in AP4M1 (MIM #612936), in which several patients displayed microcephaly in addition to other neurological signs [111]. Similarly, Abou Jamra et al. [108] found microcephaly, along with a variety of other symptoms in patients with AP4B1 (MIM #614066) and AP4S1 mutations (MIM #614067). Surprisingly, microcephaly of unknown onset associated with ID has also been noted in Hermansky-Pudlak syndrome 2 (HPS2; MIM #608233), not commonly thought to be associated with neurodevelopment; in this disorder, mutations in AP3B1 principally lead to oculocutaneous albinism, immunodeficiency, and hematological abnormalities [112].

As in the case of the TRAPP and AP protein families, mutations in another SLC family member, SLC9A6, are also associated with other forms of microcephaly (congenital, progressive, or of unknown onset) and severe ID in the X-linked mental retardation known as Christianson syndrome (MRXSCH; MIM #300243) [113-115].

In early infantile epileptic encephalopathy (EIEE49, MIM #617281), caused by mutations in DENND5A (a gene encoding a GA-localized GTPase activator thought to play a role in neuronal differentiation), almost all affected individuals show microcephaly in addition to various other neurological signs, including seizure onset in the neonatal period, global developmental delay with ID and lack of speech, hypotonia, spasticity, etc. [116, 117]. DENND5A acts as a GEF for the neuronal GA-specific GTPase RAB39 [118], which is involved in regulating the number and morphology of neurite growth cones, and in synapse formation and maintenance, including alpha-synuclein homeostasis. However, deficits in RAB39B are also associated with an X-linked form of ID with macrocephaly rather than microcephaly (MRX72; MIM #300271) [119, 120]. Macrocephaly is also a feature of RAB39B-associated early-onset Parkinsonism or Waisman syndrome (WSMN; MIM #311510) [121, 122].

Mutations in another GTPase activator, HERC1, which is localized to the cytoplasm and GA/vesicular compartments and acts as a GEF for ARF1, and possibly RAB3A and RAB5 [123], also result in a macrocephaly syndrome with dysmorphic facies and psychomotor retardation (MDFPMR; MIM #617011) [124-126].

Similarly, in autosomal dominant mental retardation 48 (MRD48; MIM #617751) mutations in the gene encoding the small Rho GTPase RAC1, which appears to interact with various GA proteins and to play a distinct role in neuronal progenitor proliferation, lead to ID with either microcephaly or macrocephaly, as well as a developmental delay, seizures, absent speech, and a number of defects on MRI [127]. Interestingly, SPATA13 (also known as ASEF2), encoding a GEF for both RAC1 and CDC42 (involved in TKS) that is enriched in the prenatal frontal cortex and plays a role in dendritic spine formation and cell migration, possibly by regulating actin cytoskeletal remodeling [128, 129], is among a set of 26 newly identified ID genes [130].

Another GA-associated gene implicated in corticogenesis is the gene for the pseudokinase STRADA (also known as STE20-related kinase adaptor alpha, or LYK5), deletions in which give rise to macrocephaly in a syndrome encompassing polyhydramnios, megalencephaly, and symptomatic epilepsy (PMSE; MIM #611087) [131, 132]. STRADA appears to form a complex with STK25 and the golgin GM130 to regulate GA morphology and positioning, thus controlling neuronal differentiation and integration during development [133].

Not all GA proteins appear to be associated with microcephalic or macrocephalic syndromes when mutated. However, a number of GA proteins do have other significantly neurodevelopmental effects involving either the central or peripheral nervous system, and lead to phenotypes that include ID, ataxias, peripheral neuropathies, or sensorineural loss. Some of these are described below and summarized in Table 3 (excluding GA-localized enzymes whose role lies primarily in protein or lipid biosynthesis/modification, and are thus likely to have metabolic effects rather than mechanistic ones).

Many GA proteins other than those described above are implicated in various forms of ID, even without being associated with microcephaly or macrocephaly. For example, mutations in the RAB-GDP dissociation inhibitor GDI1 are responsible for a nonsyndromic X-linked form of ID (MRX41/MRX48; MIM #300849) [134-136].

A non-RAB molecular partner of BIG2 (ARFGEF2), which, as we have discussed above, causes a periventricular heterotopia and progressive microcephaly, is another GA protein, FILAMIN alpha (FLNA), that plays a role in neuronal migration by regulating the relocalization of BIG2 and the activation of ARF1 at the cell membrane [137]. Mutations in FLNA lead to X-linked periventricular nodular heterotopia (PVNH1; MIM #300049), characterized by brain structural abnormalities and ID, at least in males, but with no known microcephaly (review [138]).

Mutations in the gene for phosphofurin acidic cluster-sorting protein, PACS1, a trans-Golgi protein involved in the trafficking of FURIN, lead to Schuurs-Hoeijmakers syndrome, an autosomal dominant mental retardation (MRD17; MIM #615009), characterized by ID, dysmorphic features, and brain MRI defects, in addition to skeletal defects [139].

Similarly, autosomal recessive mutations in SCYL1, which encodes a protein located at the interface between the GA and COPI-coated vesicles and is likely involved in trafficking functions, lead to another form of autosomal recessive spinocerebellar ataxia (SCAR21; MIM #616719), characterized by cerebellar atrophy, peripheral neuropathy, and mild ID [140].

Several mutations in the BICD2 gene, which encodes the golgin BICAUDAL2 that interacts with the dynein-dynactin motor complex and RAB6A, have been shown to cause an autosomal dominant form of spinal muscular atrophy (SMALED2; MIM #615290) [141-143]. Similarly, mutations in KIF1C, a kinesin family motor protein that interacts with RAB6 to mediate GA tethering [144], lead to an autosomal recessive form of hereditary spastic paraplegia, spastic ataxia 2 (SPAX2; MIM #611302) [145-147].

Mutations in the TANGO2 gene, which encodes a protein thought to be involved in loading cargo into vesicles at the ER but is also localized at the GA, lead to a complex syndrome characterized by metabolic encephalomyopathic crises, recurrent, with rhabdomyolysis, cardiac arrhythmias, and neurodegeneration (MECRCN; MIM #616878), with multiple neurodevelopmental manifestations including seizures and ataxia [148, 149].

Mutations in GOSR2, a GA SNARE complex member, lead to early-onset ataxia and progressive myoclonic epilepsy (EPM6; MIM #614018) in childhood and skeletal deformities by adolescence [150, 151].

A cis-Golgi protein encoded by RETREG1 or FAM134B has been found to be mutated in hereditary sensory and autonomic neuropathy type IIB (HSAN2B; MIM #613115), characterized by childhood-onset autonomic neuropathy, and heavily impaired sensory and sometimes motor function [152].

Among the GA-associated genes involved in sensorineural loss, OSBPL2, which encodes a fatty-acid receptor, oxysterol-binding protein-like protein 2, is mutated in a nonsyndromic autosomal dominant form of deafness with childhood onset (DFNA67; MIM #616340) [153, 154]. In HPS1 (MIM #203300), a syndrome usually characterized in humans by albinism and excessive bleeding, as in HPS2 (discussed above), mutations in the AP3D1 adaptor -complex subunit give rise to a phenotype of hearing loss, in addition to neurodevelopmental delay and gener-alized seizures (HSP10; MIM #617050) [155]. Several of the genes and syndromes described above also have an impact on vision (several AP and COG subunits, -RAB3GAP2, and TBC1D20, to name a few).

Sometimes, mutations in GA genes that are not thought to be associated with neurological symptoms might, in fact, have unnoticed neurodevelopmental effects, as in the case of the HPSs, as seen above. In other cases, disorders considered to be neurodegenerative might have a neurodevelopmental component in rare, early-onset forms, such as some spinocerebellar ataxias: type 2 (MIM #183090, ATXN2), types 5 and 14 (MIM #600224; MIM #615386, SPTBN2), and type 8 (MIM #610743, SYNE1) [156].

The GA also plays an essential role in the biosynthesis or modification of the cargo being transported. While these are generally considered to be “metabolic” disorders and are related more to the modification of trafficked molecules than to trafficking per se, mutations in many of these also have neurodevelopmental consequences ([72] for a brief overview). For instance, it is the site of synthesis of some glycolipids and sphingomyelin, a major component of myelin and essential for normal morphogenesis or the plasticity of membrane-rich cell types such as neurons (during migration, rapid axonal growth, or dendritic arborization) or oligodendrocytes (for the generation of the myelin sheath), from lipid precursors synthesized in the ER (glycerol phospholipids, cholesterol, and ceramide).

Mutations in genes involved in lipid biosynthesis and processing have also been associated with several neurodevelopmental defects. For example, a mutation in ACER3, encoding an ER and Golgi ceramidase involved in sphingolipid metabolism, was identified in a family with progressive impairment and regression of neurodevelopment during the first year of life, leukodystrophy-like signs, and peripheral neuropathy (PLDECO, MIM #617762) [157]. Another GA gene involved in lipid metabolism and that leads to neurodevelopmental deficits is that encoding DDHD domain containing 2 (DDHD2), a phospholipase with a major role in brain triglyceride metabolism [158], mutations in which result in hereditary spastic paraplegia type 54 (SPG54, MIM #615033) [159]. Mutations in the PGAP2 and PGAP3 genes (post-GPI attachment to proteins 2 and 3), which are essential for the glycolipid remodeling of GPI anchor proteins, have been associated with developmental delays and ID, in addition to microcephaly in several families (HPMRS3, MIM #614207 and HPMRS4, MIM #615716) [160-164], while PGAP1 has also been linked to ID [165, 166].

Among the cargo-modifying enzymes present in the GA are a very large complement of different glycosyltransferases and glycosidases, in addition to enzymes responsible for adding or removing phosphate, sulfate, lipids and other groups from proteins. To give a few examples, mutations in the acetylglucosaminyltransferase gene, MGAT2, a member of the N-glycosylation pathway, lead to CDG type IIa (CDG2A, MIM #212066) characterized by seizures, developmental delays, and ID in addition to other symptoms [167]. Another member of the N-glycosylation pathway is β1,4 galactosyltransferase I, implicated in severe forms of CDG-IId (CDG2D, MIM #607091) with neurological deficits, hydrocephalus, myopathy, and defective blood-clotting [168]. CDG type II is also caused by mutations in the mannosidase gene, MAN1B1, which lead to a syndrome characterized by ID and truncal obesity, while INPP5E (inositol polyphosphate-5-phosphatase E) mutations have been implicated in the neurodevelopmental disorders Joubert (JBTS1, MIM #213300) and MORM syndromes (MORMS, MIM #610156) [169, 170], the latter of which also includes ID and truncal obesity. Several types of muscular dystrophy with neurological involvement are also caused by defects in GA glycosylation genes (e.g., FKRP, LARGE, and POMGnT1). Similarly, a variety of members of SLC families, which transport various metabolites (sugars, acetylCoA, zinc, etc.), are involved in CDGs and other neurodevelopmental disorders including spastic paraplegias, ID, epilepsy, etc. [2].

Another GA protein family is that of the copper-transporting P-type ATPases, ATP7A and B (review [171]), which are specifically trafficked by AP1S1, whose mutation causes MEDNIK (mentioned above). Under basal conditions, ATP7A and B, which are ion pumps, are located in the trans-Golgi network and translocate copper across intracellular membranes into the secretory pathway for incorporation into copper-dependent enzymes. Their mutation, while having multisystemic consequences due to copper deposition in the liver and skin, for example, in addition to the brain, has been linked to Menkes disease (MNK, MIM #309400), occipital horn syndrome (OHS, MIM #304150), or spinal muscular atrophy (SMAX3, MIM #300489) (ATP7A), all disorders with childhood onset, or dystonia and neurodegeneration of either childhood- or late-onset (ATP7B).

What is the explanation for the exceptionally large spectrum of diseases and symptoms associated with the defective GA proteins listed above, even when one limits the discussion to those with neurodevelopmental consequences?

The GA is not only involved in but lies at the crossroads of a number of distinct and crucial cellular processes, some of which have only recently begun to be elucidated. In addition, individual GA proteins themselves often participate in more than one function or pathway, depending on the cellular context, developmental stage, and pathophysiological state, and some, such as the small RAB GTPases, belong to extended families with a whole gamut of functions within and outside the GA. However, much is still not known about the precise functions and molecular partners of the great majority of GA proteins, and existing information often comes from lower species or organs outside the nervous system. Despite these overlaps and restrictions, obtaining even an incomplete understanding of some of these functions and proteins may provide insights into the complex repercussions of their mutation or deregulation, including the various neurodevelopmental processes impacted. Figure 1 illustrates the relationship between some of the principal GA functions and the different stages of neurodevelopment in which they are implicated, taking the central nervous system as an example. Understanding these functions and how they evolve over the life span of the cell and the organism would also help to elucidate why mutations in some families of GA proteins appear to cause a spectrum of deficits rather than discrete phenotypic entities (e.g., both congenital microcephalies and POM in the case of defective TRAPP, AP, or SLC complex subunits).

A distinctive characteristic of the GA in the cells of mammals and other vertebrates is its structure. It is made up of stacks of flattened cisternae, as in protists, plants, and invertebrates, but these stacks are linked to each other to form a “ribbon” [172, 173]. Cisternal stacking is necessary for normal trafficking, while the ribbon structure is thought to play a role in higher-order functions such as mitosis and apoptosis, cell polarity/migration, and stress responses (review [174, 175]). A number of GA proteins are involved in regulating this typical structure and the integrity of the GA. Prominent among these are various members of the coiled-coil golgin family of proteins (e.g., GM130, GIANTIN, P115, GOLGIN-45, GMAP210, GIRDIN, etc.), many of which interact with coiled-coil GRASP proteins and small -GTPases or GTP-binding proteins of the RAB, ARF, and ARL families to control cisternal structure and stacking in the GA. (review [174-176]). Many of the GA-associated GTPases are, in turn, regulated by families of regulatory proteins, such as the GEFs and GAPs, which act at various points and target effector molecules in the trafficking pathway. Among the GA-associated members of these large families of proteins, some are specifically expressed in neural tissue, e.g., RAB6B [177] and to some extent RAB6C [178]. Several golgins (e.g., Bicaudal D and GMAP210) and RAB GTPases (e.g., the multifunctional RAB6) are involved in interactions between the GA and the microtubule cytoskeleton, including molecular motors such as DYNEIN, and help in the organization of the GA ribbon by recruiting GA stacks to the microtubule-organizing center (MTOC). Some, such as GM130, through their interactions with other polarity-determining kinases as well as their role in microtubule nucleation and organization [179-182], dictate the polarity of the GA and thus of the cell or neuron as a whole [183], thereby controlling its correct differentiation and function. The fragmentation of the GA ribbon through the phosphorylation of the GRASPs and other molecules might also allow GA repositioning and thus migration, growth, or differentiation [184] (fragmentation is also essential for the progression of the cell cycle as discussed below). In addition to structural changes, GM130 and GRASP65 also form part of signaling cascades that regulate cell migration [181].

GA structure and integrity and many of the large array of molecules that control them are also essential for the various steps involved in vesicular transport: coat components and adaptors that control vesicle targeting and recognition, the coiled-coil golgins necessary for their capture and the SNAREs and COGs that help in tethering and fusion, RAB, and other small GTPases that mediate the choreographed attachment and detachment of molecules at each step, adaptors, activators, and inhibitors for each of the above, etc. Interfering with these proteins or their interactions results in disrupted GA structure/organization and abnormal protein and lipid transport (e.g., the effects of disrupting various RABs on GA structure, reviewed in [185]) important for both neuronal and oligodendrocyte function, including the rapid membrane expansion needed for the establishment of neuronal outgrowth and connectivity and myelination. Mutations in these genes are involved in a number of Golgipathies with diverse symptoms, including the large majority of the POM syndromes with the white-matter defects listed above (DMC, WARBM, Martsolf syndrome, COH, PCCA2 and other PCHs, microcephalies related to mutations in the TRAPP complex, and COG-related CDGs, to name a few) as well as a range of spinal muscular atrophies and paraplegias.

Other functional consequences could be the disruption of neuronal migration to the appropriate target layers or regions, due to both membrane trafficking deficits and disruption of the MTOC function of the GA and thus of cellular polarity [4]. Certain syndromes characterized by epilepsy or seizures could therefore be a result of either the aberrant connectivity caused by such migration defects, or the defective transport of neurotransmitters or their receptors/inhibitors to the synapse. Additionally, the occurrence of central hypogonadism, dwarfism, obesity, etc. of unknown etiology in certain POM Golgipathies could conceivably be due to deficits in neuroendocrine secretory activity controlling bodily hormonal homeostasis.

Closely related to the structure and position of the GA is its role in the cell cycle. The typical ribbon structure of the GA in mammalian cells is usually localized close to the centrosome, a position regulated, among others, by a number of GA-associated proteins including those involved in its structure and trafficking function (review [186]). During mitosis, the GA loses its pericentriolar position and its ability to function as an independent MTOC and undergoes fragmentation to yield small tubular-reticular structures and dispersed vesicles. While GA fragmentation could be considered a logical prerequisite for the partitioning of the GA into daughter cells, it appears it is in fact a prerequisite for mitosis itself and that blocking it, e.g., by blocking GRASP phosphorylation, acts as a “Golgi checkpoint” for the progression of mitosis [187-189]. Failure of the GA to fragment results in the failure of Aurora-A recruitment at the centrosome, essential for mitosis [190]. Several other GA proteins (e.g., GM130 and golgin-84) and non-GA proteins are also activated during fragmentation. A number of RAB GTPases including several GA-associated RABs as well as the small G-protein ARF1, along with its activators and effectors, play essential roles at various steps of mitosis. Surprisingly, several GA proteins acquire new locations and functions during mitosis (review [188]). Some, like GM130, whose role in this regard has been described above, may also play very specific roles in neurogenesis and neural progenitor division, by influencing cell polarity (review [191]) or the symmetry of mitosis, which play a crucial role in the generation of different types of neuronal progenitors while still maintaining the pool of stem cells. For example, the asymmetric division of neural progenitor cells requires the interaction of the golgin GCP60 (also known as ACBD3 or PAP7) and Numb [192].

While the pathophysiological mechanisms underlying the syndromes characterized by macrocephaly mentioned above are not clear, most of the mutations detected result in a loss of function, and it is conceivable that uncontrolled proliferation or the dysregulation of the timing of symmetric versus asymmetric divisions during neurogenesis could lead to the overproduction of certain neurons. In another instance of the overlap between the trafficking function of the GA and its role in neurogenesis, the trans-Golgi BIG2 protein encoded by ARFGEF2, essential for protein trafficking, is also involved in normal neural progenitor cell proliferation, migration, and axonal growth in the developing mammalian cortex through its interaction with RAB11, another trans-Golgi protein. The early microcephaly and aberrant corticogenesis noted in ARPHN, caused by ARFGEF2 mutations [59-62], could, for example, be due to defective neurogenesis and migration rather than to defects in neuronal maturation or myelination.

It should not be forgotten that mutations in GA proteins whose function is essential early in neurodevelopment, e.g., during neurogenesis, might simply be embryonic-lethal and thus pass unremarked.

At the other end of the spectrum from cell proliferation and membrane expansion, the GA is also involved in normal or pathological cell death, or the destruction of cellular contents. Both apoptosis and autophagy are normally occurring processes by which unwanted cells or cellular components are eliminated while limiting tissue damage (i.e., leakage of cellular contents). They occur during neurodevelopment, when the large excess of new neurons or membranes produced is culled during the maturation of the central nervous system. In addition, they are brought into play for the removal of damaged cells during conditions of stress or disease. In fact, the GA is increasingly thought to play a role in sensing stress and reestablishing cellular homeostasis; prolonged “Golgi stress,” like ER stress, while not due to an intrinsic GA defect, nevertheless leads to impaired or abnormal GA function, contributing to pathogenesis and cell death also by novel mechanisms other than apoptosis [193].

In autophagy, or more specifically, macroautophagy, the cellular contents to be degraded are enveloped in a “phagophore” or “isolation membrane” and delivered to acidic -lysosomes for destruction. There is evidence that the -phagophore might bud directly from GA membranes or, alternatively, from the ER or ER-to-Golgi intermediate compartment [194, 195]. The GA might additionally be directly involved in the degradation of undesirable cellular contents [196].

Several GA-associated RAB GTPases and their partners, normally involved in various stages of trafficking, also play key roles in the formation of the autophagosome [197-201], as do SNAREs, which are involved in autophago-some fusion (review [202]). Among other GA proteins, -BECLIN1, a trans-Golgi protein involved in endosome-to-GA recycling, plays a crucial early role in autophagosome formation (review [203]) by interacting with UVRAG, which normally mediates GA-to-ER retrograde transport by tethering COPI-coated vesicles but is displaced during autophagy [204]. Similarly, the trans-Golgi membrane-bound protein ATG9 is found in vesicles that contribute to autophagosome formation [199], and the regulation of its trafficking plays a crucial role in the induction of autophagy pathways [205, 206]. Several coat adaptor proteins of the AP1, 2, and 4 classes are known to interact with ATG9, and are also necessary for autophagosome formation [207, 208].

Apoptosis, resulting from mitochondrial membrane permeabilization in response to internal or external signals, and leading to cell death by strictly regulated molecular and cellular mechanisms, is mediated by members of the caspase and BCL-2 families of proteins. Much of the evidence for the involvement of GA proteins in apoptosis comes from neurodegenerative disorders such as Alzheimer, Parkinson, and Huntington diseases, where the GA undergoes stress due to an accumulation of aberrant and misfolded proteins or peptides, also causing it to fragment (review [209-211]). However, some of the early machinery for apoptosis localizes to the GA and is involved in cleaving various GA proteins that also play a role in neurodevelopment (e.g., GRASP65, GM130, and GIANTIN). Their cleavage triggers the mitosis-like disassembly of the GA and subsequent apoptosis (review [212]). Some GA proteins also have a larger signal transduction role; GMAP210 and golgin-160 as well as cleaved p115 fragments are translocated to the nucleus during apoptosis, triggering further apoptotic changes [213, 214]. It is certainly possible that mutations that interfere with the cleavage of such GA proteins or that lead to the production of truncated proteins would either curtail or augment apoptosis, with consequences for neurodevelopment ranging from insufficient neurogenesis to excess neurons that are not culled and that lead to macrocephaly, aberrant connectivity, and related disorders.

As mentioned previously, the GA is also intimately involved in the posttranslational modification of proteins and the biosynthesis of some lipids. In terms of the physiological effects of mutations in these proteins versus mutations in proteins involved in trafficking per se, there might be a convergence, since the functional consequences of the lack of delivery of a modified protein to its target intracellular compartment and the lack of its modification in the first place are likely to be similar, at least in some respects. In addition, the highly specialized nature of the nervous system implies a corresponding adaptation of GA trafficking function in the component cells (e.g., the occurrence of Golgi outposts in dendrites) and, correspondingly, defects in GA enzyme function often lead to neurological consequences (review [2, 4]). A prime example of this convergence is provided by the CDGs, which could be related to structural/mechanical deficits (e.g., due to the mutation of COG subunits necessary for vesicle tethering) or enzymatic deficits (due to the failure of glycosylation).

While the large number of neurodevelopmental Golgipathies described in this review contain some surprises, it appears as if the majority have simply been hiding in plain sight. The GA is involved in a very large range of functions spanning every stage of a cell’s life span. An even larger number of overlapping GA-associated proteins and pathways carry out or regulate these functions. It is therefore to be expected that, as our knowledge of these proteins and pathways grows and modern sequencing methods provide a molecular diagnosis for more obscure syndromes, the range of phenotypes that we know to be caused by defective GA proteins and their partners will expand. This is especially true when one considers the essential roles played by the GA in the formation, growth, and maturation of the brain and nervous system. In fact, the spectrum of syndromes in which mutations in GA-associated proteins are implicated, and the overlap in the cellular roles played by these proteins, evoke the intriguing idea that the GA might modify its role in keeping with the needs of the cell or the organism during its life span. For instance, the same proteins that are actively involved in the initiation or regulation of the cell cycle during neurogenesis, when abundant proliferation is required, might be repurposed to lend a hand in neuronal migration, apoptosis, or maturation, meet secretory trafficking needs during synaptogenesis and myelination as well as normal cellular function for the major part of the life span of the cell, and, finally, play a role in repair and damage control during aging [215]. From this perspective, this vast array of neurodevelopmental Golgipathies appears not as isolated syndromes that differ from each other, but as points that fall along a few key functional continua, with flexible phenotypes that primarily reflect the developmental stage at which function is interrupted or altered. Such an organelle-centric approach might eventually make it easier to understand the pathophysiology of these disorders.

We are grateful to Cécile Martel for invaluable support. This work was supported by the Institut National pour la Santé et la Recherche Médicale (INSERM), the Centre National de la Recherche Scientifique (CNRS), the Université Paris 7, DHU PROTECT, and European and national grants from the French National -Research Agency (ANR-13-RARE-0007–01 to AV, SP and VEG (ERA-NET E-Rare 2013, EuroMicro project), ANR-15-NEUR--0003–01 to PG (ERA-NET Neuron, MicroKin project), and ANR-16-CE16–0024–01 to VEG (PRC GENERIQUE 2016, MicroGol project).

The authors declare there are no conflicts of interest.