LUBAC: a new player in polyglucosan body disease

Protein metabolism involves co-ordinated actions of synthesis and degradation. Ubiquitination is the major mechanism in regulating the degradation. However, it is also a major modifier of protein function regulation in numerous pathways [1,2]. During protein ubiquitination, an 8 kDa protein moiety, ubiquitin is attached to another protein via an enzymatic cascade. First, the enzyme E1 binds to ubiquitin in an ATP dependent manner, which is subsequently transferred to another enzyme E2 and finally, the third enzyme, E3 ubiquitin ligase, attaches the ubiquitin to its substrate [2]. There are three main types of E3 ligases: really interesting new genes (RING) type directly transfers ubiquitin from E2 to substrate; RING between RING (RBR) binds ubiquitin first and then transfers it to the substrate; finally, homologous to E6AP C-terminus (HECT) operates via a mechanism similar to RBR [3] (Figure 1). Irrespective of types, most E3 ligases catalyze the formation of an isopeptide bond between the C-terminus glycine of ubiquitin and a lysine of the substrate and in many cases extends the reaction between ubiquitin molecules to generate polyubiquitin chains [4]. Recently, a protein complex, linear ubiquitin chain assembly complex (LUBAC), consisting of two RBR type of ligases, heme-oxidized IRP2 ubiquitin ligase 1 (HOIL-1L) and HOIL-1L interacting protein (HOIP) and an adapter protein SHANK-associated RH domain interacting protein (SHARPIN) was shown to form an unusual peptide bond between the C-terminal carboxyl group of the donor ubiquitin glycine and the N-terminus amino group of the acceptor ubiquitin methionine, forming a linear polyubiquitin chain [5,6]. The major involvement of such linear ubiquitination has so far been in immune signaling and cell death-related pathways [7,8].

Glycogen, with up to 55 000 glucose units, is the largest molecule in the cytoplasm and serves as a critical energy storage primarily in the liver and muscle but also in the brain, heart and kidney [9,10]. Three Nobel prizes have been awarded in the field of glycogen metabolism and the biology is generally considered well understood. Glycogen synthesis is in particular thought of as settled science resulting from concerted actions between glycogen synthase (GS) which synthesizes glycogen chains and glycogen branching enzyme (GBE). Together, they assure regular branching, and radial, spherical growth of the molecule. Recent discoveries through the genetics of certain rare diseases shed light on additional proteins that operate alongside GS and GBE to ensure correct glycogen architecture [11-15]. In the absence of any of these proteins, some glycogen branches become overlong, driving the affected molecules (now called polyglucosans) to precipitate. Over time, the precipitating molecules accumulate into larger inclusions (PBs) that drive disease in affected organs [16,17]. Intriguingly, all these new proteins that regulate glycogen structure are either E3 ligases or proteins that regulate E3 ligases: (A) Lafora disease (LD) is a fatal progressive myoclonus epilepsy caused by mutations in the EPM2A or EPM2B genes encoding the proteins laforin and malin, respectively. Malin is a RING-type ligase, which tightly interacts with laforin. Extensive studies with these proteins show that malin's essential function is in glycogen structure regulation, [18,19]. Using in vitro experiments, malin was shown to incorporate lysine 48 (K48) or K63-linked ubiquitin chains on substrates leading to their proteasomal or autophagosomal degradation [20]. (B) HOIL-1L deficiency results in a disease in which polyglucosans accumulate in skeletal and cardiac muscle, resulting in skeletal and cardiomyopathy and heart failure necessitating cardiac transplantation [21]. While all HOIL-1L deficient patients accumulate polyglucosans, some depending on the mutation, also develop an immunological disorder [22–24]. A glycogen metabolism-related substrate of HOIL-1L is unknown, however, certain proteins are shown to undergo K48-mediated ubiquitination by the ligase followed by proteasomal degradation (reviewed in [25]). Interestingly, a loss of function mutation in HOIP also results in polyglucosans accumulations [26]. (C) Finally, loss of function mutation in KLHL24 gene leads to polyglucosans accumulations and fatal cardiomyopathy. The KLHL24 gene encodes KLHL24 which is a constituent of a Cullin-RING E3 ubiquitin ligase complex—CUL3-RBX1-KLHL24 complex, degrades keratin, type I cytoskeletal 14 protein via K48-linked ubiquitination and is involved in fatal epidermolysis bullosa [27,28]. How deficiencies of each of these ubiquitin ligases, and through what substrates and pathways, result in derangement of glycogen structure are not known.

In this review, we focus on LUBAC and its connection to PB disease. First, we describe LUBAC's known structure and function. Next, we detail the polyglucosan pathology of human patients and mouse models. Finally, we review potential therapeutic approaches for affected patients, relying on gene-based information and lessons learned in LD. For reasons mentioned above, this review cannot go far down the molecular pathway from LUBAC deficiency to PB formation, but hopes to serve as a launching point for research in this area.

The three components of LUBAC share similarities in domain structures and functions [29,30]. HOIL-1L has an N-terminal LTM followed by UBL domain, an NZF domain and a C-terminal RBR region. The 123 kDa HOIP comprises an N-terminus PUB domain, a classical ZF motif, two NZF domains, two UBA domains (UBA1 and UBA2) and a C-terminal RBR domain followed by a unique region named linear ubiquitin chain determining domain (LDD). SHARPIN is the simplest among the three containing only an N-terminal PH domain followed by LTM, an UBL and NZF domain. The LUBAC complex is formed by the interactions of HOIL-1L and SHARPIN UBLs with UBAs of HOIP [31] as well as by HOIL-1L and SHARPIN's LTMs interactions [32]. A cysteine residue in the RING2 of HOIP together with a C-terminal region makes the LDD and forms the minimal catalytic core for LUBAC [33]. In isolation, this region of HOIP is sufficient forming M1-linked linear ubiquitin chains, although much less efficiently than the RBR–LDD together [34,35].

As stated above, the main catalytic core for M1-linked ubiquitination resides with HOIP's RBR and LDD. A bipartite binding between HOIP-RING1 to E2 and HOIP-IBR to the activated ubiquitin molecule is suitable for transferring the E2-ubiquitin to the RING2 domain [36]. The RING2 domain interacts with the canonical hydrophobic patch and polar arginine residues of the donor ubiquitin. A beta-hairpin located within the LDD further positions the carboxylate group of the C-terminal G76 of the donor ubiquitin towards the catalytic cysteine (C885) of HOIP. The acceptor ubiquitin is locked in place by HOIP's RING2-LDD so that only the N-terminal Met1 and none of the lysines of this ubiquitin molecule are in close proximity to the catalytic cysteine. A conserved histidine residue (H887) of HOIP's RING2-LDD forms a hydrogen bond with M1 of the acceptor ubiquitin and acts as a base to deprotonate and activate the amino group nucleophile of the M1 residue [37].

A key question in the LUBAC field is whether there is a catalytic role for HOIL-1L in M1 ubiquitination. While the answer is not completely established, based on recent studies, the following functions of the ligase are starting to emerge:

• (1) Attaching the first ubiquitin prior to M1-linked chain formation: In in vitro ubiquitination assays, HOIP‘s RBR–LDD can form free ubiquitin chains. However, only the full-length HOIP/HOIL-1L complex or HOIP-RBR–LDD/HOIL-1 complex is capable of ubiquitinating a physiological protein substrate [34]. Furthermore, full-length HOIL-1L inhibits free ubiquitin chain formation in solution. In fact, mutation in the active site of HOIL-1L (C460A) impairs the ubiquitination of substrate but enhances the formation of free ubiquitin chains by HOIP in solution [34]. These observations suggest that HOIL-1L is critical in targeting the first ubiquitin as well as regulating HOIP's free chain-forming activity at least in in vitro setting. Indeed, in cell cultures, it was shown that the HOIP/HOIL-1L complex has higher affinity for NEMO compared with HOIP alone [38].

• (2) A potential role in K63/M1 hybrid chain formation? It was recently reported that most of the M1-linked ubiquitin chains, generated by LUBAC, are hybrids containing both M1 and lysine 63 (K63)-linked polyubiquitin chains [39,40]. Based on these findings, a second mechanism for M1 ubiquitination was proposed, namely that LUBAC substrates are first ubiquitinated by K63-linked small polyubiquitin chains and M1-linked chains are added subsequently. Whether or not HOIL-1L is involved, is not yet known.

• (3) Unique oxyester linked ubiquitination by HOIL-1L: The same group as above recently reported that an additional slowly migrating HOIL-1L band on the Western blot is actually a monoubiquitinated species of the protein [41]. It was furthermore shown that this could be hydrolyzed by hydroxylamine, a chemical that selectively cleaves ester bond but not isopeptide (such as K-linked ubiquitination) or peptide (such as M1-linked ubiquitination) bonds. These results suggested that HOIL-1L monoubiquitinates itself through an unusual oxyester linkage (O-linked ubiquitination) [42,43] rather than a canonical lysine-mediated bond. This oxyester-linked monoubiquitinated species of HOIL-1L was enriched within LUBAC. Apart from oxyester-mediated autoubiquitination, it has been suggested that some populations of HOIL-1L substrates also undergo O-linked ubiquitination prior to M1-linked chain attachments [41,44]. An interesting addition came from a recent report, which showed that HOIP produces M1-linked linear ubiquitin chains onto the LUBAC substrate; this is then branched by the HOIL-1L mediated addition of O-linked ubiquitin molecules, suggesting HOIP–HOIL-1L together form a single catalytic center for substrate ubiquitination [45].

• (4) Monoubiquitination and inhibition of LUBAC by HOIL-1L: HOIL-1L autoubiquitination was separately reported by a second group [32,46]. Here, the majority of ubiquitinated HOIL-1L species were resistant to hydroxylamine treatment but digested by a deubiquitinase that cleaves K-linked ubiquitin linkages. Their experiments identified ubiquitinated lysine residues of HOIL-1L, HOIP and SHARPIN and a serine residue of HOIP. Strikingly, a mutant HOIL-1L where all the lysine residues were changed to arginine, failed to monoubiquitinate HOIL-1L but increased the activity of LUBAC. Furthermore, HOIL-1L was found to monoubiquitinate HOIP and SHARPIN, priming those HOIP substrates for further linear ubiquitin chain generation. Thus, this HOIL-1L-aided and HOIP-mediated M1-linked ubiquitination of LUBAC components curbs its function.

LUBAC is a unique E3 ubiquitin ligase complex, which can determine the type of ubiquitin chain irrespective of which E2 is being used [5]. This is due to a unique interaction between ubiquitin and HOIP's RING2-LDD [33]. Solitarily, HOIP's UBAs inhibits its catalytic activity [5,33,35]. The binding of SHARPIN and HOIL-1L UBLs to HOIP-UBA1 and UBA2, respectively, induces a conformational change within HOIP [31,47]. The RBR and LDD of HOIP are now structurally oriented to facilitate E2 loading and to perform the catalytic reaction. Furthermore, SHARPIN and HOIL-1 interact with each other through LTMs, which folds into a single globular domain upon heterodimerization producing a larger buried surface area than UBA–UBL interactions [32]. This stabilizes the trimeric LUBAC. The UBLs are now secured to HOIP's UBAs, forming a tight complex with the conformation needed for catalysis. Once LUBAC starts chain linkage, these are stabilized by multiple ubiquitin–LUBAC interactions. The purified NZF domain of HOIL-1L binds to M1-linked ubiquitin chains with higher affinity than K-linked chains [48,49]. This might be integral to the retention of LUBAC to the ubiquitinated substrate and the further addition of M1-linked ubiquitins to the substrate. In contrast, HOIP preferentially binds K63-linked ubiquitin oligomers [39]. SHARPIN NZF binds to both K63-linked and M1-linked ubiquitin chains [49,50].

LUBAC is emerging as an unusual protein complex where two RBR E3 ligases could form multiple unique ubiquitin linkages (O-linked, K-linked and M1-linked) (Figure 2A). Based on the available information, we propose scenarios for LUBAC catalysis (Figure 2B): (i) First, UBLs of SHARPIN and HOIL-1L interact with UBA1 and UBA2 of HOIP, respectively, to remove the auto-inhibitory state of HOIP; (ii) Next, HOIL-1L and SHARPIN interact via the LTMs, forming a stable globular tethering domain that further stabilizes the trimeric LUBAC and orients the complex for E2 recruitment and further catalysis. In this tight complex, a few scenarios are possible: (A) an oxyester-linked or lysine-linked ubiquitin bond is formed between the LUBAC substrates and ubiquitin by HOIL-1L, followed by the formation of K63/M1 hybrid chains. Note: Which E3 ubiquitin ligase attaches the K-63-linked polyubiquitin chains, is not known; (B) a lysine-linked ubiquitin bond is formed by HOIL-1L, followed by the formation of M1 chains by HOIP; (C) lastly, HOIP directly attaches M1-linked linear ubiquitin chains onto the substrate, which is then branched by HOIL-1L via an oxyester linkage. The tight control of LUBAC's activity comes primarily from two sources: (A) HOIL-1L monoubiquitinates all three components of LUBAC either via the oxyester bond or the lysine-mediated isopeptide bond triggering the M1-linked ubiquitination onto LUBAC subunits and inhibiting its function; (B) LUBAC-specific deubiquitinase OTULIN removes/trims M1 chains on LUBAC subunits, enhancing their function or on certain LUBAC substrates restricting their signaling [51,52]. Similarly, CYLD trims K63 and M1 chains on LUBAC substrates suppressing LUBAC mediated inflammatory pathways [53,54]. Intriguingly, both OTULIN and CYLD bind to the HOIP-PUB domain [55-58]. Therefore, given the complexity, there might be a possibility of their involvement in PB generation as well.

Considering the extensive involvement of LUBAC in immune signaling pathways [59], it is not surprising that LUBAC deficient patients and mice show immunodeficiency and autoimmune defects. Intriguingly, mutations in two LUBAC components—HOIL-1L and HOIP also showed PB accumulation in human patients (described below) or in a mouse model in the case of SHARPIN (our unpublished results).

Compared with only two patients with HOIP, multiple patients have been reported with homozygous or compound heterozygous mutations in the HOIL-1L gene [21]. OMIM has registered HOIL-1L deficiency as PGBM1 WITH OR WITHOUT IMMUNODEFICIENCY (OMIM entry #615895). All patients of PGBM1 exhibit PBs and have skeletal and/or cardiac myopathy, progressive muscle weakness and early disease onset - certain cases even before 1 year of age. So far, one patient has been described with additional mild cognitive impairment and periventricular white matter abnormalities on brain MRI [60]. The histopathological characteristics of PGBM1 differs from PGBM2 (caused by mutation in GYG1 gene) aiding in differential genetic diagnosis [61]. In contrast with large round, oval-shaped inclusions of PGBM2, PGBM1, shows smaller dotty inclusions within the cytoplasm and depletion of normal glycogen [61]. A recent article potentially explains this observation [62]. The study found that several glycogen metabolism, autophagosome-related and ubiquitin-proteasomal proteins accumulate within the PGBM1 PBs suggesting that sequestration of glycogen metabolism-related enzymes could cause the remarkable decrease in normal glycogen.

A subset of patients suffer immunodeficiency, systemic inflammation of multiple organs and autoinflammation (combined hyper-inflammation and immunodeficiency with early childhood death). It is intriguing that not all patients have the immune pathology. Overall, mutations near the N-terminus lead to the immune dysfunction and PB disease, whereas mutations farther from the N-terminus are associated only with the PB disease [63]. However, there are exceptions, which need to be explained, e.g. the homozygous A18P N-terminus mutation associated with solely PB disease. The explanation can be gleaned by combining the human mutation data with murine results. The first HOIL-1L deficient mouse model (HOIL-1L^−/−) was generated by deletion of exons 7 and 8, which throws the gene's coding sequence out of the frame and generates a premature termination codon. It was anticipated that nonsense-mediated decay (NMD) would completely eliminate the transcript and result in a full knockout. This mouse developed PBs but had limited immunopathology [64]. The second mouse model (HOIL-1L^null/null) was a true complete knockout, and was embryonic-lethal [32]. The explanation of the difference between the two models proved to be the retention of a smaller transcript sufficient to form LUBAC by LTM mediated interactions [32]. This suggest that LUBAC complex formation, even in small amount, could avert the immune disease, but not the PB disease. Returning to the human situation and the A18P mutation, this mutation sufficiently destabilizes and diminishes LUBAC complexes to induce PB formation, but retains some amount of LUBAC to avoid immune disease.

There are multiple mouse models for HOIL-1L. The first model, as mentioned, had the C-terminal E3 ligase domain of HOIL-1L deleted but expressed a ∼30 kDa fragment containing only the N-terminal region [65]. HOIP and SHARPIN protein levels were severely reduced but not completely abolished. Immunologically, these mice were healthy, except for susceptibility to certain bacterial pathogens. Older mice had PB accumulations in multiple organs [64] including the brain (our unpublished results). Complete HOIL-1L knockout was generated subsequently [32,66]. This led to embryonic lethality at E10.5 due to endothelial cell death. In a third model, only the first RING1 domain of HOIL-1L was deleted [46]. These mice expressed lower levels of HOIL-1L, but HOIP and SHARPIN were comparable to WT. The LUBAC complex was formed, and M1-linked ubiquitination function was elevated in all tested tissues. Finally, a C458S mutant was generated [41], which inactivates the E3 ligase activity. It will be critical to determine whether these last two mice develop the polyglucosan phenotype.

The first patient identified with HOIP deficiency had a homozygous missense mutation in the PUB domain [26]. The affected individual expressed 50% less HOIP mRNA, although no protein was detected by Western blotting. Both HOIL-1L and SHARPIN protein levels were also reduced. LUBAC was not formed as evidenced by co-immunoprecipitation experiments. However, SHARPIN interactions with non-ubiquitinated species of HOIL-1L remained intact. This could be due to the diminished amount of ubiquitinated HOIL-1L in the absence of HOIP, indicating that the amount of HOIP, perhaps influences the ubiquitination states of HOIL-1L as well. The patient accumulated PBs in muscle with lower extremity muscle weakness without any clinical signs of cardiomyopathy. The second patient had two point mutations, which resulted in aberrantly spliced RNA lacking either exon 7 or exon 9 [67]. The peripheral blood mononuclear cells showed reductions in protein amount for full-length HOIP, SHARPIN and HOIL-1L and complete abolition of ubiquitinated HOIL-1L. The interactions between the components of the trimeric LUBAC were also compromised. Muscle biopsy was not performed, and the presence of PBs could, therefore, not be assessed.

B cell and T-reg cell-specific deletion of HOIP reduced the amounts of HOIL-1L and SHARPIN [68,69]. Similar to HOIL-1L, a complete HOIP knockout was embryonically lethal between E11.5 and E12.5 due to angiogenic defect [70]. Interestingly, HOIP functions seemed to be important in muscle as indicated by a study with Drosophila HOIP ortholog, LUBEL, down-regulation of which caused decreased accumulation of ubiquitinated proteins in ageing muscles [71].

Chronic proliferative dermatitis (cpdm) was shown to result from a spontaneous mutation in the Sharpin gene that deletes a single nucleotide in exon 1 [72]. This results in an early premature stop codon and complete absence of SHARPIN protein. Sharpin^cpdm pups are viable, however, adults display numerous conditions necessitating euthanasia by 9–12 weeks of age. These include systemic inflammation of multiple organs, severe dermatitis with inflamed and scaly skin, defective lymphoid organ developments in spleen, lymph nodes and absence of Peyer's patches in adult mice [72–74]. In our unpublished work with these mice, we document the presence of PBs in this genotype as well.

Research in the LUBAC field unveiled its critical role in diverse signaling pathways, especially immune-related ones. However, its function in glycogen metabolism is completely unstudied, though obviously present. One key question is whether HOIL-1L is a main player in this role, which seems to be the case, since all HOIL-1L deficient patients develop PBs. With various possibilities for LUBAC catalysis identified (Figure 2), HOIL-1L's ligase function, previously considered possibly dispensable, is presently emerging front and center in LUBAC function. Hence, identifying new physiological substrates for LUBAC that opens a pathway to glycogen metabolism is now of primary importance. Therefore, the future of LUBAC related PB research should be guided by using existing HOIL-1L mouse models and creating tissue-specific knockout models for HOIP to decipher their involvement in PB pathogenesis.

In recent years, there has been extensive development of LUBAC inhibitors affecting the linear ubiquitination activity. Despite being effective, they all present pitfalls. BAY 11-7082, a known anti-inflammatory compound inhibits HOIP catalysis possibly by preventing the transfer of ubiquitin from E2 to the catalytic cysteine of the protein, but is non-specific [75]. Gliotoxin, a fungal metabolite, showed great promise; however, had cytotoxic effects [76]. Bendamustine, a drug usually used for leukemia and lymphoma, was effective in suppressing LUBAC activity, but inhibited other E3s as well [77]. Furthermore, these compounds were found to be ineffective in suppressing NF-κB activation at non-toxic concentrations [78]. The most effective compounds in this regard were HOIPINs, which covalently bind C885 of HOIP and specifically inhibit LUBAC [78,79]. Finally, a series of covalent inhibitors targeting the RBR region of HOIP [80] and peptide inhibitors blocking HOIP/HOIL-1L UBA–UBL interactions showed specificity [81]. Nonetheless, further characterization is warranted. Recently, thiolutin was identified as a potential inhibitor of LUBAC, which inhibited tumor growth in a preclinical mouse model of lymphoma [82].

In the absence of any definite treatment and with the advancements of gene therapies for rare diseases in recent years [16,83,84], LUBAC mediated PB disorders could benefit from gene replacement approaches.

The observation that the polyglucosans phenotype in patients and in mouse models have been primarily associated with mutations in HOIL-1L suggests that this gene is an ideal candidate for future targeted gene therapy. The size of the HOIL-1L cDNA (2.4 kb) is suitable for packaging in AAV9, presently the flagship vector for gene therapy. Intravenous delivery would be needed to access skeletal muscle and heart. Possibly, even partial expression in cardiac muscle may suffice to rescue a patient's heart, and at least partially improved the skeletal myopathy. AAV9 crosses the blood-brain barrier [85,86] and if there is, any central nervous system disease, so far suggested in one patient, may partially correct the functional deficiency in that organ as well.

Another new therapeutic avenue could be targeting the GS gene for down-regulation. Since polyglucosans, by definition, are glycogen molecules with overlong branches, it was theorized that down-regulating GS activity might correct the problem. This was confirmed in LD mouse models. Down-regulating GS was achieved with (1) crosses between LD models and mice deficient of GS or proteins that activate this enzyme [87-90]; (2) delivering Gys1 targeting CRISPR/Cas9 to the brain with AAV9 [91] and (3) antisense oligonucleotides (ASO) against GS gene [92]. In all cases, PBs were drastically reduced and consequent neuropathological and behavioral phenotypes improved. Since the basis of this approach is blind to the basic mechanisms of PB formation, the same ASO could be directly repurposed for any PB disease, including HOIL-1L deficiency.

Glycogen metabolism in the past, and ubiquitination in the present, are some of the most studied fields of biology. They clearly intersect, including at LUBAC, and when this does not happen, fatal cardio-musculoskeletal and neurological diseases result. Uncovering the ubiquitination pathways and mechanisms in this space will be important for further understanding of the multiple functions of ubiquitination in glycogen metabolism, and their roles in skeletal muscle, heart and brain health. LUBAC is the newest, and possibly most complex, player in ubiquitination. Much about its role in immune pathways has already been uncovered, but multiple other cellular pathways unsurprisingly utilize this intricate machinery. Ongoing deconstruction of the molecular basis of LUBAC, and pathways it regulates, are exciting new frontiers in ubiquitin biology research in health and disease.

The authors declare that there are no competing interests associated with the manuscript.

This work was funded by the National Institutes of Health under award P01NS097197 to B.M. and Dr. Roy Elterman Pediatric Neurology Research Fund awarded to S.M.

Open access for this article was enabled by the participation of University of Texas Southwestern Medical Center in an all-inclusive Read & Publish pilot with Portland Press and the Biochemical Society under a transformative agreement with EBSCO.

A.A., B.M. and S.M. wrote and edited the manuscript.

CUL3

Cullin 3

HOIPINs

HOIP inhibitors

KLHL24

Kelch like protein 24

LTM

LUBAC tethering motif

LUBEL

linear ubiquitin E3 ligase

NEMO

NF-κB essential modulator

NF-κB

nuclear factor Kappa B

NZF

Npl4-type zinc finger

OTULIN

OTU deubiquitinase with linear linkage specificity

PH

pleckstrin homology

PUB

PNGase and UBA or UBX containing protein

UBA

ubiquitin associated

UBL

ubiquitin like

ZF

zinc finger