The DUB Club: Deubiquitinating Enzymes and Neurodevelopmental Disorders

Abstract

Protein ubiquitination is a widespread, multifunctional, posttranslational protein modification, best known for its ability to direct protein degradation via the ubiquitin proteasome system (UPS). Ubiquitination is also reversible, and the human genome encodes over 90 deubiquitinating enzymes (DUBs), many of which appear to target specific subsets of ubiquitinated proteins. This review focuses on the roles of DUBs in neurodevelopmental disorders (NDDs). We present the current genetic evidence connecting 12 DUBs to a range of NDDs and the functional studies implicating at least 19 additional DUBs as candidate NDD genes. We highlight how the study of DUBs in NDDs offers critical insights into the role of protein degradation during brain development. Because one of the major known functions of a DUB is to antagonize the UPS, loss of function of DUB genes has been shown to culminate in loss of abundance of its protein substrates. The identification and study of NDD DUB substrates in the developing brain is revealing that they regulate networks of proteins that themselves are encoded by NDD genes. We describe the new technologies that are enabling the full resolution of DUB protein networks in the developing brain, with the view that this knowledge can direct the development of new therapeutic paradigms. The fact that the abundance of many NDD proteins is regulated by the UPS presents an exciting opportunity to combat NDDs caused by haploinsufficiency, because the loss of abundance of NDD proteins can be potentially rectified by antagonizing their UPS-based degradation.

The modification of protein function through covalent attachment of ubiquitin (Ub) via lysine (K) residues is a widespread event involved in every eukaryotic cellular process (1). As a posttranslational mechanism, ubiquitination is used where rapid changes in protein and cell function are essential, for example, enabling signaling cascades, cell cycle changes, cell differentiation, and plasticity (2,3). The specificity of protein ubiquitination is bestowed via approximately 600 enzymes called E3 ubiquitin ligases, that dictate which, when, where, and how proteins are ubiquitinated (1). Ubiquitination is reversible, and deubiquitination is also critically important. Only approximately 90 deubiquitinating enzymes (DUBs) are currently known, but recent identification of noncanonical DUB families suggests that others may exist. This comparatively small number of DUBs are charged with the challenge of precisely regulating all ubiquitinated proteins (4). DUBs target many substrates and function as regulators of cell, developmental, and disease processes (5,6).

Ubiquitination is a diverse and multifunctional posttranslational modification (7). Proteins can be monoubiquitinated or polyubiquitinated, the latter involving sequential attachment of additional Ub moieties to either the N-terminal methionine (M1) or one of seven internal lysine residues of Ub itself (K6, K11, K27, K29, K33, K48, K63). The different Ub chain lengths and linkage types provide a “ubiquitin code” that is deciphered by a host of Ub-binding receptors, which dictate downstream outcomes (7,8). Ubiquitination can facilitate protein-protein interactions, protein trafficking, subcellular localization, and enzymatic activity, among other outcomes, in addition to its well-established role in directing protein degradation. Classically, the attachment of a poly-Ub chain of four or more K48- or K11-linked Ub molecules targets proteins to the 26S proteasome for degradation, a process known as the Ub proteasome system (UPS) (1). In addition, ubiquitination of proteins trafficked through vesicular systems instructs their delivery to the lysosome for degradation (9). In turn, many DUBs antagonize these degradation pathways by removing Ub from specific subsets of proteins in a precise and regulated manner (10).

DUBs are either cysteine- or metalloproteases and divided into seven subfamilies (11). The sole metalloprotease subfamily is the Jad1/Pad/Mpn domain–containing metalloenzymes (JAMMs). The cysteine protease subfamilies include the ubiquitin specific proteases (USPs), ovarian tumor proteases (OTUs), Machado-Joseph disease domain proteases (MJDs), Ub C-terminal hydrolases (UCHs), Josephins motif interacting with Ub-containing novel DUB family (MINDYs), and the zinc-finger and UFSP domain protein (ZUFSP) (11). The sole function of some DUBs is to maintain the free Ub pool in cells, such as UCHL1, which releases free Ub from newly synthesized Ub polymers, and USP14 and UCHL3, which recycle Ub from proteins at the proteasome (11). However, many DUBs are known to target specific sets of ubiquitinated proteins derived via affinities for particular Ub chain lengths or linkage types (e.g., OTU family members) or auxiliary protein-protein interaction motifs (e.g., USP family members) (11–17).

The process and functions of protein ubiquitination and deubiquitination and their roles in neurodegenerative disorders have been reviewed in depth (1,4,7,18,19). Here, we focus on the role of DUBs in the developing brain. We describe genetic evidence from human and mouse implicating DUBs in a spectrum of neurodevelopmental disorders (NDDs) including developmental delay (DD); intellectual disability (ID); epilepsy; motor, speech, language, and visual difficulties; and behavioral disturbances. We highlight how NDD DUB substrates provide insight into the pathological mechanisms converging on loss of protein abundance and synthesize this knowledge to discuss therapeutic paradigms relevant to individuals with NDD DUB mutations and, more broadly, to NDDs caused by haploinsufficiency.

DUBs IN NDDs

Mutations in 12 DUBs are currently known to cause NDDs (Table 1). Half of these belong to the USP subfamily. Heterozygous loss of function (LoF) mutations (encompassing gene deletion, nonsense, and frameshift mutations) of USP7 have been reported in 23 individuals with DD/ID, speech delay, autism spectrum disorder (ASD), impulsivity, compulsivity, and aggression (20,21). Brain imaging revealed decreased white matter and thinning of the corpus callosum. A thin cortex and reduced brain size is observed in brain-specific Usp7 knockout mice and attributed to elevated apoptosis (22). Mutations in the X chromosome gene Usp9x cause an X-linked NDD in hemizygous males (23–25) and also an NDD syndrome in heterozygous females (26,27). This phenomenon is resolved by the fact that males are affected by hypomorphic (i.e., partially nonfunctional) missense mutations, which are tolerated in heterozygous female carriers (23,24), while females are affected by de novo heterozygous complete LoF mutations (26). Because USP9X escapes X-inactivation, the heterozygous LoF alleles result in an unusual state of female X-linked haploinsufficiency (26). The core neurologic phenotypes caused by USP9X mutations in females (33 individuals) and males (16 individuals) are shared, including DD/ID, delayed speech, movement disorders, and behavioral problems including ASD (27). Frequent brain malformations include agenesis of the corpus callosum and ventriculomegaly (27). Usp9x brain-specific knockout mice show NDD hallmarks, including defective learning, abnormal communication and socialization, dysgenesis of the corpus callosum, and ventriculomegaly (28–31). These mice display defects in neural stem cell (NSC) polarity, proliferation, and differentiation and neuronal cell migration, growth, and connectivity (23,28–30,32–34). De novo heterozygous LoF mutations of USP15 have been identified in 11 individuals with ASD (35,36). Intriguingly, Usp15 knockout mice display altered neuroinflammatory responses that potentially contribute to the NDD phenotype (37). Biallelic LoF mutations in USP18 have been identified in 6 individuals with pseudo-TORCH syndrome (38,39). TORCH syndrome results from placental-to-fetal microbe transmission, and pseudo-TORCH syndrome describes neonates with TORCH features in the absence of infection. Brain characteristics include hemorrhage, microcephaly, white matter loss, cerebral atrophy, and calcifications. Together with other congenital features, neonatal lethality results. A hallmark of pseudo-TORCH syndrome is hyperactivation of the brain’s innate immune system. USP18 is a unique DUB that cleaves the Ub-like molecule ISG15 from substrates to limit interferon-mediated inflammation (38,40). A hyperactive innate immune response occurs in Usp18 knockout mice, which feature hydrocephalus and reduced life span (41). Two hemizygous LoF mutations affecting 4 individuals with ID, speech, and behavioral problems have been identified in the X chromosome gene USP27X (42). Two additional individuals with NDD have been identified with hemizygous deletions involving USP27X (43). Knockout mice have not been generated; however, overexpression of Usp27x inhibits neurogenesis in mice (44). Biallelic LoF mutations in USP45 cause Leber congenital amaurosis, a severe form of inherited retinal dystrophies that results in irreversible childhood blindness (45). Usp45 knockout mice feature severely disrupted eye function caused by loss of photoreceptor cone cells (45).

Table 1.

DUBs Involved in NDDs

DUB Class	Gene	Inheritance	NDD Phenotype (OMIM)	Other Disease Associations	Key References
USP	USP7	AD	Hao-Fountain syndrome (HAFOUS) (616863)	Cancer, neurodegeneration	(20–22)
	USP9X	XL	Males: Mental retardation X-linked family 99 (MRX99) (300919) Females: USP9X-female syndrome/Mental retardation X-linked family 99-female syndrome (MRX99F) (300968)	Cancer, neurodegeneration	(23–34)
	USP15	AD	Autism spectrum disorder (NA)	Cancer	(35–37)
	USP18	AR	Pseudo-TORCH syndrome 2 (617397)	Cancer, autoimmune	(38–41)
	USP27X	XL	Mental retardation X-linked family 105 (MRX105) (300984)	Cancer	(42–44)
	USP45	AR	Leber congenital amaurosis 19 (618513)	Cancer, inflammatory bowel disease	(45)
OTU	OTUD5	XL	Multiple congenital anomalies with neurodevelopmental syndrome – X-linked (MCAND) (301056)	Cancer,	(46–48)
	OTUD6B	AR	Intellectual developmental disorder with dysmorphic facies, seizures, and distal limb anomalies (IDDFSDA) (617452)	Cancer	(49–51)
	OTUD7A	AD	Microdeletion 15q13.3 (612001)	Cancer	(52–56)
JAMM	STAMBP	AR	Microcephaly-capillary malformation syndrome (MICCAP) (614261)	Cancer, neurodegeneration	(57,58)
MJD	ATXN3	AD	Machado-Joseph disease type I (109150)	Cancer, neurodegeneration	(66,67)
UCH	UCHL1	AR	Spastic paraplegia 79–autosomal recessive (SP79-AR) (615491)	Neurodegeneration	(59–65,70)

List of known NDD DUBs. Inheritance patterns shown are AD, AR, and XL. Phenotypes are referenced with OMIM entries.

AD, autosomal dominant; AR, autosomal recessive; DUB, deubiquitinating enzyme; JAMM, Jad1/Pad/Mpn domain-containing metalloenzymes; MJD, Machado-Joseph deubiquitinase; NA, not applicable; NDD, neurodevelopmental disorder; OTU, ovarian tumor proteases; UCH, ubiquitin C-terminal hydrolase; USP, ubiquitin specific protease; XL, X chromosome–linked.

Mutations in three OTU subfamily members cause NDDs (Table 1). Hemizygous missense mutations in the X chromosome gene OTUD5 cause a male-limited NDD named multiple congenital anomalies-neurodevelopmental syndrome. Ten unique missense variants affecting 26 individuals have been described (46–48). Mutations are hypomorphic, and outcomes range from DD to neonatal lethality. Common features include brain malformations (ventriculomegaly, hydrocephalus), congenital heart disease, and craniofacial and genitourinary defects. Both knockout and disease-relevant knockin mouse models are embryonic lethal, while induced pluripotent stem cell (iPSC) models show reduced differentiation into NSCs (47). Biallelic LoF mutations in OTUD6B have been discovered in 15 individuals featuring ID/DD, seizures, absent speech, hypotonia, and microcephaly (49–51). Variable brain structural abnormalities include white matter volume loss, corpus callosum dysgenesis, and dilatation of lateral ventricles. Knockout mice are nonviable, but embryos displayed reduced size with brain and organ abnormalities (49). OTUD7A was identified as the critical gene responsible for brain phenotypes associated with heterozygous 15q13.3 deletions found across hundreds of individuals with NDDs including ID, ASD, epilepsy, and schizophrenia (52). In support, deletions involving only OTUD7A, as well as OTUD7A LoF mutations, have been identified in individuals with NDDs (53,54). Furthermore, both the syntenic 15q13.3 microdeletion mouse model and the Otud7a knockout mouse model recapitulate the human 15q13.3 syndrome, and re-expressing OTUD7A in cortical neurons from the microdeletion model completely rescues neuronal dendritic spine and branching defects (55,56).

Mutations in members of the JAMM-containing metalloenzyme, UCH, and MJD domain protease subfamilies also cause NDDs (Table 1). Homozygous mutations in STAMBP have been identified in 19 individuals with microcephaly-capillary malformation characterized by microcephaly with progressive cortical atrophy, intractable epilepsy, DD, and small capillary malformations on the skin (57). Knockout mice exhibit neuronal damage and elevated apoptosis in the hippocampus (58). Homozygous missense and splice-site mutations in the UCH subfamily member UCHL1 have been reported in 10 cases with spastic paraplegia-79, an early-onset neurodegenerative syndrome (59–62). This childhood syndrome features blindness, cerebellar ataxia, nystagmus, dorsal column dysfunction, and spasticity. The progressive ataxia, muscular, and lethal phenotypes are observed in mutant and knockout mouse models (63,64), which display nerve fiber loss, axonal swelling, degeneration in the spinal cord, and impaired neuromuscular denervation and synaptic transmission (65). Finally, a polyglutamine-coding CAG trinucleotide repeat expansion in the MJD domain protease subfamily member ATXN3 causes an autosomal dominant neurodegenerative disorder called spinocerebellar ataxia 3 (SCA3), which can have infantile or childhood onset called SCA type 1 (66). SCA type 1 is a progressive neuropathy causing severe movement disturbances (ataxia, weakness, dysarthria, spasticity), visual defects, and cognitive decline, culminating in death in early adulthood (66). A polyglutamine-expanded Atxn3 knockin mouse displays adult-onset motor decline; however, a larger expansion of the polyglutamine tract reflective of SCA type 1 is likely required to better model earlier onset (66). Nonetheless, this model reveals accumulation and aggregation of ATXN3 in adult brain cells, neuropathology in both neurons and glia, cerebellar degeneration with loss of Purkinje cells, and evidence of neuroinflammation (67). Collectively, these data highlight that compromised function of DUBs caused by LoF or hypomorphic mutations can cause NDDs, and current mouse models of these disorders are providing valuable insight into the molecular, cell, and physiological mechanisms of pathology.

CANDIDATE DUBs IN NDDs

Recent genetic and functional data have identified at least 19 additional DUBs as NDD gene candidates (Table 2). The chromosome 21 gene USP16 is implicated in Down syndrome. In the Ts65Dn mouse model, which is trisomic for 132 human chromosome 21 gene orthologs, trisomy of USP16 was shown to drive NSC defects (68). Duplications of <1 Mb of sequence including USP16 have also been identified in 2 individuals with NDDs (43). A homozygous missense mutation in USP11 has been identified in an individual with ID and brain malformation (69). Conditional brain knockout of Usp11 in mice causes behavioral, learning, and memory defects underpinned by defective generation and migration of cortical neurons (69). Unlike wild-type Usp11, re-expression of the Usp11 NDD variant in the knockout model failed to rescue these defects, establishing LoF of USP11 as the pathogenic mechanism. Three other DUBs including UCHL1 (described above), OTUD7B, and OTUB1, have been shown to interact with polymorphisms (via chromatin conformation capture and expression quantitative trait loci analysis) identified in genome-wide association studies incorporating multiple psychiatric and behavioral disorders (70). In support, OTUD7B depletion in NSCs induces differentiation (71). USP46 and USP8 are involved in synaptic transmission. Both localize to the postsynaptic density and facilitate recycling of AMPA receptors back to the synaptic membrane (72,73). Usp46 hypomorphic and knockout mice show depression-like phenotypes (74,75), while Usp8 knockout mice are embryonic lethal (76). USP14 is also involved in synapse development. Hypomorphic mice have an early-onset progressive ataxic phenotype associated with defects in synaptic transmission (77), while knockout results in additional NDD phenotypes (78). In utero knockdown of USP1, USP4, and USP20 in the mouse cerebellum disrupts granule neuron morphogenesis, while knockdown of USP30 and USP33 disrupts granule neuron migration (79). CYLD also regulates axonal outgrowth, structure of the postsynaptic density, and synaptic transmission (80,81). Additional UCH family members are also implicated in brain development. Knockout of Uchl3 in mice leads to learning, memory, and synaptic transmission defects, while knockout of Uchl5 causes lethality with severe defects in embryonic brain development (82,83). Finally, mutation of the DUB Otulin also has a severe NDD phenotype in mice (84). LoF of any of these candidate NDD DUBs may plausibly give rise to a human NDD.

Table 2.

Additional DUBs That Regulate Neurodevelopment

DUB Class	Gene	Inheritance	NDD Phenotype (OMIM)	Other Disease Associations	Key References
USP	USP1	n/a	n/a	Cancer, Fanconi anemia	(79)
	USP4	n/a	n/a	Cancer	(79)
	USP8	n/a	n/a	Cancer, Cushing disease	(73,76)
	USP11	XL	Intellectual disability	Cancer	(69)
	USP14	n/a	n/a	Cancer, neurodegeneration	(77,78)
	USP16	AD/IC	Down syndrome (190685)	Cancer	(43,68)
	USP20	n/a	n/a	Cancer	(79)
	USP30	n/a	n/a	Neurodegeneration	(79)
	USP33	n/a	n/a	Cancer	(79)
	USP46	n/a	n/a	n/a	(72,74,75)
	CYLD	n/a	Frontotemporal dementia and/or amyotrophic lateral sclerosis 8 (FTDALS8) (619132)	Brooke-Spiegler syndrome, cylindromatosis, trichoepithelioma, cancer, neurodegeneration	(80,81)
OTU	ALG13	XL	Developmental and epileptic encephalopathy 36 (DEE36) (300884)	Congenital disorder of glycosylation	(85)
	OTUD7B	Complex	Neuropsychiatric disorder	Cancer	(70,71)
	OTUB1	Complex	Neuropsychiatric disorder	Cancer	(70)
	OTULIN	n/a	n/a	Autoinflammatory disease	(84)
JAMM	PRPF8	AD	Retinitis pigmentosa 13 (600059)	Glaucoma, cancer	(90,91)
	EIF3F	AR	Mental retardation–autosomal recessive 67 (MR-AR67) (618295)	Cancer	(86–89)
UCH	UCHL3	n/a	n/a	Cancer	(82)
	UCHL5	n/a	n/a	Cancer	(83)

List of candidate NDD DUBs. Inheritance patterns shown are AD, AR, XL, and IC. Phenotypes are referenced with OMIM entries.

AD, autosomal dominant; AR, autosomal recessive; DUB, deubiquitinating enzyme; IC, isolated case; JAMM, Jad1/Pad/Mpn domain-containing metalloenzymes; n/a, not applicable; NDD, neurodevelopmental disorder; OTU, ovarian tumor proteases; UCH, ubiquitin C-terminal hydrolase; USP, ubiquitin specific protease; XL, X chromosome–linked.

In addition to the above candidate NDD DUBs, mutations in ALG13, EIF3F, and PRPF8 are known to cause NDDs, but their DUB activity remains speculative. Over 50 individuals, predominantly female, have been identified with de novo missense variants in the X chromosome gene ALG13, causing an NDD featuring an infantile epileptic encephalopathy with DD/ID and hypotonia (85). Although ALG13 is best known as a uridine diphosphate-N-acetylglucosaminyltransferase, the individuals display no biomarkers of congenital disorders of glycosylation, casting intrigue over the relevance of its uncharacterized OTU deubiquitinase domain (85). A recurrent autosomal recessive mutation in EIF3F also causes an NDD. The homozygous p.Phe232Val mutation has been identified in 29 individuals (86,87) with DD/ID, seizures, behavioral disturbances, hearing loss, hypertonia, eye abnormalities, and sleeping problems. EIF3F is a translational initiation factor (88), but its deubiquitinating activity has not been well established (89). Mutations in PRPF8 are a well-known cause of autosomal dominant retinitis pigmentosa, a degenerative eye disorder that can present in childhood as early-onset severe retinal dystrophy. PRPF8 encodes a spliceosome factor, and although it harbors a JAMM domain–containing metalloenzyme domain, its deubiquitinating activity has yet to be displayed (90,91).

The repositories of transcriptomes, genomes, and exomes can be leveraged to prioritize candidate DUBs in NDDs, with the prediction that an NDD gene should be highly expressed in the brain and intolerant to genetic variation. Cortical expression of DUBs derived from the Genotype-Tissue Expression database (92) reveals that most of the ~90 DUBs are expressed in the cortex, but some to a much higher degree than others (Figure 1A; Table S1). The gnomAD V2 resource of genome variation of individuals devoid of NDDs enables predictions of which DUBs are intolerant to genetic variation (Figure 1B, C; Table S1) (93). Combining this information provides a rationale to prioritize investigations on candidate NDD DUBs including USP11, CYLD, and PRPF8, and identifies potential new NDD DUBs such as ZRANB1, PAN2, COPS5, VCPIP1, USP5, USP24, and USP48 (Figure 1D). In aggregate, these genetic and functional data further highlight the essential requirements of DUB function during brain development and suggest that more NDD DUBs are on the verge of discovery.

Figure 1.

Brain expression and mutational constraint of DUBs. Metrics for NDD DUBs and candidate NDD DUBs are highlighted. (A) Cortical mRNA expression of DUBs. mRNA expression was extracted from the GTEx database and displayed as log₂ TPM. Low expression is defined as < −3.32 (1 transcript per 10,000,000 reads). (B, C) Mutational constraint of DUBs. Metrics for intolerance to loss of function variants (B) and missense variants (C) were extracted from gnomAD. (B) Stringent intolerance to loss of function is defined as a LOEUF score < 0.35 (93). (C) Stringent intolerance to missense variation is defined as a Z-score > 3 (93). (D) Combined analysis of brain expression and mutational constraint. Note that all known NDD DUBs and candidate NDD DUBs are expressed in the cortex. Only 4 of 12 NDD DUBs meet the criteria of Z-score > 3 and LOEUF score < 0.35 (USP9X, USP7, USP15, and OTUD5), which cause dominant or X-linked NDDs as predicted for intolerant alleles. These data therefore support several candidate NDD and other DUBs as potentially dominant drivers of NDDs, being defined as highly intolerant to variation and highly expressed in the cortex. DUBs, deubiquitinating enzymes; gnomAD, the Genome Aggregation Database; GTEx, Genotype-Tissue Expression; LOEUF, loss of function observed/expected upper bound fraction; mRNA, messenger RNA; NDD, neurodevelopmental disorder; TPM, transcripts per kilobase million.

MOLECULAR AND CELLULAR FUNCTIONS OF NDD DUBs

Most NDD DUBs are best known for their ability to antagonize protein degradation via the UPS (Figure 2A). The major exceptions are UCHL1, STAMBP, and USP18. UCHL1 functions to maintain free Ub by releasing it from newly translated multimeric Ub peptides, ribosomal-Ub fusion peptides, and other small Ub adducts (94). Loss of UCHL1 function thus results in a global reduction in the efficiency of ubiquitination machinery (94). The major function of STAMBP is to antagonize vesicular trafficking of membrane-associated proteins to the lysosome. It removes mono- and K67-linked Ub (95), which are common membrane receptor protein internalization signals that instruct delivery to the lysosome for degradation. STAMBP-mediated deubiquitination of cargo redirects their trafficking back to the plasma membrane for ongoing signaling (95), and as such, STAMBP mutations likely cause pathology via defective signaling of receptors in the developing brain. USP18 cleaves the Ub-like molecule ISG15 from protein targets, a modification not directly linked to the UPS (40). Instead, protein ISGylation promotes protein-protein interactions or enzyme activity and functions specifically in the innate immune response (40). USP18 therefore likely affects central nervous system development in a non–brain cell autonomous manner. The remaining NDD DUBs including USP7, USP9X, USP15, USP27X, USP45, OTUD5, OTUD6B, OTUD7A, and ATXN3, are best characterized for their ability to antagonize the UPS through deubiquitination of their substrates. This function is not exclusive, as demonstrated by the USP family members, which generally show a broad spectrum of Ub chain length and linkage preferences (11). For example, although USP9X protects the majority of its substrates from the UPS by removing K48- and K11-linked Ub, it also displays activity against other Ub linkages and chain lengths and regulates processes including vesicular trafficking, subcellular localization, enzyme activity, protein interactions, and chromatin architecture (32,96). Regarding cellular functions of the NDD DUBs, most of the limited knowledge available derives from the study of cancer. The emerging landscape of DUB functions is broad and aligns with use of ubiquitination in dynamic processes including signaling cascades, gene expression and cell proliferation, differentiation, and response mechanisms (Figure 2B).

Figure 2.

Molecular and cellular roles of NDD DUBs. (A) Major known roles of NDD DUBs and candidate NDD DUBs in the Ub system. The process of protein ubiquitination is shown in blue. The sequential activity of E1, E2, and E3 Ub ligases conjugate Ub to substrates. Three major outcomes are shown including 1) trafficking of vesicular proteins to the lysosome via endocytic, mitophagy, and autophagy pathways; 2) alteration to protein function, e.g., enzyme activity, protein localization, or protein interactions; and 3) driving proteasomal degradation. Diverse Ub chain topologies can drive different outcomes, but chains of four or more K48-linked or K11-linked Ubs direct proteasomal degradation. NDD DUBs predominantly antagonize these outcomes by removing Ub. Note that DUBs can also stimulate Ub processes by contributing to the free Ub pool (e.g., UCHL1). (B) Current knowledge of the major cellular roles of NDD DUBs and candidate NDD DUBs. The roles of many of these DUBs are poorly characterized. DUB, deubiquitinating enzyme; EGF, epidermal growth factor; IFN, interferon; mRNA, messenger RNA; mTOR, mechanistic target of rapamycin; NDD, neurodevelopmental disorder; NFKB, nuclear factor-κB; TGFβ, transforming growth factor β; TNFα, tumor necrosis factor α; Ub, ubiquitin.

NDD DUBs CONTROL ABUNDANCE OF PROTEINS ENCODED BY NDD GENES

The remainder of this review focuses on the function of NDD DUBs in the UPS and explores mechanisms that converge on the regulation of protein abundance in the developing brain. For descriptions of non-UPS roles of DUBs in the brain, the reader is directed to a recent review (97). USP9X, USP7, and OTUD5 are currently the only NDD DUBs to have protein substrates identified in the context of brain development (Figure 3A). USP9X deubiquitination of the mTOR (mechanistic target of rapamycin) signaling component RAPTOR protects it from proteasomal degradation, thus promoting mTOR signaling and proliferation of NSCs (34). Other substrates depleted in Usp9x knockout embryonic brains include CTNNB1 (WNT signaling), ITCH, and NUMB (NOTCH signaling), and these signaling pathways are defective in USP9X-deficient NSCs (31). CTNNB1 also functions in cell adhesion, and NSC adhesion and polarity is defective in the absence of USP9X (29,33). USP9X also binds DCX, a master regulator of neuronal migration (23,98). C-terminal missense mutations in USP9X that cause NDDs result in loss of this interaction and reduced neuronal migration. Conversely, missense mutations in DCX that cause NDDs specifically disrupts interaction with USP9X (23,98). USP9X regulates additional substrates in developing neuronal synapses. USP9X deubiquitination of ANK3 (and other synaptic proteins including ANK2, SHANK3, and TNKS2) is required to antagonize its proteasomal degradation and is essential for synapse formation (30). Some USP9X mutations causing NDDs disrupt this interaction (30). The degradation of PRICKLE1 and PRICKLE2 is also antagonized by USP9X in the synapse and is required for control of neuronal excitability (25). A USP9X mutation in a male with epilepsy disrupts interaction with PRICKLE1 (25). Thus, USP9X is required to maintain substrate abundance to control different aspects of brain development that appear relevant to the pathology of NDDs.

Figure 3.

NDD DUBs control the dosage of NDD proteins in cells of the developing brain. (A) Overview of the known NDD DUB substrates identified in the context of cells of the developing brain and the cellular processes they regulate. (B) Interactomes of NDD DUBs. All known DUB-interacting proteins were extracted using the BioGRID resource. The number of proteins encoded by NDD genes within each NDD DUB interactome is shown (yellow) and the OR factor provided to highlight enrichment (p value represents an exact hypergeometric probability test). The connectivity of the NDD interactome components (PPI p value) were assessed using STRING. The proportion of haploinsufficient NDD proteins within the NDD DUB interactomes is shown (green) and defined using the gnomAD metric LOEUF score < 0.35. BioGRID, Biological General Repository for Interaction Datasets; DUBs, deubiquitinating enzymes; gnomAD, the Genome Aggregation Database; LOEUF, loss of function observed/expected upper bound fraction; NDD, neurodevelopmental disorder; OR, overrepresentation; PPI, protein-protein interaction; STRING, Search Tool for the Retrieval of Interacting Genes/Proteins.

Similar to USP9X, the NDD DUBs USP7 and OTUD5 also regulate brain development by promoting substrate abundance. USP7 protects the transcriptional repressor REST and transcriptional activators c-MYC and SOX2 from proteasomal degradation in NSCs to maintain an undifferentiated state (99). In hypothalamic neurons, USP7 protects TRIM27 from proteasomal degradation. In turn, TRIM27 opposes the delivery and degradation of endosomal cargo at the lysosome (20). USP7 also regulates the proteasomal degradation of the p53 E3-ligase MDM2. In the developing brains of mice lacking USP7, p53 levels become elevated, leading to widespread apoptosis (20). Studies of OTUD5 mutant iPSCs and NSCs revealed that it protects the chromatin regulators ARID1A, ARID1B, HCFC1, and HDAC2 from proteasomal degradation associated with decreased capacity of iPSCs to differentiate into NSCs. Substrates of the remaining NDD DUBs are undetermined in the context of the developing central nervous system.

Although knowledge of NDD DUB substrates in the developing brain is currently limited, it is striking that 70% of the known substrates are themselves encoded by NDD genes (USP9X/USP7/OTUD5 = 14/20) (Figure 3A). Because the genes encoding these substrates cause NDDs through LoF and DUBs promote the function of these genes by promoting the abundance of their encoded proteins, it follows that the function of NDD DUBs may converge to maintain the abundance of NDD protein networks. Support of this mechanism is derived by interrogating NDD DUB interactomes assembled en masse through proteome-scale interaction databases such as BioGRID, BioPlex, and STRING (12,13,100). For example, the BioGRID-derived NDD DUB interactomes are enriched with proteins encoded by NDD genes (Figure 3B; Tables S2 and S3). Furthermore, about half of the NDD genes within these interactomes are intolerant to LoF mutations, suggesting that they are haploinsufficient, thus requiring precise control of their abundance. The fact that the predominate role of several NDD DUBs is to promote the abundance of their interactors through antagonism of proteasomal (or lysosomal) degradation (Figure 2A) illustrates that NDDs caused by LoF of NDD DUBs may be, at least in part, underpinned by loss of abundance of proteins encoded by haploinsufficient NDD genes.

FUTURE DIRECTIONS

Systematic genome sequencing of individuals with NDDs continues to resolve the contributions that individual DUBs play in brain development. However, even for well-established DUBs, technical and experimental limitations have restricted our knowledge of their complex cell and developmental functions and protein-substrate interactions. The current knowledge of NDD DUBs is a piecemeal of studies focusing on different DUBs, different interactors, different cell types (and so on), predominantly generated in the context of cancer. How DUBs spatiotemporally regulate networks of protein substrates, particularly during physiological brain development, remains largely unexplored. Identification of DUB substrates is challenged by rapid substrate degradation and interaction kinetics. Indeed, traditional immunoprecipitation techniques used to capture interactomes yielded far lower numbers of DUB interactors than predicted (13,14,24,69). Innovations in identifying transient and low-affinity protein-protein interactions and proteomic techniques are providing new opportunities to prospectively identify and interrogate DUB interactions. Proximity biotinylation protein-protein interaction techniques such as BioID, are proving ideal for capturing transient and unstable interactions within the in situ spatially resolved environments of cells and organisms (30,101,102). Coupling these techniques with quantitative proteomics allows identification of interacting proteins en masse (24,69,101,102). The ubiquitination status of interactomes and entire proteomes (aka ubiquitinomes) can also be resolved using techniques such as Ub-remnant (di-glycine) immunoprecipitations (69,103), while the linkage topology of Ub chains decorating proteins can be assessed globally using Ub-TUBEs, Ub linkage–specific antibodies, and Ub-clipping technologies, among others (7,47,104–110). CRISPR (clustered regularly interspaced short palindromic repeats)–Cas9 and iPSC technologies have also made studying LoF alleles of DUBs in brain cell contexts more tractable. Entire proteomes and the impact that DUB LoF has on its interacting proteins can be derived by combining the above techniques with traditional paradigms of translational and proteasomal inhibition. Such advances are already driving new knowledge of DUB function, but the challenge is to apply them in the context of the developing brain using in vivo–based small animal and in vitro–based stem and neural cell models (47,69,111,112). Collectively, the ability to deep dive into the discovery of NDD DUB interactomes and their functional relationships in the developing brain can now be approached at the proteome scale with high temporal and spatial resolution.

THERAPEUTIC IMPLICATIONS

While it is apparent that DUBs regulate networks of proteins, the rate-limiting components of an NDD DUB interactome network may represent major drivers of phenotypic features and, as such, represent key therapeutic targets. ANK3, for example, is a key target in USP9X NDDs, because restoration of its abundance rescues synaptic defects in neurons lacking USP9X (30). The fact that USP9X regulates the abundance of ANK3 by antagonizing its UPS-dependent degradation reveals that increasing ANK3 abundance could be achieved by targeting other regulatory points in the UPS, for example, inhibiting upstream ANK3 cognate E3-ligases such as CDC20-APC (30). However, the emerging mechanism by which NDD DUBs function to antagonize the UPS degradation of proteins encoded by other NDD genes highlights a grander opportunity: NDDs caused by haploinsufficiency may be amenable to therapies that increase abundance of proteins encoded on wild-type alleles by inhibiting their UPS-based degradation (Figure 4). Antagonizing the UPS could be approached by targeting the regulatory elements imparting specificity, either by inhibiting cognate E3-ligases or activating cognate DUBs. The large numbers of DUBs and E3-ligases (~700 in total), coupled with their unique protein specificities, provide an extensive scope of therapeutic targets (113). The cancer field long viewed the UPS as a fertile therapeutic landscape (113). Small molecule inhibitors disrupting specific E3-ligase–substrate interactions now represent true precision therapies in pre/clinical trials (114,115), and developing modifiers of E3-ligase and DUB activity is an active field of research and industry (116–119). Indeed, the UPS is heralded as a means to “drug the undruggable (120)” because of the diverse ways in which it is being harnessed to protect or execute a protein’s existence. Adapting these technologies to increase the abundance of haploinsufficient proteins is largely unexplored. The identification and study of NDD DUBs is illuminating this possibility, and their ongoing study will resolve how the UPS can be targeted as another approach to overcome NDDs caused by haploinsufficiency.

Figure 4.

A therapeutic paradigm to overcome haploinsufficiency in neurodevelopmental disorders using antagonism of UPS. The major outcome of haploinsufficiency is loss of protein dosage. If the dosage of a haploinsufficient protein is also regulated by UPS, then antagonism of UPS may restore protein dosage and overcome the haploinsufficient state. UPS antagonism should be approached by targeting the elements of UPS harboring specificity, either by inhibiting cognate E3-ubiquitin ligases or activating cognate DUBs. With ~700 E3-ubiquitin ligases and DUBs, there is a large scope of targets that represent a fertile landscape of therapeutic exploration and are supported by successful application in the field of oncology. DUBs, deubiquitinating enzymes; LoF, loss of function; NDD, neurodevelopmental disorder; UPS, ubiquitin proteasome system; WT, wild-type.