Linear ubiquitin chain‐binding domains

Ubiquitin modification (ubiquitination) of target proteins can vary with respect to chain lengths, linkage type, and chain forms, such as homologous, mixed, and branched ubiquitin chains. Thus, ubiquitination can generate multiple unique surfaces on a target protein substrate. Ubiquitin‐binding domains (UBDs) recognize ubiquitinated substrates, by specifically binding to these unique surfaces, modulate the formation of cellular signaling complexes and regulate downstream signaling cascades. Among the eight different homotypic chain types, Met1‐linked (also termed linear) chains are the only chains in which linkage occurs on a non‐Lys residue of ubiquitin. Linear ubiquitin chains have been implicated in immune responses, cell death and autophagy, and several UBDs ‐ specific for linear ubiquitin chains ‐ have been identified. In this review, we describe the main principles of ubiquitin recognition by UBDs, focusing on linear ubiquitin chains and their roles in biology.


Introduction
Ubiquitin, an 8.5 kDa protein conserved from yeast to human, is utilized to post-translationally modify substrate proteins [1,2]. Covalent attachment of ubiquitin to a protein (ubiquitination) regulates their activity, stability, and folding, and thus biological functions, in different species [3]. Unlike other post-translational modifications, such as acetylation or phosphorylation, ubiquitination adds a complex protein tag to its substrate. To date, more than 10 000 ubiquitination sites in over 4000 human proteins have been uncovered by mass spectrometry analysis [4].
Like other post-translational modifications such as phosphorylation and methylation, ubiquitination is reversible (Fig. 1A). Substrate ubiquitination is achieved by a multistep ATP-dependent enzymatic cascade catalyzed by an E1 activating enzyme, an E2 conjugating enzyme and an E3 ligase [5], whereas deubiquitination is performed in a single step by dubiquitinases (DUBs) (Fig. 1A). In humans, two genes encode E1s,~25 genes encode E2s, and > 600 genes encode E3s. The nature of the ubiquitin modification depends on the E2 enzyme and the E3 ligase. Ubiquitin itself can also be post-translationally modified by phosphorylation, acetylation, sumoylation, and neddylation (Fig. 1B) [6]. The roles of post-translationally modified ubiquitin are not well-understood. However, it is clear that phosphorylation of ubiquitin by PTENinduced "Putative Kinase 1" (PINK1) plays a critical role in regulating mitophagy, a mitochondrial damage response [7,8].
A single ubiquitin moiety can be conjugated on a substrate (monoubiquitination) (Fig. 1C). Modification of ubiquitin by phosphorylation, acetylation, sumoylation, or neddylation can occur before or after the monoubiquitination event is completed (Fig. 1C). Ubiquitin chains are established via conjugation to intrinsic ubiquitin residues, including seven Lys residues (Lys6, Lys11, Lys27, Lys29, Lys33, Lys48, or Lys63) or the N-terminal Met1 residue [6,9,10]. This process yields eight possible linkage types within homologous, mixed, or branched chains (Fig. 1D,E) [11]. Ubiquitin chains may contain a mixture of modified and nonmodified ubiquitins (Fig. 1F,G). This variation is important for mediating spatiotemporal regulation of cellular signaling. Ubiquitination not only alters the substrate protein's binding partners, but may also directly induce conformational changes within the substrate protein [12,13], influencing critical steps of cellular signaling pathways.

Ubiquitin-binding domains (UBDs)
A wide range of proteins with "Ubiquitin-Binding Domains" (UBDs) recognize diverse ubiquitination types on specific substrate sites, thereby forming distinct signaling complexes ( Fig. 2A-D) [14]. Thus far, over 20 different families of UBDs have been identified [15,16]. UBD-containing proteins interact with ubiquitin monomers or chains on substrates. Ubiquitin-UBD interactions may occur in trans (between two molecules) ( Fig. 2A) or in cis (within a molecule) (Fig. 2B). Recruitment of UBD-containing proteins into a signaling complex via ubiquitin-binding regulates downstream signaling pathways [15]. Ubiquitination of a protein with an intrinsic UBD domain may influence its folding [17], which in-turn affects its activity [13] or its interacting partners [16,18,19] to impact signaling cascades.
As mentioned above, recent studies have shown that ubiquitin can be phosphorylated, acetylated, sumoylated, and neddylated. These modifications of ubiquitin potentially alter the UBD-binding surface or chain formation [20][21][22], therefore it is tempting to speculate that there are distinct UBDs recognizing chains with modified ubiquitins as a monomer or as a chain (Fig. 2C,D). The identification of such UBDcontaining proteins would extend our knowledge of ubiquitin biology and its roles in regulating physiology and disease. Alternatively, the post-translational modification of ubiquitin might inhibit canonical ubiquitin-UBD interactions.
UBDs that bind to ubiquitin chains are often selective for a specific type of chain linkage. This selectivity may arise from the recognition of a unique orientation of the chain and distinct surfaces on the ubiquitin moieties, or via direct interaction with the linker region connecting the two ubiquitins. Based on the structural analysis of UBDs, required length of ubiquitin chains to confer specificity to a UBD is suggested to be rather short, either dimeric or trimeric [9,19,[23][24][25]. UBDs in ubiquitin-binding proteins generally occur as single domains or in tandem ( Fig. 2A). Tandem "Ubiquitin-Interacting Motifs" (UIMs) in Rap80 and "Arsenite-Inducible RNA-Associated Protein-Like" (AIRAPL) proteins bind cooperatively to Lys63-and Lys48linked ubiquitin chains respectively [26,27]. An individual UBD may also bind multiple ubiquitin molecules, such as the single "Motif Interacting with Ubiquitin" (MIU) domain within the "MIU-containing Novel DUB family member" (MINDY), which binds to three ubiquitin molecules linked through Lys48 [28]. Also, a double-sided UIM in the "Hepatocyte growth factorregulated tyrosine kinase substrate" (Hrs) protein is known to bind two independent ubiquitin molecules on either side of an a-helix [29].

NEMO-UBAN
The UBAN domain in "NF-jB essential modulator" (NEMO)/"IjB kinase gamma" (IKKc) was the first linear ubiquitin chain-specific binding domain identified [47] (Fig. 4A), and has been structurally analyzed by x-ray crystallography (Fig. 5A) [19,48]. The C-terminal zinc finger (ZF) of NEMO co-immunoprecipitates with polyubiquitinated signaling components, and thus NEMO was speculated to interact with Lys63-linked ubiquitin chains. Unexpectedly, NEMO-UBAN interacts with linear diubiquitin chains with approximately 100-fold higher affinity compared to Lys63-linked diubiquitin chains [19,49]. NEMO-UBAN forms a parallel homodimer that adopts a highly symmetric coiled-coil structure, resulting in the simultaneous binding of two linear diubiquitin chains along NEMO-UBAN (Fig. 5A). NEMO binds to the distal and proximal ubiquitins within a linear diubiquitin via distinct surfaces centered on Ile44 and Phe4 respectively (Fig. 5A). Furthermore, tight interactions between NEMO-UBAN and the C-terminal tail of the distal ubiquitin (residues Leu71-Arg74) that forms a linker between two ubiquitin molecules ensures its specificity for linear ubiquitin chains. NEMO plays an important role as part of the "IjB Kinase" (IKK) complex in the "Nuclear Factor-KappaB" (NF-jB) activation signaling cascade, both as a linear ubiquitination substrate and linear ubiquitin chain binding protein (Fig. 3A). The interaction of linear ubiquitin chains with NEMO-UBAN alters the conformation of NEMO [17,19], which may have an allosteric effect, regulating IKK kinase activity and downstream signaling. Genetic [19] and pharmacological [50] inhibition  of the interaction between NEMO-UBAN and linear ubiquitin chains preclude the full activation of NF-jB, revealing an important role for the NEMO-linear ubiquitin chain interaction in regulating the NF-jB pathway. Whether NEMO-UBAN binds to linear ubiquitination formed on NEMO itself, and whether this interaction affects the oligomeric state of the IKK complex, remain unclear. Patients suffering from incontinentia pigmenti and "Ectodermal Dysplasia with Immuno Deficiency" (EDA-ID) display mutations in the NEMO gene, including at positions within the UBAN domain (Fig. 4A, Fig. 5A, Table 1) [51][52][53]. Interestingly, some of these mutations (Asp311Asn, Glu315Ala, and Arg319Gln, corresponding to Asp304Asn, Glu308Ala and Arg312Gln in mouse NEMO as in Fig. 5A) prevent the interaction between NEMO-UBAN and linear ubiquitin chains, suggesting that loss of this interaction might underlie this disease [19].

ABIN-1, 2, 3-UBAN
The three "A20-Binding Inhibitors of NF-jB" (ABIN) proteins are known to inhibit the NF-jB pathway and the ABIN-UBAN domains are important for this regulation (Fig. 3A and 4B) [54]. The ABIN proteins are implicated in immunity and cell death based on the analysis of genetically modified mouse models and cellular assays [47]. The three ABIN proteins appear to play nonredundant roles in NF-jB signaling. For example, NF-jB activation induces the expression of ABIN-1 and ABIN-3 but not ABIN-2 [47], suggesting distinct functions in cell signaling regulation.
The UBAN domains in the ABIN proteins were first identified in 2008 [54] and shown to negatively regulate NF-jB signaling. However, a structural understanding of the ABIN-UBAN domain remained unclear until recently. A crystal structure of the ABIN-2-UBAN domain in complex with a linear triubiquitin chain indicated that this domain forms a coiled-coiled homodimer that provides binding sites for two of the ubiquitins, similar to NEMO-UBAN (Fig. 5B,C) [24]. From each triubiquitin chain, the distal ubiquitin binds the UBAN domain through its Ile44 surface. The middle ubiquitin employs both the Phe4-and Ile-44 centered surfaces, acting as the proximal ubiquitin for the same UBAN domain and the distal ubiquitin for the second UBAN dimer from the neighboring asymmetric unit. The third ubiquitin in the chain forms the proximal ubiquitin for binding to the second UBAN dimer (Fig. 5B). Although each ABIN2-UBAN dimer appears to bind one linear triubiquitin in solution as well as in each asymmetric unit of the crystal, the crystal packing reveals a second triubiquitin chain on the other side of the ABIN2-UBAN dimer (Fig. 5B). Thus, high local concentrations of linear ubiquitin chains might lead to the formation of a complex in which two linear ubiquitin chains bind either side of a ABIN2-UBAN dimer. It would be important to understand how triubiquitin chains may support oligomerization of UBAN-containing proteins like ABIN-2. For instance, diubiquitin versus triubiquitin chain binding to ABIN-2 might trigger differential readouts in downstream signaling activation, since only the latter would be predicted to induce oligomerization.

OPTINEURIN-UBAN
The "Optic Neuropathy-Inducing" (OPTINEURIN)-UBAN domain was initially predicted by bioinformatic analysis together with other UBANs in ABIN proteins and NEMO (Fig. 4C) [54]. A structural study of OPTINEURIN-UBAN in complex with linear tetraubiquitins revealed a binding mode similar to that of linear ubiquitin with NEMO and ABIN-2 [23]. In this crystal structure, two tetraubiquitin chains within neighboring asymmetric units interact with opposing faces of an UBAN dimer, with two ubiquitins from each chain interacting with one face of a UBAN dimer (Fig. 5D). OPTINEURIN is a multifunctional protein,  involved in protein trafficking by vesicles, autophagy and signal transduction. OPTINEURIN functions as an autophagy receptor to regulate selective autophagy at least partially through its UBAN domain, and its "microtubule associated protein 1 light chain 3 (LC3)interacting region" (LIR) (Fig. 3B, Fig. 4C) [60]. Furthermore, the OPTINEURIN-UBAN/linear ubiquitin chain interaction regulates the selective autophagy pathway through "TANK Binding Kinase 1" (TBK1) kinase-dependent phosphorylation [35]. OPTI-NEURIN is also a negative regulator of the NF-jB signaling cascade (Fig. 3A).
Some of the patients suffering from glaucoma or amyotrophic lateral sclerosis (ALS) display mutations in OPTINEURIN (OPTN). A Glu478Gly mutation within the UBAN domain of OPTINEURIN (Fig. 4C,  Fig. 5D, Table 1) [61], which recapitulates a mutation in ALS patients, disrupts its interaction with linear tetraubiquitin chains and abolishes OPTINEURINmediated inhibition of NF-jB signaling [23]. Together, these data suggest that linear ubiquitin chain recognition by OPTINEURIN is important for human health.
The UBAN domains in NEMO, ABINs and OPTI-NEURIN ( Fig. 4A-C) display a similar interaction mode with linear ubiquitin chains (Fig. 5A-D), yet they have either positive or negative roles in TNFinduced immune signal regulation (Fig. 3A). It will be of interest to determine how their specific UBANdependent interacting partners and biological functions are precisely regulated to activate or inhibit the downstream signaling cascade.

HOIL-1L-NZF
HOIL-1L is a component of the LUBAC complex responsible for linear ubiquitination [62]. HOIL-1L itself is not considered to have the central catalytic 'center' of LUBAC, but HOIP, for generating linear ubiquitin chains. Linear ubiquitin-binding by HOIL-1L is mediated by its "Npl4 zinc finger" (NZF) domain and its C-terminal a-helical tail extension, referred to as the NZF-tail (Fig. 4D) [63]. The two domains are linked via a loop region, which itself does not interact with the ubiquitin chain (Fig. 5E) [63]. The NZF domain binds the canonical Ile44 hydrophobic surface on the distal ubiquitin within a diubiquitin chain, whereas the NZF tail binds the Phe4 surface of the proximal ubiquitin (Fig. 5E). Unlike the UBAN domains, HOIL-1L does not make extensive interactions with the C-terminal tail of the distal ubiquitin. Thus, the specificity of HOIL-1L-NZF for linear ubiquitin chains appears to be determined by the relative spatial orientation and the spacing between the distal and proximal ubiquitins. The physiological relevance of the HOIL-1L NZF is not fully understood besides its implication in NF-jB activation [63]. The interaction between linear ubiquitin chains and the HOIL-1L-NZF might be required for LUBAC-dependent linear ubiquitination. Alternatively, this interaction may protect chains from DUBdependent hydrolysis, resulting in positive regulation of downstream signaling cascades. Furthermore, HOIL-1L might be recruited to the TNFR complex via the NZF domain.
As part of LUBAC, HOIL-1L regulates immune signaling cascades as well as selective autophagy by linearly ubiquitinating key signaling components, namely NEMO and RIPK2 (Fig. 3A). HOIL-1L NZF mutants that no longer interact with linear ubiquitin chains cannot fully activate NF-jB signaling in gene reporter assays [63], strongly suggesting that binding between HOIL-1L and linear ubiquitin chains is critical for regulation of the NF-jB signaling pathway. Mutations in the RBCK1 gene encoding HOIL-1L have been identified in patients suffering from polyglucosan body myopathy with immunodeficiency (Fig. 4D, Table 1) [64][65][66].

A20-ZF7
A20 protein was initially found to regulate NF-jB and cell death signaling pathways [67]. A20 harbors multiple functionally distinct domains that are important for regulating ubiquitin-dependent signaling cascades (Fig. 4E). For instance, A20 was first shown to acts as a ubiquitin E3 ligase to target "Receptor Interacting Protein Kinase 1" (RIPK1) for proteasomal degradation, and also as a DUB to hydrolyze Lys63-linked ubiquitin chains on RIPK1 and negatively regulate NF-jB activation [68]. Later, it was shown that A20 interacts with RNF11 and an E3 ligase Itch to negatively regulate NF-jB signaling [69], demonstrating its role as an E3 ligase may depend on these E3 ligases. More recently, A20-ZF7 (Fig. 4E) has been shown to have catalytically independent mechanisms of action [70,71] by supporting interactions between A20 and linear ubiquitin chains on TNFR1 and RIPK1 within a signaling complex (Fig. 3A) [72]. A20 ZF7 simultaneously binds the Ile44 hydrophobic patch on the distal ubiquitin within diubiquitin, and forms a hydrogen bonding network with a region on the a-helix of proximal ubiquitin, consisting of residues Gln31/Asp32 and the backbones of residues Ala28/Gly35 (Fig. 5F). Moreover, Leu71, Arg72, and Leu73 from the C-terminal tail of distal ubiquitin interact with A20 ZF7. Mutations in the linear ubiquitin-binding A20-ZF7 domain were found to be associated with B cell lymphomas (Fig. 4E, Table 1) [71,[73][74][75]. Mutations within the catalytic "Ovarian Tumor" (OTU) domain and the ZF4 domain of A20 have been identified in patients with autoinflammatory syndrome, Hodgkin's and non-Hodgkin's lymphoma, and diabetes [73,76,77].
UBAN and ZF domains use distinct mechanisms to interact with linear ubiquitin chains, however, they both recognize the same surface on the distal ubiquitin within a linear ubiquitin chain, suggesting that one  [93] c.2272C > T p.R758X Nonsense Familial cylindromatosis [88] For A20, R a = sequence variation present in non tumor cells suggesting a polymorphism [73].

2755
The The DUB catalytic domains for linear ubiquitin chains

OTULIN-OTU
Although the E3 ligase complex LUBAC specific for linear ubiquitination was discovered in 2006 [62], linear ubiquitin chain-specific DUBs were only recently discovered in 2013 [43,44]. The catalytic domain within the DUB OTULIN recognizes the linear ubiquitin dimer and specifically hydrolyzes linear ubiquitin chains ( Fig. 4F and 5G) [43,44]. OTULIN binds to distinct surfaces on distal and proximal ubiquitins within diubiquitin, and specifically recognizes residues at the C-terminal tail of the distal ubiquitin that link the two ubiquitin molecules [43]. Whereas the distal ubiquitin uses its Ile44 hydrophobic surface for binding, the proximal ubiquitin binds the OTU domain via its a-helix and Phe4-centered surface (Fig. 5F). Importantly, Glu16 from the proximal ubiquitin takes part in the activation of OTULIN, revealing a substrateassisted catalysis mechanism [27]. OTULIN negatively regulates various signaling cascades in which linear ubiquitination plays a role, such as immune signaling cascades and the cell death pathway mediated by the "TNF Receptor" (TNFR) or "Nucleotide-binding Oligomerization Domain-containing protein 2" (NOD2) (Fig. 3A). OTULIN is also known to directly interact with HOIP [78][79][80]. Interestingly, loss-of-function OTULIN mutations in mice lead to embryonic death between E12.5-E14 due to angiogenic deficits, indicating a role for OTULIN in vascularization [44].
Mutations within the catalytic OTU domain in OTULIN were identified in autoinflammation and dermatosis syndrome patients (Fig. 4F, Table 1) [81,82]. The autoimmune disease-related OTULIN mutations Leu272Pro and Tyr244Cys are not directly involved in interactions with linear ubiquitin chains but they are located on the helix arm that forms part of the binding pocket for the distal ubiquitin [43] (Fig. 5G). The Leu272Pro mutation of OTULIN suppresses the DUB-mediated hydrolysis of linear ubiquitin chains in vitro, supporting the physiological relevance and molecular function of linear ubiquitin binding by OTULIN [81].

CYLD-USP
CYLD is a tumor suppressor DUB that utilizes its catalytic specific protease (USP) domain to recognize ubiquitin chains and regulate immune signaling cascades (Fig. 4G). CYLD displays dual specificity for, and thus hydrolyzes, linear and Lys63-linked ubiquitin chains [80,83]. In the crystal structure of CYLD-USP in complex with linear diubiquitin, distal ubiquitin uses a hydrophobic patch centered on Ile44, as well as its C-terminal tail including residues Val70-Gly76, to interact with the USP domain (Fig. 5H). The CYLD-USP binding region on proximal ubiquitin involves residues from the Phe4 patch (Fig. 4E). Moreover, similar to the OTU domain, Glu16 of the proximal ubiquitin is shown to contribute to the hydrogen bonding with the CYLD-USP domain and thus to the catalytic activity of the CYLD enzyme. CYLD functions as an inhibitor in the TNFR and NOD2 signaling cascades by deubiquitinating both linear and Lys63-linked ubiquitin chains, which are important for mediating these signaling cascades (Fig. 3A) [46]. Interestingly, CYLD makes a signaling complex with HOIP via "Spermatogenesis Associated 2" (SPATA2) at TNFR [84][85][86][87]. The interplay between the deubiquitinase CYLD or OTULIN and the ubiquitin E3 ligase HOIP in a complex is an interesting aspect to consider for the regulation of downstream signaling cascades. Several mutations in CYLD, mostly within the USP domain, have been identified in patients with familial cylindromatosis (Fig. 4G, Table 1) [88][89][90][91][92][93].
Although OTULIN and CYLD have overlapping functions in the regulation of immune signaling cascades (Fig. 3A), their roles in vivo seem to be distinct based not only on their association with different human diseases but also on the distinct phenotypes of their genetically modified mouse models. Cyld À/À mice show no gross defects and are born at the expected Mendelian ratios [94], whereas OTULIN loss-of-function mutant mice (Gumby mice) are embryonic lethal [44]. Their distinct functions in biology likely depend on their different noncatalytic functions and interacting partners, as well as the regulation of their expression and activity.

Concluding remarks
The first linear ubiquitin chain-specific UBD was discovered in NEMO and shown to regulate the inflammatory NF-jB signaling cascade [19]. Since that discovery, roles for linear ubiquitination have been demonstrated in various diverse signaling pathways. Thus far, all proteins with linear ubiquitin chain