HOIL-1-catalysed ubiquitylation of unbranched glucosaccharides and its activation by ubiquitin oligomers

HOIL-1, a component of the Linear Ubiquitin Assembly Complex (LUBAC), ubiquitylates serine and threonine residues in proteins, forming ester bonds (Kelsall et al, 2019, PNAS 116, 13293-13298). Here we report that mice expressing the E3 ligase-inactive HOIL-1[C458S] mutant accumulate polyglucosan in brain, cardiac muscle and other organs, indicating that HOIL-1’s E3 ligase activity is essential to prevent these toxic polysaccharide deposits from accumulating. We found that HOIL-1 monoubiquitylates glycogen and α1:4-linked maltoheptaose in vitro and identify the C6 hydroxyl moiety of glucose as the site of ester-linked ubiquitylation. The HOIL-1-catalysed monoubiquitylation of maltoheptaose was accelerated >100-fold by Met1-linked or Lys63-linked ubiquitin oligomers, which interact with the catalytic RBR domain of HOIL-1. HOIL-1 also transferred preformed ubiquitin oligomers to maltoheptaose en bloc, producing polyubiquitylated maltoheptaose in one catalytic step. The Sharpin and HOIP components of LUBAC, but not HOIL-1, bound to amylose resin in vitro, suggesting a potential function in targeting HOIL-1 to unbranched glucosaccharides in cells. We suggest that monoubiquitylation of unbranched glucosaccharides may initiate their removal by glycophagy to prevent precipitation as polyglucosan.


Introduction
Mutations that reduce the expression of HOIL-1 (haem-oxidised IRP2 ubiquitin ligase-1), also called RBCK1 (RING-B-Box-coiled-coil protein interacting with PKC 1 or RANBP2-Type and C3HC4-type zinc finger-containing protein 1) cause both auto-inflammation and immuno-insufficiency in humans (Boisson et al, 2012;Phadke et al, 2020) and immunoinsufficiency in mice (MacDuff et al, 2015;Tokunaga et al, 2009). However, HOIL-1 deficiency in humans also leads to cardiomyopathy and death from heart failure in early adulthood (Boisson et al., 2012;Fanin et al, 2015;Krenn et al, 2018;Nilsson et al, 2013;Phadke et al., 2020;Wang et al, 2013), which is unrelated to the immune defects, and arises from the progressive accumulation of toxic polyglucosan bodies in cardiac muscle and other tissues, such as the brain, with some patients also displaying cognitive impairment (Chen et al, 2021;Phadke et al., 2020). Mice expressing low levels of HOIL-1 (Fujita et al, 2018) also form toxic polyglucosan bodies in cardiac muscle (MacDuff et al., 2015), brain and spinal cord (Sullivan et al, 2018), but it is the brain that is affected predominantly in mice, the animals displaying defects in learning, memory and motor coordination (Sullivan et al., 2018).
Polyglucosan bodies are dense inclusions of starch-like polysaccharide that are insoluble because they lack the α1:6 branch points found in the glucose polymer glycogen.
Consequently, they have defective metabolism compared to glycogen and are resistant to digestion with α-amylases (Hedberg-Oldfors & Oldfors, 2015). The frequency of the α1:6 branch points, which occur after about every twelve α1:4-linked glucose units, determines the topology, structure, and solubility of glycogen. Mutations in glycogen metabolising enzymes cause a variety of glycogen storage diseases characterised by aberrant glycogen deposits (Kanungo et al, 2018). HOIL-1, is a component of the linear ubiquitin chain assembly complex, LUBAC (Dittmar & Winklhofer, 2019;Kirisako et al, 2006;Liu & Pan, 2018). It is well established that this trimeric complex, comprising HOIL-1, HOIP (HOIL-1 interacting protein) and Sharpin (Shank-associated RH domain interactor) is required in signal transduction pathways that generate inflammatory mediators or lead to cell death (Gerlach et al, 2011;Ikeda, 2015;Ikeda et al, 2011;Sasaki & Iwai, 2015;Shimizu et al, 2015;Tokunaga et al, 2011). In these pathways, Met1-linked ubiquitin (M1-Ub) oligomers generated by the E3 ligase HOIP interact with several proteins that regulate innate immunity, such as the NEMO regulatory subunit of the canonical IκB kinase (IKK) complex. This leads to IKK activation and the phosphorylation and activation of its substrates, which include the transcription factors NFκ B and IRF5 (Lopez-Pelaez et al, 2014;Ren et al, 2014;Scheidereit, 2006) that stimulate expression of the mRNAs encoding many inflammatory mediators.
Like HOIP, HOIL-1 is a member of the "RING in-between RING" (RBR) subfamily of E3 ubiquitin ligases but, unlike HOIP, it does not generate Met1-linked ubiquitin oligomers. Instead, it catalyses the attachment of ubiquitin to serine and threonine residues in proteins, forming ester bonds, and its substrates include the IRAK1 and IRAK2 components of Myddosomes that have critical roles in inflammatory mediator production (Kelsall et al, 2019). The failure to generate ester-linked ubiquitin chains in knock-in mice expressing the E3 ligase-inactive HOIL-1[C458S] mutant can enhance or reduce the production of inflammatory cytokines, depending on the ligand, receptor and immune cell type (Petrova et al, 2021).
Since HOIL-1 ubiquitylates the hydroxyl side chains of serine and threonine residues (Kelsall et al., 2019;Rodriguez Carvajal et al, 2021), we hypothesized that it might also ubiquitylate the hydroxyl groups in glucose and so be able to ubiquitylate glycogen directly.
Here, we report that HOIL-1 does indeed ubiquitylate glycogen and the smaller model oligosaccharide maltoheptaose in vitro and identify the site of ubiquitylation. Based on these and other unexpected findings reported in this paper, we propose a new role for HOIL-1 and LUBAC in the ubiquitylation of unbranched glycogen molecules that may initiate their elimination from cells before they precipitate as toxic polyglucosan deposits.

Polyglucosan accumulates in the brain and heart of HOIL-1[C458S] mice.
HOIL-1 knockout (KO) mice exhibit early embryonic lethality (Fujita et al., 2018;Peltzer et al, 2018), but HOIL-1 "KO" mice expressing low levels of a truncated protein comprising the N-terminal Ubiquitin-Like (UBL) and Npl4 zinc finger (NZF) domains, but lacking the Cterminal region including the catalytic RBR domain (Fujita et al., 2018), accumulate polyglucosan in their brain, spinal cord and heart (MacDuff et al., 2015;Sullivan et al., 2018). Because HOIL-1 stabilises its binding partner HOIP, the expression of HOIP and Sharpin are also greatly reduced in these HOIL-1 "KO" mice. It was therefore unclear whether the accumulation of polyglucosan was caused by the loss of the E3 ligase activity of HOIL-1, the loss of the E3 ligase activity of HOIP or reduced expression of the non-catalytic domains of HOIL-1, HOIP or Sharpin. In contrast, HOIL-1, HOIP and Sharpin are expressed at normal levels in the E3 ligase-inactive HOIL-1[C458S] knock-in mice that we have described previously (Kelsall et al., 2019). We therefore investigated whether polyglucosan was present in the tissues of these mice. We found that polyglucosan did indeed accumulate in the brain, heart and other tissues of the HOIL-1[C458S] mice, but not in their wild-type littermates. Particularly high levels of polyglucosan were present in the hind brain (pons) and the hippocampus, the extent of deposition being similar in mice varying in age from 0.5-1.5 years ( Fig. 1A-D). Polyglucosan was present in lower amounts in the brain cortex (Figs. EV1A and EV1B). It was also present in small amounts in the heart, reaching a maximum level after about a year ( Fig. 1E and 1F), as well as in the lungs and liver (Figs. EV1C-F).
These results establish that the E3 ligase activity of HOIL-1 is required to prevent the accumulation of polyglucosan in several tissues.

HOIL-1 ubiquitylates glycogen in vitro
We considered how HOIL-1 might prevent the deposition of polyglucosan and wondered whether HOIL-1 might not only ubiquitylate the hydroxyl side chains of serine and threonine residues in proteins, but also the hydroxyl moieties present in glucose. To investigate whether HOIL-1 was capable of ubiquitylating glycogen directly we initially used an in vitro fluorescence-based assay in which commercially available bovine liver glycogen was incubated with HOIL-1, E1, E2, Mg 2+ -ATP and fluorescent Cy5-labelled ubiquitin.
Following SDS-PAGE, the glycogen (detected by Periodic acid-Schiff [PAS] staining), is too large to enter the gel and remains at the origin ( Fig. 2A), as does a portion of the fluorescent Cy5-labelled ubiquitin (Fig. 2B, lanes 5 and 6). The comigration of Cy5-labelled ubiquitin and glycogen required the presence of every reaction component. This high molecular mass Cy5-ubiquitin signal was lost if the reaction was incubated with α -amylase to degrade the glycogen (Fig. 2B, lanes 7 and 8) or incubated with the oxyester-specific nucleophile hydroxylamine (Fig. 2B, lanes 9 and 10). These experiments indicated that HOIL-1 had ubiquitylated glycogen and that ubiquitin was attached to it via an oxyester bond.

Small malto-oligosaccharides are mono-ubiquitylated by HOIL-1 in vitro.
The majority of the glucose units in glycogen are linked via α -1,4-glycosidic bonds. The linear oligosaccharide maltoheptaose contains seven glucose units linked by such bonds (Fig.   2C) and was used as a simple model substrate. The replacement of glycogen by maltoheptaose in the in vitro ubiquitylation reaction led to the formation of a single more slowly migrating form of ubiquitylated maltoheptaose in a concentration and time-dependent manner (Figs. EV2A and EV2B). As observed for glycogen, this adduct was sensitive to treatment with both hydroxylamine and α -amylase, although the ubiquitin was still bound to maltose and maltotriose after digestion with α -amylase, because complete hydrolysis to glucose did not take place (Zakowski & Bruns, 1985) (Fig. 2D, lanes 7 and 8). Mass spectrometry analysis of ubiquitylated-maltoheptaose established that it was a monoubiquitylated species (Fig. 2E) N, 13 C] ubiquitin-maltoheptaose shows the characteristic peak patterns of ubiquitin and any peak movements could be reassigned readily. Two major amide chemical shifts differences can be observed at the C-terminus, namely G75 and G76, when compared to free ubiquitin (Fig. 3A). Most notably, the amide 15 N and 1 HN signals of G76 are shifted upfield by -5.4 ppm and downfield by 0.3 ppm, respectively, which is consistent with the average C-terminal amide values when the charged carboxylate is removed by modification (Ulrich et al, 2008). Taken together with the observation that amide linewidths are comparable to free ubiquitin, these data indicate that a single ubiquitin moiety is conjugated to the oligosaccharide via its C-terminal glycine residue. Interestingly, the C-15 N and 1 HN signals of G76 exhibit multiple chemical shift environments, with at least four resolved positions, suggesting different isomers of the conjugate, with individual maltoheptaose molecules being singly ubiquitylated at distinct sites along the oligosaccharide.
Comparison of the anomeric region of the 1D 1 H NMR spectrum for the ubiquitylated maltoheptaose with free maltoheptaose (Fig. 3B), shows that while the anomeric protons of the α and β reducing end sugar are largely unaffected, there are a number of significant chemical shift changes for the anomeric protons of the other glucose residues. As the anomeric position of these sugar residues cannot be targeted by the E3 ligase, the chemical shift changes reflect the effects of ubiquitylation on other positions within these sugars.
To identify unambiguously which position(s) on the glucose units were modified with ubiquitin, we exploited the hybrid labelling scheme of the protein-carbohydrate conjugate and used an 1 H-13 C HMBC experiment to detect the 3-bond scalar coupling between the sugar protons and the 13 C carbonyl of G76. These 3 J COCH couplings are typically 2-3 Hz and correlations in the HMBC would reveal the specific ligated positions. Only one significant correlation could be observed at 4.30 and 4.50 ppm (Fig. 3C), corresponding to two protons of a CH 2 group, which could be readily assigned using 1 H correlation NMR spectroscopy to the sidechain 6-CH 2 position of the glucose residues.
To confirm these assignments, we prepared an unlabelled version of ubiquitylated maltose and compared the 1 H-13 C HSQC spectrum with that for free ubiquitin (Fig. 3D). The conjugated 6-CH 2 group is clearly identifiable at 4.30 and 4.50 1 H ppm (67.0 13 C ppm) and can be assigned to the non-reducing glucose unit. These chemical shifts match those for 6-CH 2 groups typically found in (α1-6) glucose chains (Dobruchowska et al, 2012), thereby confirming the 6-position modification. We also note a minor set of proximal CH 2 peaks in this region of the spectrum, which could either represent a different conformation of the nonreducing 6-CH 2 group due to restriction in free bond rotation from the attached ubiquitin or some ligation to the reducing end sugar 6-CH 2 group, although there are no apparent chemical shift changes for the anomeric proton at the reducing end (Fig. 3B). Peaks for a second 6-CH 2 group are also observed at 3.85 and 3.90 1 H ppm (63.5 13 C ppm) chemical shifts, which correspond to an unmodified glucose 6-CH 2 OH group (Bekiroglu et al, 2003) and this belongs to the reducing end carbohydrate residue of maltose. The ring CH protons are clearly visible between 71 and 81 ppm and these show no significant chemical shift changes when compared to free maltose, suggesting that any ubiquitylation of the ring CHOH groups is absent, or is minimal and beyond detection here.
Consistent with the NMR analysis we found that cyclomaltoheptaose lacking the reducing C1 hydroxyl and non-reducing C4 hydroxyl groups could still be ubiquitylated It should be noted that the hydroxymethyl (CH 2 -OH) group of a glycosyl unit is identical to the hydroxymethyl side-chain of the amino acid serine, a previously identified target for HOIL-1 ubiquitylation (Kelsall et al., 2019).

HOIP and Sharpin bind to amylose resin.
Since the rate of ubiquitylation of glycogen was slow, taking many hours to convert most of the ubiquitin in the assay to monoubiquitylated maltoheptaose ( Fig. EV2B), we wondered whether a mechanism(s) might be needed to accelerate the HOIL-1-catalysed ubiquitylation of glycogen in cells, if this reaction was to be of physiological significance. We therefore initially investigated the binding of HOIL-1 and other proteins to the α1:4-linked oligosaccharide amylose to evaluate their glucan-binding properties (Fig. 4A). Unlike maltose/maltodextrin binding protein (MBP), HOIL-1 did not bind to amylose-agarose resin, similar to other negative control proteins, such as glutathione S-transferase (GST). However, to our surprise, the other two components of LUBAC, Sharpin and HOIP, both bound to amylose resin, the latter behaving similarly to MBP in the pull-down assays (Fig. 4A). This suggested that one role for the interaction of HOIL-1 with HOIP and Sharpin in LUBAC may be to enable HOIP and/or Sharpin to target HOIL-1 to glycogen in cells.  4D). None of the truncated HOIP constructs bound to amylose, implying that noncontiguous regions most likely contribute to glucan binding (Fig. 4E).

Recently, DeepMind and EMBL's European Bioinformatics Institute created the
AlphaFold Protein Structure Database, making freely available structural predictions for all proteins in the human proteome (Jumper et al, 2021). The AlphaFold aligorithm predicts that the NZF domain of Sharpin contacts an interface between the PH-like domain and the UBL domain ( Fig. 4F), so assuming the structural prediction is correct, this region may form the amylose-binding interface, and removal of the NZF domain may increase accessibility to amylose to enhance binding.

Allosteric activation of HOIL-1 by Met1-linked and Lys63-linked ubiquitin oligomers.
The experiments described in the preceding section suggested a role for HOIP in recruiting HOIL-1 to glycogen. We therefore wondered whether HOIP might also facilitate the HOIL-1-catalysed ubiquitylation of glycogen in other ways. HOIL-1 contains an Npl4 Zinc Finger (NZF) domain that is reported to bind M1-Ub dimers about 10-fold more strongly than K63-Ub dimers (Sato et al, 2011). We therefore wondered whether M1-Ub oligomers, formed by the action of HOIP, might bind to the NZF domain of HOIL-1, inducing a conformational change that accelerated the HOIL-1-catalysed ubiquitylation of maltoheptaose. We found that the inclusion of M1-Ub or K63-Ub dimers ( 7C and 7D). Performing time-course assays allowed the visualisation of a dithiothreitol-sensitive tetra-ubiquitin adduct on UBE1, UBE2L3 and HOIL-1 that was prominent at early timepoints but decreased after 2 min (Fig. 7E left panel). When wild-type HOIL-1 was replaced with a catalytically inactive C460S or C460A mutant the transient adducts on UBE1 and UBE2L3 could not be discharged and persisted, while the use of the C460S mutant similarly stabilised a dithiothreitol-insensitive oxyester-linked tetraubiquitin adduct on HOIL-1 (Fig. 7E, middle and right-hand panels). These experiments are consistent with the HOIL-1-catalysed conjugation of ubiquitin oligomers to maltoheptaose en bloc. The HOIL-1-initiated polyubiquitylation of glycogen (and proteins) in cells, might therefore be catalysed in a single step, and not sequentially by HOIL-1 catalysed monoubiquitylation followed by the addition of further ubiquitin molecules to monoubiquitylated glycogen through the action of other E3 ligases.

Discussion
Ever since the discovery of lysosomal α 1:4 glucosidase deficiency (Pompe's disease (Hers, 1963)) it has been known that glycogen-like molecules are continually being transported from the cytosol to the lysosomes, where they are taken up and hydrolysed by lysosomal α 1:4 glucosidase enabling the glucose units in glycogen to be recycled. More recently, it has been recognised that this process occurs by a mechanism akin to autophagy and has therefore been termed glycophagy . In this process, it has been proposed that starchbinding domain-containing protein 1 (Stbd1), a protein localised to perinuclear or lysosomal membranes, acts as a cargo receptor to recruit glycogen to lysosomes (Jiang et al, 2010).
Autophagy requires the ubiquitylation of organelles or pathogenic bacteria, prior to their uptake into endo-lysosomes where they are destroyed (Noad et al, 2017;Polajnar et al, 2017;van Wijk et al, 2017), and the possibility that glycophagy requires the ubiquitylation of glycogen-associated proteins has been considered by others (Brewer & Gentry, 2019;Sanchez-Martin et al, 2020). However, glycophagy has generally been considered to be a device for removing normal glycogen from cells to prevent its accumulation in excess of the cell's requirements, enabling the excess glucose to be recycled for use by other cells . However, it is possible that instead, or in addition, glycophagy is a quality control mechanism that functions to remove abnormal forms of glycogen from cells, for example glycogen molecules with few α 1:6 branch points that may normally be formed adventitiously in trace amounts by errors of metabolism. If not removed rapidly, these starch-like molecules precipitate as polyglucosan deposits, which gradually accumulate in tissues until they reach levels that cause serious damage to tissue functions as shown, for example, by the fatal diseases that arise in humans and mice with deficiencies in glycogen branching enzyme, lysosomal α1:4 glucosidase or HOIL-1.
Until now, the quality control mechanism by which unbranched glucose polymers can be distinguished from normal glycogen has been obscure but, based on the results presented in this paper, we suggest the following new working hypothesis. We suggest that α1:4-linked oligosaccharides that have failed to form α1:6 branch points when they attain a certain length may then undergo HOIL-1 catalysed ubiquitylation that triggers their recognition by the glycophagy machinery, enabling their uptake into lysosomes and hydrolysis. Ubiquitylation might also enhance the solubility of unbranched glycogen preventing its precipitation until uptake into lysosomes has taken place.
HOIL-1 is associated with HOIP and Sharpin in LUBAC and, in the present study we found that both HOIP and Sharpin interact with amylose in vitro, suggesting that one of their functions may be to target HOIL-1 to glycogen in cells. We also found that small M1-Ub oligomers produced by HOIP are powerful allosteric activators of the HOIL-1-catalysed ubiquitylation of maltoheptaose, indicating a second way in which HOIP may facilitate the ubiquitylation of unbranched α1:4-linked glucosaccharides. Interestingly, these studies revealed that the allosteric binding site for these M1-Ub or K63-Ub dimers and tetramers was not the previously described M1-Ub-binding NZF domain of HOIL-1, but a region within the RBR domain.
In conclusion, our results are the first documented example of sugar ubiquitylation, introducing a new aspect to ubiquitylation research. Recent work from the Randow laboratory has described the ester-linked ubiquitylation of bacterial lipopolysaccharide by the E3 ligase RNF213 (Otten et al, 2021). However, the authors did not identify the site of ubiquitin attachment, so it is unclear whether it is the lipid or sugar moieties of lipopolysaccharide that undergo ubiquitylation. Nevertheless, their paper and the present study, indicate that ester-linked ubiquitylation of non-proteinaceous biological molecules is a tactic employed by more than one family of ubiquitin ligases to enable the ubiquitylation of amine-free substrates

Experimental Procedures
Histological examination of mouse tissues. Ni 2+ -Sepharose chromatography. Lys63-linked and linear di-and tetra-ubiquitin chains were produced and purified according to previously reported methods (Dong et al, 2011;Komander et al, 2008). All proteins were prepared in PBS, 1 mM TCEP and stored in aliquots at -80°C.
Where noted, HOIL-1 activity was stimulated by addition of chain-type specific ubiquitin dimers or tetramers at a concentration of 2 μM. In these experiments the ubiquitin chains were pre-incubated for 30 min at 30°C with all other reaction components except ATP. Reactions were initiated by adding Mg 2+ -ATP to a final concentration of 5 mM. In order to slow the reaction rate and allow better visualisation of the differences between ligase activation in the presence of different chain types, maltoheptaose concentration was dropped to 10 mM and reactions proceeded at the lower temperature of 30°C for the indicated times.

Preparation of purified Ub-maltoheptaose
Ubiquitylation reactions were carried out as described previously and terminated through the addition of 20% (v/v) trifluoroacetic acid (TFA) to a final concentration of 2%.
Ubiquitylated-maltoheptaose was separated from unconjugated free ubiquitin by reverse phase-high performance liquid chromatography (RP-HPLC) using a Dionex Ultimate 3000 System. A Thermo BioBasic column (250 x 10 mm) was equilibrated in aqueous buffer containing 20% (v/v) acetonitrile supplemented with 0.1% (v/v) TFA. A flow rate of 2.3 mL/min and an increasing gradient of acetonitrile from 20% to 60% over 60 min was utilised for sufficient separation. Separated fractions were validated using MALDI-TOF mass spectrometry before being pooled and freeze-dried.

MALDI-TOF mass spectrometry
For MALDI-TOF sample preparation a mixture containing a 1:1 ratio of 2% TFA (v/v) and 2,5 Dihydroxyacetophenone (DHAP) matrix solution (7.6mg of 2,5 DHAP in 375 μ L 100% ethanol and 12 μ L of 12 mg/mL diammonium hydrogen citrate) was added to the sample in a 1:1 ratio. 1 μ L of the solutions were spotted in duplicate onto an MTP AnchorChip 1,536 Target (Bruker Daltronics). Samples were air dried at room temperature prior to analysis. All spectra were acquired using a Rapiflex MALDI-TOF mass spectrometer (