Ubiquitination of glycogen and metabolites in cells and tissues

Abstract

Ubiquitin signalling covers a wide range of protein modifications, but its scope may still be underestimated, owing to the ubiquitination of non-proteinaceous substrates, such as sugars, lipids and nucleotides¹. The breadth of ubiquitinated non-protein substrates, their abundance and their cellular roles are currently unclear, as current ubiquitinomic and proteomic techniques do not detect non-proteinaceous modifications. Here we report non-protein ubiquitin clipping (NoPro-clipping) as a mass spectrometry-based technique that combines ubiquitin clippases with sortase labelling. Targeted and untargeted workflows unveil a wide range of ubiquitinated substrates in mammalian cells and in mouse and human tissues. We find ubiquitinated glycogen in glycogen-containing tissues in mice, with the highest abundance in liver and skeletal muscle. Ubiquitination can deliver glycogen to lysosomes and leads to decreased glycogen levels. Glycogen ubiquitination is modulated in glycogen storage diseases and is regulated by the Met1–polyubiquitin machinery. Notably, glycogen depletion in the liver during fasting coincides with increased glycogen ubiquitination, suggesting that ubiquitin is a previously unknown component of physiological glycogen catabolism. We also reveal ubiquitination of endogenous glycerol and spermine in cells and tissues. NoPro-clipping thus reveals unexpected endogenous non-proteinaceous targets of ubiquitination, broadening the role of ubiquitin from a protein modifier to a general modifier of biomolecules.

Integration of ubiquitin clippases with sortase-based labelling and mass spectrometry reveals that a wide range of non-peptide molecules are ubiquitinated in vivo, and points to a potential role for ubiquitination in physiological glycogen breakdown.

Main

Ubiquitination is a post-translational modification of proteins that regulates their function and fate, and has key roles in protein turnover by proteasomal or lysosomal degradation^2–4. During canonical ubiquitination, ubiquitin is attached to Lys residues via its C-terminal Gly-Gly (GG) motif, forming chemically stable isopeptide bonds⁵. Ubiquitination can also occur on Cys, Ser and Thr residues^6–9 forming thiol- or oxy-ester linkages that are more labile and difficult to detect by mass spectrometry workflows. Recent findings suggest the presence of non-proteinaceous ubiquitination, mostly driven by ester linkages between ubiquitin and other biomolecules, such as lipids, sugars and nucleotides^10–15. However, the evidence for non-proteinaceous ubiquitination in cells is scarce and relies on techniques such as western blotting, often diagnostically exploiting the low stability of ubiquitin ester linkages to hydroxylamine treatment^10,11. Several in vitro studies report E3 ligase reactions in which ubiquitin can be attached to glucosaccharides¹², glycans¹³, nucleic acids^14–17 and exogenous, amine-group containing small molecules^18,19.

Existing high-resolution and highly sensitive MS workflows are blind to ubiquitination of biomolecules, other than proteins, in complex mixtures. Whereas proteomic workflows excel at detecting and quantifying miniscule amounts of individual peptides in complex mixtures, protein-modified metabolites or non-proteinaceous macromolecules pose challenges for mass spectrometry techniques. Thus whether ubiquitin is attached to non-proteinaceous material under endogenous settings has not been independently verified, and the abundance and roles of ubiquitinated non-proteinaceous material in cells remains unknown.

We developed non-protein ubiquitin clipping (NoPro-clipping) to enable detection of non-proteinaceous ubiquitination in cells and tissues. NoPro-clipping uses bacterial ubiquitin clippases (Ub-clippases)²⁰ that cleave ubiquitin internally to leave its C-terminal GG motif attached to a substrate^21,22. Clippase activity is complemented with sortase labelling^23,24 to selectively peptide-label and enrich the generated GG-modified non-protein substrate. Together, both activities enable sensitive proteomic workflows to detect, identify and quantify a wide range of non-proteinaceous ubiquitination events in mammalian cells and tissues.

Detecting non-protein ubiquitination

The workflow described in Fig. 1a illustrates how we converted non-proteinaceous ubiquitinated substrates into GG-labelled small molecules for detection by metabolomic-based mass spectrometry. In brief, a cell lysate is cleared of small molecules (less than 7 kDa) using size-exclusion columns. The separated protein fraction that includes ubiquitinated non-protein species is subjected to ubiquitin clipping (Ub-clipping), using either Lb^pro* (ref. ²²) or BpJOS, a highly active ubiquitin- and NEDD8-specific Josephin-domain clippase²⁰. Clippase treatment converts the cellular ubiquitome to a GG-ome, generating small GG-modified biomolecules, which are separated from proteins in a second filtration step (removing molecules with molecular mass more than 3 kDa) before detection and characterization by microflow liquid chromatography–tandem mass spectrometry (LC–MS/MS) (Fig. 1a).

Fig. 1. NoPro-clipping can detect ubiquitinated non-protein substrates.

a, Schematic of NoPro-clipping, applicable to in vitro samples, cell lysates and tissue samples to detect diagnostic GG-modified small molecules via mass spectrometry: (1) direct NoPro-clipping of in vitro standards; and (2) NoPro-clipping of cell lysates, with or without addition of in vitro standards used as controls or for quantification. b, In vitro-generated His-tagged ubiquitinated maltoheptaose (5 µM) was incubated with 100 nM BpJOS, 10 µM Lb^pro* or without Ub-clippase (Ctrl) for the indicated time at 37 °C in 50 mM ammonium acetate (pH 6). Representative example of an experiment performed in triplicate. Uncropped blots are provided in Supplementary Fig. 9. c, Chemical structure of GG-modified maltose with the GG tag highlighted in orange. d, Representative extracted ion chromatogram (EIC) of GG-maltose following NoPro-clipping of in vitro-ubiquitinated maltose. e, Schematic depicting the sequential digestion of ubiquitinated glycogen by α-amylase and Ub-clippases, creating GG-modified glucosaccharides. f, In vitro-ubiquitinated glycogen is sensitive to α-amylase and BpJOS. Top, glycogen is stained by PAS staining. Bottom, FITC-labelled ubiquitin detected by fluorescent read-out. Representative example of an experiment performed in triplicate. Uncropped blots are provided in Supplementary Fig. 9. Arrows indicate the expected positions of glycogen and ubiquitinated glycogen. Dashed boxes indicate where small ubiquitinated glucosaccharides (<25 kDa) run on a gel. g, EICs of GG-modified glucosaccharides from in vitro-ubiquitinated glycogen processed with the workflow shown in a. Representative extracted chromatograms from one injected sample. h, Schematic of sortase-mediated peptide labelling (sortagging) of GG-substrates and LC–MS acquisition. i, Chemical structure of the depsi ClipTag peptide used to label GG-modified molecules via sortase. The depsi ester bond is highlighted by the red box. Bio, biotin. j, Representative EICs of ClipTag-glucosaccharides from in vitro-ubiquitinated glycogen processed with NoPro-clipping (workflows in a,h).

Source data

The availability of standards is essential for mass spectrometry method development. The RING-between-RING (RBR) E3 ligase HOIL-1L predominantly forms oxy-ester linkages^7,25, and can ubiquitinate a range of hexoses and pentoses in vitro^12,26. We used HOIL-1L to generate monoubiquitinated maltose and maltoheptaose to assess the stability of these analytes and optimize the NoPro-clipping workflow (Extended Data Fig. 1 and  Methods). Notably, ubiquitinated maltoheptaose (and all other standards tested so far) remained susceptible to Ub-clipping, a key pre-requisite for our method (Fig. 1b and Extended Data Fig. 1d). GG-modified saccharide species were characterized by mass spectrometry, which informed on detection limits, liquid chromatography retention times, and diagnostic MS2 fragmentation patterns (Fig. 1c,d and Extended Data Fig. 2). In each case, the GG-modified species were only detected after Ub-clipping (Extended Data Fig. 2a–f), and their MS2 spectra matched expected fragment ions (Extended Data Fig. 2g–j), with the GG fragment serving as a useful diagnostic tool. A ubiquitinated N-acetyllactosamine (LacNAc) standard was added to cell lysates, and the corresponding GG-saccharide was detected via mass spectrometry, confirming that the workflow is able detect non-protein ubiquitinated substrates in complex biological samples (Extended Data Fig. 2k).

Extended Data Fig. 1. Stability of ubiquitinated maltoheptaose and susceptibility to Ub-clipping.

a. Temperature sensitivity of ubiquitinated maltoheptaose. 5 µM His-Ub-maltoheptaose in PBS was either boiled at 95 °C for 10 min or not boiled, separated by SDS-PAGE then visualised with Coomassie. Representative example of experiments performed in triplicate. See Supplementary Fig. 9 for uncropped blot. b. Stability of ubiquitinated maltoheptaose in different buffers and pH. 5 µM His-Ub-maltoheptaose in either phosphate buffered saline (PBS, pH 7.4), 50 mM ammonium formate (pH 4, 6 or 9), 50 mM ammonium acetate (pH 4, 6 or 9), 50 mM ammonium carbonate (pH 4, 6 or 9) or 50 mM ammonium bicarbonate (pH 4, 6 or 9) was incubated for 2 h at 37 °C, separated by SDS-PAGE and visualised with Coomassie. Representative example of experiments performed in triplicate. See Supplementary Fig. 9 for uncropped blot. c. Deubiquitinase treatment of ubiquitinated maltoheptaose. 5 µM His-Ub-maltoheptaose in 4 mM DTT-containing PBS was incubated with 250 nM catalytic domains of USP10, or USP21 or kept untreated (ctr) at 37 °C for the indicated amount of time, separated by SDS-PAGE and visualised with Coomassie. Representative example of experiments performed in triplicate. See Supplementary Fig. 9 for uncropped blot. d. Ub-clipping of ubiquitinated maltoheptaose at different pH. 5 µM His-Ub-maltoheptaose in 50 mM ammonium acetate (pH 4 or 6) was incubated with either 0.1 µM BpJOS, 10 µM or 0.1 µM Lb^pro*, or kept untreated (ctr) at 37 °C for the indicated amount of time, separated by SDS-PAGE and visualised with Coomassie. Representative example of experiment performed in triplicate. See Supplementary Fig. 9 for uncropped blot.

Extended Data Fig. 2. NoPro-clipping of ubiquitinated sugars and MS/MS spectra.

a. MS1 detection of GG-maltose after ubiquitinated maltose was subjected to NoPro-clipping, ± Lb^pro*. b. Annotated MS2 spectra of GG-maltose. c–j. MS1 detection of c. GG-maltoheptaose, d. GG-N-acetyllactosamine (LacNAc), e. GG-N-acetylglucosamine (GlcNAc) and f. GG-¹³C6-glucose when in vitro ubiquitinated maltoheptaose, LacNAc, GlcNAc or ¹³C6-glucose were subjected to NoPro-clipping, ± Lb^pro* or BpJOS. MS2 spectra of, g. GG-maltoheptaose, h. GG-LacNAc, i. GG-GlcNAc and j. GG-¹³C6-glucose from c, d, e, f, respectively. The diagnostic GG ion 133.06 is highlighted in orange. k. NoPro-clipping of ubiquitinated LacNAc spiked into cell lysate. Shown are EICs of GG-LacNAc from in vitro ubiquitinated LacNAc either spiked into NoPro-clipping buffer or into HeLa cell lysate (lysed in NoPro-clipping buffer), ±Ub-clipping with either Lb^pro* or BpJOS.

Source data

HOIL-1L also ubiquitinates glycogen, a cellular storage form of glucose with a size greater than 5 MDa, in vitro¹². Glycogen, visible as a high-mass (more than 250 kDa) periodic acid Schiff (PAS) signal, was ubiquitinated with fluorescent ubiquitin, creating a fluorescent glycogen band (Extended Data Fig. 3a). α-Amylase treatment of ubiquitinated glycogen collapses the glycogen, and produces short ubiquitinated glucosaccharides visible as lower-mass (less than 25 kDa) fluorescent bands above monoubiquitin (Fig. 1e,f, boxed region in lane 5). BpJOS clipping of ubiquitinated glycogen did not affect the HMW PAS signal, but collapsed the fluorescent signal into a monoubiquitin band without higher-mass species (Fig. 1e,f, lane 9 boxed). Digestion of ubiquitinated glycogen by α-amylase, followed by BpJOS clipping (Fig. 1e,f, lane 13), generated GG-modified glucosaccharides that were no longer visible via SDS–PAGE, but were detectable as GG-modified maltotriose and maltotetraose by microflow liquid chromatography–mass spectrometry (LC–MS) (Fig. 1g and Extended Data Fig. 3b). Thus, we successfully transformed ubiquitinated non-proteinaceous macromolecules to species that were compatible with detection by mass spectrometry.

Extended Data Fig. 3. Sortase-mediated peptide labelling is compatible with Ub-clipping.

a. Glycogen from bovine liver was incubated with FITC-labelled ubiquitin, E1, UBE2L3, HOIL-1L and M1 di-ubiquitin at 30 °C for 7 h, +/− 10 mM ATP. After gel electrophoresis, fluorescent ubiquitin was visualized on a ChemiDoc and glycogen was stained via PAS stain. Depicted is one representative blot from n = 2 independent experiments. b. NoPro-clipping of in vitro ubiquitinated (FITC-labelled) glycogen from a, +/− α-Amylase and +/− BpJOS. Shown are MS1 extracted ion chromatograms (EICs) for the expected GG-glucosaccharide masses. c. Intact MS detection of His-Ub-maltose and His-UbΔGG following Ub-clipping via incubation at 37 °C for 1 h with 100 nM BpJOS. Addition of the 25 μM depsi ClipTag peptide is detected but only creates transpeptidation products when incubated for 1 h at 37 °C in the presence of 2.5 μM sortase, producing the expected ClipTag-maltose product. d. Sortase labelling of Ub-clipped in vitro standards. MS1 detection of the GG-saccharide and reciprocal ClipTag-saccharide ± 2.5 μM sortase and 25 μM depsi ClipTag. e. In vitro generated, FITC-labelled ubiquitinated glycogen was treated with α-amylase and processed via NoPro-clipping. 25 μM ClipTag peptide was added to all samples, with or without sortase (2.5 μM).

Source data

Sortase labelling increases sensitivity

In a final optimization step, we exploited the GG-remnant of ubiquitin that is conjugated to non-protein substrates. GG-labelled non-proteinaceous molecules were converted to peptidic species to enable enrichment (in this case, solid-phase peptide capture) and nanoflow peptide-centric LC–MS acquisition, markedly improving detection limits and sensitivity of the method. Peptidic conversion was facilitated by Staphylococcus aureus sortase A (hereafter sortase), a bacterial transpeptidase that can be used to specifically extend N-terminal GG remnants (Fig. 1h). During a sortase reaction, the enzyme recognizes a LPXTG motif in a supplied peptide, which it cleaves between the Thr and Gly, to form a thioester intermediate. In a second step, this LPXT intermediate is discharged onto the N-terminal Gly of a GG-labelled substrate^23,24, making the reaction an ideal companion for Ub-clipping.

We synthesized a biotinylated LPXTG motif-containing peptide (hereafter called a ClipTag peptide) with an ester bond (depsi bond) between Thr and Gly, to drive the sortase reaction to completion²⁷ (Fig. 1i). In vitro-ubiquitinated maltose, LacNAc and N-acetylglucosamine (GlcNAc) were used to confirm the compatibility of Ub-clipping with sortase labelling (Extended Data Fig. 3c,d), and extended to in vitro-ubiquitinated glycogen (Extended Data Fig. 3e). From 1/100th of the modified glycogen sample used in the original metabolomic workflow (Fig. 1g), we could detect an increased number of ClipTag-modified glucosaccharide species at greater intensity and dependent on sortase activity (Fig. 1j and Extended Data Fig. 3e). NoPro-clipping was of sufficient sensitivity to detect low-abundant ubiquitinated non-proteinaceous species.

Induced ubiquitination of glycogen

Glycogen storage is an important function in many organs, such as liver, muscle and brain²⁸, and imbalances in glycogen metabolism lead to liquid–liquid phase-separated glycogen foci and polyglucosan aggregates, a hallmark of glycogen storage diseases (GSDs)²⁹. Overexpression of glycogen synthase (GYS1 or GYS2) can induce glycogen foci in cell lines, which can be visualized as puncta by co-expression of a carbohydrate-binding CBM20 domain fused to GFP³⁰. To increase our chances of detecting ubiquitinated glycogen in cells, we engineered an induced-proximity system, in which co-localization of glycogen and HOIL-1L is induced by a FKBP/FRB chemical dimerizer system³¹ (Fig. 2a). We co-expressed FKBP–GFP–CBM20 and mCherry–FRB–HOIL-1L in human Huh7 liver cells (Fig. 2b). A diffuse GFP signal indicated a lack of glycogen foci in Huh7 cells, consistent with previous reports³⁰. However, GFP puncta appeared following treatment with the dimerizer rapamycin, which co-localized with mCherry–FRB–HOIL-1L, and immunostained for ubiquitin (Fig. 2b). GYS1 or GYS2 overexpression induced GFP-positive glycogen foci in all conditions, which, after dimerizer treatment, co-localized with HOIL-1L and stained positive for ubiquitin (Extended Data Fig. 4a,b). Ubiquitin staining was dependent on expression of both dimerizer constructs and a functional ubiquitin system (assessed via the E1 inhibitor TAK243 (ref. ³²)) (Fig. 2b and Extended Data Fig. 4a,b). Rapamycin, a regulator of energy metabolism in cells, did not induce ubiquitin co-localization at CBM20 foci in cells lacking the HOIL-1L construct (Fig. 2b and Extended Data Fig. 4b).

Fig. 2. Proximity-induced ubiquitination of glycogen via HOIL-1L.

a, Schematic of the FKBP/FRB-dimerizer system to induce proximity of HOIL-1L and glycogen via the CBM20 domain. b, Confocal fluorescence microscopy of the human liver Huh7 cell line stably expressing both FKBP–GFP–CBM20 and mCherry–FRB–HOIL-1L (rows 1–3), or FKBP–GFP–CBM20 alone (row 4). Cells were treated with 500 nM dimerizer (rapamycin) and co-treated with 1 μM TAK243 as indicated, for 4 h. Ubiquitin was labelled using the FK2 antibody and nuclei were stained with DAPI. Representative example of an experiment performed in triplicate. See Supplementary Figs. 5 and 6 for uncropped images. Similar experiments were performed using GYS1 or GYS2 overexpression to confirm CBM20-positive puncta formation (Extended Data Fig. 4). Scale bar, 20 μm. c, EICs of ClipTag-modified glucosaccharides from in vitro (FITC-labelled) ubiquitinated glycogen (left; 10 µl of ubiquitinated glycogen reaction mixture (Methods) was used as input) and from Huh7 cell lysates (right; 10 cm dish at 80% confluency), in which HOIL-1L was recruited to glycogen (b, second row) after α-amylase treatment and NoPro-clipping. Representative example of an experiment performed in triplicate. d, Annotated comparison of MS2 spectra of ClipTag-maltotriose from in vitro-generated ubiquitinated glycogen (bottom) and from Huh7 cell lysates, in which HOIL-1L was recruited to glycogen (middle). Top, chemical structure showing how fragmentation generates a-, b- and y-ions.

Source data

Extended Data Fig. 4. Formation of CBM20-positive puncta following GYS1 or GYS2 overexpression and ClipTag-maltotriose detection.

a. Confocal fluorescence microscopy of Huh7 cell lines stably expressing both FKBP-GFP-CBM20 and mCherry-FRB-HOIL-1L. GYS2 was transiently over-expressed for 24 h prior to drug treatment. Cells were then either treated with 500 nM rapamycin, 500 nM rapamycin and 1 μM TAK243, or kept untreated for 4 h. Ubiquitin was immunostained using the FK2 antibody and nuclei were stained with DAPI. Data depicted are from n = 1 independent experiment. b. Confocal fluorescence microscopy of Huh7 cell lines as described in a, additionally including Huh7 cells expressing FKBP-GFP-CBM20 alone. GYS1 was transiently over-expressed for 24 h prior to drug treatment. Cells were then either treated with 500 nM rapamycin, 500 nM rapamycin and 1 μM TAK243, or kept untreated for 4 h. Ubiquitin was immunostained using the FK2 antibody and nuclei were stained with DAPI. Data are representative of n = 3 independent experiments. See Supplementary Figs. 5, 6 for uncropped images. c. EICs of ClipTag-maltotetraose from Huh7 cells stably expressing FKBP-GFP-CBM20 and mCherry-FRB-HOIL-1L following NoPro-clipping. GYS1 was transiently over-expressed in all cells for 24 h prior to drug treatment. Cells were then either treated with 500 nM rapamycin or left untreated for 5 h, harvested and processed by NoPro-clipping with α-amylase and ± 100 nM BpJOS, followed by ClipTag-labelling ± sortase (25 μM ClipTag peptide added to all samples). Data are representative from n = 3 independent replicates. d. NoPro-clipping was performed as in c, with or without α-amylase treatment (12 U/sample at 37 °C for 2 h). Both conditions were pre-treated with 500 nM rapamycin, then Ub-clipped and sortase peptide labelled. Shown are EICs of ClipTag-maltotriose. e. NoPro-clipping was performed as in c, with or without rapamycin treatment of cells (but with α-amylase, BpJOS and Sortase treatment for all conditions). One no-rapamycin sample was spiked with 0.5 µM FITC-labelled ubiquitin and 1.15 mg/mL of glycogen from bovine liver to test for the spontaneous formation of ubiquitinated glycogen in cell lysates. Shown are EICs of ClipTag-maltotriose.

Source data

When NoPro-clipping was applied to this cellular system, we identified the same range of ClipTag-glucosaccharides that were generated from in vitro-ubiquitinated glycogen (Fig. 2c). MS2 fragmentation (exemplified for ClipTag-maltotriose) exhibited clear diagnostic a- and b-ions corresponding to the N terminus of the ClipTag peptide and several y-ions corresponding to the C terminus of the ClipTag peptide conjugated to mono- or di-hexoses (Fig. 2d). The MS2 spectrum of ClipTag-maltotriose from cells matched that of an in vitro-generated standard (Fig. 2d), and detection of the MS1 peak required the presence of both dimerizer constructs and treatment with dimerizer, α-amylase, BpJOS and sortase (Extended Data Fig. 4c,d). The addition of recombinant ubiquitin and glycogen to cell lysates did not produce a signal (Extended Data Fig. 4e). ClipTag-glucosaccharide abundance increased with prolonged dimerizer treatment (Extended Data Fig. 5a,b).

Extended Data Fig. 5. Time-dependent ubiquitination of glycogen and stable-isotope incorporation.

a. Western blot confirming the stable expression of FKBP-GFP-CBM20 and mCherry-FRB-HOIL-1L, or FKBP-GFP-CBM20 alone, in Huh7 cells. Cells were treated with 500 nM rapamycin for the indicated amount of time and were harvested for immunoblotting, with Vinculin serving as a sample processing control. See Supplementary Fig. 10 for uncropped blots. b. Time-dependent accumulation of ubiquitinated glycogen (measured as ClipTag-glucosaccharides) in the HOIL-1L recruitment system (matching Western blot in a). Varying times of 500 nM dimeriser treatment are indicated. Data represents n = 2 technical replicates. c. Schematic of isotope labelling with ¹³C6-glucose in Huh7 cells, followed by HOIL-1L* overexpression and NoPro-clipping. d. Immunoblot detection of HOIL-1L* (3xFLAG-diUb-HOIL-1L) transiently over-expressed in Huh7 cells, with Vinculin serving as a sample processing control. See Supplementary Fig. 11 for uncropped blots. e. EICs of ClipTag-maltotriose for expected ¹³C-labelling patterns (¹³C0-maltotriose, ¹³C6-maltotriose, ¹³C12-maltotriose and ¹³C18-maltotriose) from Huh7 cells incubated with ¹³C-U-glucose as depicted in c. Labelling led to a mass shift by +18 and +12 Da, corresponding to complete labelling of all three or two out of three glucosyl units, respectively. f. MS2 spectral comparison between unlabelled and ¹³C18-labelled ClipTag-maltotriose, from Huh7 cells treated as shown in c. Several y-ions possessing one or two hexoses exhibited +6 and +12 mass shifts respectively, confirming the endogenous labelling of ubiquitinated glycogen from ¹³C6-glucose.

Source data

Metabolic labelling of cells with ¹³C6-glucose and overexpression of an auto-activated HOIL-1L* construct (a fusion between HOIL-1L and Met1(M1)-linked diUb²⁶), generated ubiquitinated ¹³C6-glycogen in Huh7 cells (exemplified for ClipTag-maltotriose) with expected mass shifts in MS1 and MS2 spectra (Extended Data Fig. 5c–f).

Ub-glycogen triggers lysosomal trafficking

Lysosomal catabolism of glycogen was described more than 60 years ago³³ but a role for ubiquitination had not been associated with this process. We turned to an orthogonal inducible dimerization system using PYR and ABI tags and the dimerizer mandipropamid³⁴, to overcome potential issues arising from rapamycin-induced autophagy activation. Addition of mandipropamid induced co-localization of glycogen, HOIL-1L and ubiquitin, similar to the FKBP/FRB system (Fig. 3a,b; compare with Fig. 2b). Notably, we also observed co-localization with the lysosomal marker LAMP1, but not with the endosomal marker EEA1 (Fig. 3a,c and Extended Data Fig. 6a). Pre-treatment with bafilomycin A1 led to a further increase in the amount of fluorescence co-localization (Fig. 3c), whereas pre-treatment with TAK243 or expression of a catalytically-impaired HOIL-1L(H510A) mutant²⁶, did not lead to co-localization of CBM20 with lysosomes (Fig. 3d) or induce glycogen ubiquitination (Extended Data Fig. 6b). Results were consistent across Huh7 and Hela cell lines (Fig. 3 and Extended Data Fig. 6c,d), and were supported by live-cell imaging data, which also depicts the loss of GFP signal in the acidic lysosomal environment (Supplementary Video 1 and Supplementary Fig. 1). Of note, CBM20–GFP foci formed from either dimerizer system exhibited fluorescence recovery after photobleaching, similar to GYS1-induced foci³⁰, consistent with glycogen remaining soluble following ubiquitination (Extended Data Fig. 6e,f).

Fig. 3. Clearance of glycogen via lysosomes is ubiquitin-dependent.

a, Huh7 cells expressing ABI–GFP–CBM20 and mCherry–PYR–HOIL-1L with or without 500 nM mandipropamid (mandi) and/or 300 nM bafilomycin A1 (bafA1) treatment for 24 h. Immunostaining for GFP, ubiquitin (FK2 antibody) and LAMP1. n = 3 independent experiments. Arrowheads depict co-localization of CBM20, HOIL-1L, ubiquitin (orange) and LAMP1 (white). Scale bar, 5 μm. b, Ubiquitinated glycogen (ClipTag-maltotriose) quantified from Huh7 cells expressing the FKBP/FRB or ABI/PYR dimerizer systems with or without 500 nM dimerizer (rapamycin or mandi, respectively) for 4 h. Representative example of n = 3 independent experiments. c, Quantification of co-localization corresponding to a. n = 3 independent experiments, one-way ANOVA (Holm–Sidak correction). d, Quantification of co-localization of FKBP–GFP–CBM20, mCherry–FRB–HOIL-1L, ubiquitin and LAMP1 in Huh7 cells, with wild-type (WT) or H510A inactive mutant HOIL-1L with or without 6 µM TAK243 (30 min pre-treatment) and/or 500 nM rapamycin (4 h). n = 4 independent experiments for wild type and n = 3 for H510A. See Supplementary Fig. 2 for example uncropped images. In box plots in c,d, the centre line is the median, box limits show upper and lower quartiles, and whiskers extend to the furthest data points that lie within 1.5× the interquartile range. e, Quantification of glycogen and ubiquitinated glycogen in SKOV3 cells treated with or without 300 nM bafilomycin A1 for 22 h. Data represent one of n = 3 independent experiments. f, Schematic of the ABI/PYR dimerizer system. g, Glycogen in SKOV3 cells expressing the mandi system as depicted a (left; n = 3) or wild-type SKOV3 cells (right; n = 2), treated with or without 500 nM mandi for 4 h. Independent replicates; paired two-sided t-test. h, Schematic of glycogen branching basally and when GBE1 is disrupted. i, Quantification of glycogen and ubiquitinated glycogen in wild-type and GBE1^−/− SKOV3 cells. n = 3 independent experiments. j, Quantification of ubiquitinated glycogen in GBE1^−/− SKOV3 cells with or without 48 h small interfering RNA (siRNA)-mediated silencing of HOIP (also knbown as RNF31) or negative control siRNA. n = 3 independent experiments. k, Quantification of ubiquitinated glycogen in wild-type, OTULIN-knockout (KO) or PYGL-knockout SKOV3 cells. Data represent one of n = 3 independent experiments. NS, not significant (P ≥ 0.05).

Source data

Extended Data Fig. 6. Glycogen localisation to lysosomes is ubiquitin dependent.

a. Confocal microscopy of Huh7 cells expressing the FKBP-FRB dimeriser system with 4 h rapamycin treatment. GFP-CBM20, ubiquitin and HOIL-1L co-localise but no co-localisation with the endosomal marker EEA1 was observed. See Supplementary Fig. 8 for uncropped images. b. Ubiquitinated glycogen (via measurement of ClipTag-maltotriose) quantified from Huh7 cell lines stably expressing the FKBP/FRB dimeriser systems where HOIL-1L is either wild-type (WT) or possesses the H510A substitution. 500 nM rapamycin was added for 4 h prior. Representative example of n = 3 independent biological replicates with n = 3 technical replicates. c. Quantified co-localisation of FKBP-GFP-CBM20, mCherry-FRB-HOIL-1L, ubiquitin and LAMP1 in HeLa cells. Where specified, cells were incubated with 300 nM bafilomycin A1 (bafA1) and 500 nM rapamycin (rapa) 24 h prior to fixation. Data represents n = 3 independent replicates with n = 3 technical replicates each, with one-way ANOVA Holm-Sidak testing, ns denotes P > 0.05. Centre line = median; box limits = upper and lower quartiles; whiskers = connect box to the furthest data point that lies within the 1.5x interquartile range. See Supplementary Fig. 3 for example uncropped images used for quantification. d. Quantification of co-localisation of FKBP-GFP-CBM20, mCherry-FRB-HOIL-1L, ubiquitin and LAMP1 in HeLa cells. Where specified, cells were pre-treated for 30 min with 6 µM of TAK243 (TAK) and 500 nM rapamycin (rapa) was then added for 4 h. Data represents n = 3 independent replicates with n = 3 technical replicates each. Plot specifications are equivalent to c. e. Fluorescence recovery after photobleaching (FRAP) of FKBP-GFP-CBM20 foci induced either with 24 h transient overexpression of GYS1 or with 24 h addition of 500 nM rapamycin (= recruiting mCherry-FRB-HOIL-1L to glycogen) in Huh7 cells. The same recovery was observed in SKOV3 cells expressing ABI-GFP-CBM20 and mCherry-PYR-HOIL-1L for foci observed basally (control) or following 24 h of 500 nM mandipropamid (= dimeriser) treatment, indicating they are not insoluble aggregates. Lighter banding indicated 95% confidence intervals. f. Example for the fluorescence recovery of a rapamycin-induced FKBP-GFP-CBM20 foci in Huh7 cells (see e). A Fire Lookup Table (LUT) was applied to show levels of fluorescence. g. Glycogen quantification of HepG2 (liver cancer), MeWo (melanoma), and SKOV3 (ovarian cancer) cell lines. Measurements were performed on 50,000 cells and a ‘no glucoamylase’ condition was used to measure residual glucose amounts. Data are from n = 3 technical replicates. h. Quantification of co-localisation of ABI-GFP-CBM20, mCherry-PYR-HOIL-1L, ubiquitin and LAMP1 in SKOV3 cells. Where specified, cells were pre-treated for 30 min with 6 µM of TAK243 (TAK) and 500 nM mandipropamid (mandi) was then added for 4 h. Data represent n = 3 independent biological replicates with n = 3 technical replicates each. Plot specifications are equivalent to c. See Supplementary Fig. 4 for example uncropped images used for quantification. i. Ubiquitinated glycogen (via measurement of ClipTag-maltotriose), quantified from SKOV3 cells expressing ABI-GFP-CBM20 and mCherry-PYR-HOIL-1L +/− 4 h mandi treatment. Representative example of n = 3 independent biological replicates with n = 3 technical replicates. j. Western blot confirming the CRISPR-Cas9 mediated genetic disruption of Glycogen Branching Enzyme (GBE1) in SKOV3 cells, with Vinculin serving as a sample processing control. Depicted one independent experiment. See Supplementary Fig. 12 for uncropped blots k. Transmission electron microscopy of WT and GBE1^−/− SKOV3 cells. Dotted box indicates a polyglucosan body, with an inset corresponding to this region. At 25,000x magnification, glycogen granules of 20–30 nm are observed in WT but absent from GBE1^−/− SKOV3 cells. In contrast, smaller 2–5 nm granules and large polyglucosan deposits are only observed in GBE1^−/− SKOV3 cells. Data are representative of n = 2 independent experiments.

Source data

Liver cancer cell lines, including Huh7 and HepG2, have low glycogen content under standard (glucose-replete) growing conditions^35,36 and have no detectable ubiquitinated glycogen (Extended Data Figs. 4c and 6g), yet expression of auto-activated HOIL-1L* in Huh7 was sufficient to detect ubiquitinated glycogen by NoPro-clipping (Extended Data Fig. 5e). Other cancer cell lines, such as melanoma-derived MeWo and ovarian cancer-derived SKOV3 cells, were previously reported to have high glycogen content³⁷. We confirmed the presence of substantial amounts of glycogen in SKOV3 cells (Extended Data Fig. 6g) and detected endogenous ubiquitinated glycogen under standard growth conditions (glucose- and serum-containing media) (Fig. 3e). Glycogen and ubiquitinated glycogen increased threefold following treatment with the autophagy inhibitor bafilomycin A1 (Fig. 3e).

We next introduced the PYR/ABI system into SKOV3 cells, and induced co-localization of CBM20, HOIL-1L, ubiquitin and LAMP1 following addition of mandi (Fig. 3f and Extended Data Fig. 6h). Dimerization induced substantial ubiquitinated glycogen and coincided with a reduction in detectable glycogen of about 25% (Fig. 3g and Extended Data Fig. 6i).

Altered Ub-glycogen in GSDs

We next investigated the effect of glycogen structure and glycogen ubiquitination in GSD conditions. GBE1 mediates 1,6-glucoside linkage formation and creates the highly branched architecture of glycogen (Fig. 3h). When GBE1 is disrupted, glycogen is mostly unbranched, forming polyglucosan bodies and causing human pathologies such as Andersen disease (also known as GSD IV) and adult polyglucosan body disease. Knockout of GBE1 in SKOV3 cells led to a substantial (11-fold) reduction in detectable glycogen (Fig. 3h,i), as previously observed in settings of GBE1 loss, but differs from clinical presentation of GSD IV patients, who show residual (5–20%) GBE1 activity and higher detectable glycogen levels^38,39. Loss of GBE1 in SKOV3 cells also induces the appearance of polyglucosan bodies (Extended Data Fig. 6j,k), a hallmark of GSD IV. Notably, we observed a 27-fold increase in ubiquitinated glycogen in GBE1^−/− SKOV3 cells (Fig. 3i), corresponding to a 300-fold increase when normalized to detectable glycogen.

HOIL-1L is a component of the linear ubiquitin chain assembly complex (LUBAC), which is responsible for regulating M1-linked polyubiquitin in cells. Loss of LUBAC components in mice or humans leads to inflammation^40,41 but also causes amylopectinosis^42–44, a subtype of GSDs associated with defective glycogen structure. Indeed, recent data suggest that LUBAC regulates metabolism^45,46. Knockdown of the LUBAC component HOIP in GBE1^−/− SKOV3 cells reduced glycogen ubiquitination by around 60% (normalized to detectable glycogen) (compare Fig. 3j with Extended Data Fig. 7a,b), suggesting that LUBAC is involved in modifying polyglucosan.

Extended Data Fig. 7. Genetic disruption in cells and ubiquitinated glycogen in mouse tissue.

a. Western blot confirming the siRNA-mediated gene silencing of the LUBAC component HOIP in SKOV3 GBE1−/− cells, with Actin serving as a sample loading control. See Supplementary Fig. 12 for uncropped blots. b. Glycogen level quantified from SKOV3 cells that cannot express GBE1 (GBE1−/−), with or without 48 h siRNA-mediated gene silencing of the LUBAC component HOIP. Data are presented as technical replicate averages of n = 2 independent biological replicates. c. Western blot confirming the CRISPR-Cas9 mediated genetic disruption of the M1-specific deubiquitinase OTULIN and PYGL isoform of glycogen phosphorylase in SKOV3 cells, with Vinculin serving as a sample processing control. See Supplementary Fig. 13 for uncropped blots. d. Glycogen level quantified from wild type (WT) SKOV3 cells and when the M1-specific deubiquitinase OTULIN or one of the main glycogen phosphorylase isoforms expressed in SKOV3 (PYGL) are disrupted. Data are presented as n = 3 technical replicates. e. 5 mg/mL FITC-Ub (S20C) bovine glycogen was treated with 250 nM USP10 catalytic domain (AA 368-798), with OTULIN (full-length) or kept untreated (ctr) at 37 °C for the indicated amount of time, separated by SDS-PAGE, imaged (532/28 emission filter) then PAS stained for glycogen. Depicted is one representative blot from n = 2 independent experiments. See Supplementary Fig. 13 for uncropped blots. f. ClipTag-maltotriose from mouse liver could be modulated by addition of exogenous, in vitro generated ubiquitinated maltoheptaose. Mouse liver under standard feeding conditions was spiked with increasing amounts of in vitro generated His-tagged ubiquitinated maltoheptaose, digested with α-amylase and processed via NoPro-clipping and sortase labelling. Shown are EICs of ClipTag-maltotriose. g. ClipTag-maltotriose MS2 spectra acquired from mouse liver samples match the in vitro standard of ubiquitinated glycogen. Shown is an annotated MS2 spectral comparison between ClipTag-maltotriose from mouse liver under standard feeding conditions and ClipTag-maltotriose from in vitro ubiquitinated (FITC-labelled) glycogen following NoPro-clipping. h. EICs of ClipTag-maltotriose from mouse quadricep muscle after NoPro-clipping and ClipTag-labelling. Muscle samples were either directly measured, spiked with 2 nmol recombinant native ubiquitin and 4.6 mg glycogen from bovine liver prior to NoPro-clipping or spiked with 0.8 nmol His-tagged ubiquitinated maltoheptaose. One processing blank (i.e. no tissue lysate) was also spiked with 2 nmol recombinant native ubiquitin and 4.6 mg glycogen from bovine liver prior to NoPro-clipping. Importantly, the signal of ClipTag-maltotriose was not modulated by adding exogenous free ubiquitin and unmodified glycogen. i. EICs of ClipTag-maltotriose from mouse liver under standard feeding conditions processed via NoPro-clipping and ClipTag-labelling, ± BpJOS, ± ClipTag and ± α-amylase. Note that the intensity values for the no-α-amylase sample were divided by 4 to compare it to the other 4 samples which originate from the same input but were split into 4 during processing. The ClipTag-maltotriose species was only detectable with ClipTag and BpJOS treatment, while a low ClipTag-maltotriose signal was also present without exogenous α-amylase, suggestive of active glycogen catabolic enzymes present in liver lysate. j. Schematic of the mouse feeding schedule for the fasting experiment in Fig. 4b. k. Glycogen and ubiquitinated glycogen (measured as ClipTag-maltotriose) per total protein in the livers of an independent cohort of mice (as compared to Fig. 4b) that were subjected to a similar fed, fasted, and 24 hr fasted/refeed regime, with n = 5 mice per condition. Centre line = median; box limits = upper and lower quartiles; whiskers = connect box to the furthest data point that lies within the 1.5x interquartile range.

Source data

Knockout of OTULIN, the LUBAC-antagonizing M1 deubiquitinase, in wild-type SKOV3 cells, led to a 16-fold increase in glycogen ubiquitination, whereas glycogen levels remained unchanged (Fig. 3k and Extended Data Fig. 7c,d). The mechanism of OTULIN antagonism of LUBAC⁴⁷ limits its activity to cleaving M1 chains, and the enzyme is incapable of hydrolyzing the bond between ubiquitin and glucose or glycogen in vitro (Extended Data Fig. 7e), suggesting a more complex mechanism of regulation. We note that although the studied scenarios for OTULIN and HOIP relate to distinct glycogen structures, as indicated, for example, by electron microscopy (Extended Data Fig. 6k), these data directly implicate LUBAC, OTULIN and M1-linked ubiquitination in glycogen regulation.

Deletion of PYGL—a highly expressed isoform of glycogen phosphorylase that mediates the cytosolic hydrolysis of 1,4-glucoside linkages—in SKOV3 cells did not affect glycogen or ubiquitinated glycogen (Fig. 3k and Extended Data Fig. 7c,d). SKOV3 transcriptionally expresses PYGM and may still possess active glycogenolysis.

The fasting response of Ub-glycogen

The existence of endogenous ubiquitinated glycogen under basal conditions in a human cell line motivated further physiological studies. We collected organs from wild-type male C57BL/6 mice, preserving and measuring glycogen content, and processed these samples using NoPro-clipping. Glycogen levels across brain, lung, heart, liver and skeletal muscle were within expected ranges (Fig. 4a and  Methods). Notably, we detected endogenous ubiquitinated glycogen in all organs. In particular, a large quantity of ubiquitinated glycogen (ClipTag-maltotriose) was observed in the liver (Fig. 4a), a finding confirmed by MS2 spectral comparison to an in vitro standard (Extended Data Fig. 7f–i).

Fig. 4. Ubiquitinated glycogen in mammalian tissue is dynamically regulated during fasting.

a, Analysis of glycogen and ubiquitinated glycogen in tissue samples of n = 5 male, wild-type, C57BL/6 mice between 8 and 11 weeks of age. Top, amount of glycogen as a proportion of total protein from indicated mouse tissues. Bottom, ubiquitinated glycogen abundance, measured as ClipTag-maltotriose, as a proportion of total protein from indicated mouse tissues. See Extended Data Fig. 7 for controls. Quad, quadricep muscle. b, Glycogen (top) and ubiquitinated glycogen (middle, measured as ClipTag-maltotriose), and the ratio of Ub-glycogen to glycogen (bottom), in liver samples from mice that were fed (control), fasted, or fasted then re-fed, for the indicated times (see Extended Data Fig. 7j). n = 5 mice were analysed per condition; each data point represents a single mouse. The centre line is the median, box limits show upper and lower quartiles, and whiskers extend to the furthest data points that lie within 1.5× the interquartile range from the box edges. A second mouse cohort is shown in Extended Data Fig. 7k. c, Absolute quantification of ClipTag-maltose, ClipTag-maltotriose and ClipTag-maltotetraose from two mouse livers using in vitro-generated isotope-labelled ubiquitinated maltose, maltotriose and maltotetraose standards (see Extended Data Fig. 8a–d). Mouse m1 was part of the 6 h fasted cohort and was selected because it yielded the highest ClipTag-maltotriose signal; mouse m2 was part of the 24 h fasted cohort and yielded one of the lowest ClipTag-maltotriose signals. d. EICs of ClipTag-maltotriose from three human skeletal muscle biopsies (see Extended Data Fig. 8e and Methods).

Source data

The liver has a central role in glycogen metabolism, as it regulates circulating blood glucose concentrations, and releases glucose from glycogen stores to counter nutrient stress, such as fasting. We performed a fasting study with mice to quantify glycogen and ubiquitinated glycogen in livers under fed, fasted and fasting–re-feeding conditions (Fig. 4b and Extended Data Fig. 7j). As expected, liver glycogen decreased by 90% after 6 h of fasting and was at the detection limit after 24 h or 48 h of fasting (Fig. 4b). By contrast, the amount of ubiquitinated glycogen increased after 6 h of fasting compared with fed mice, before decreasing to barely detectable levels with longer fasting. There was an eightfold increase in glycogen ubiquitination (relative to detectable glycogen) within 6 h of fasting, indicating an active ubiquitination process that occurs during glycogen depletion (Fig. 4b).

Refeeding animals for 3 h after a 24 h fasting period resulted in recovery of glycogen levels, but did not recover ubiquitinated glycogen levels, indicating that freshly synthesized glycogen is not ubiquitinated. Ubiquitinated glycogen abundance recovered only after longer refeeding times (24 h) (Fig. 4b). An independent, similarly treated cohort of mice exhibited the same dynamic responses (Extended Data Fig. 7k).

To quantify ubiquitinated glycogen in mouse liver, we generated a new set of standards comprising ¹³C-ubiquitinated maltose, maltotriose and maltotetraose, in which uniformly ¹³C-labelled ubiquitin was attached to unlabelled glucosaccharides, leading to a mass shift of +4 in ClipTag-glucosaccharide species, as they contained 4 ¹³C atoms from the ubiquitin-derived GG motif (Extended Data Fig. 8a). These ubiquitinated standards behaved as expected (Extended Data Fig. 8b–d). We spiked known amounts of these standards into α-amylase-treated liver lysates from mice that were fasted for 6 h (m1) or 24 h (m2) (Fig. 4c). Unlabelled ClipTag-glucosaccharides were quantified against a standard curve of the ¹³C-labelled species. m1, which had the highest ubiquitinated glycogen concentration after a 6 h fast, was estimated to contain more than 1.35 pmol (per mg of protein) ubiquitin attached to glycogen (sum of ClipTag-labelled maltose, maltotriose and maltotetraose). Total ubiquitin levels were 103 and 109 pmol (per mg of protein) for fed and 6 h fasted conditions, respectively (Extended Data Fig. 8e). Thus, more than 1% of total ubiquitin is attached to glycogen after 6 h of fasting, which is similar to the total amount of ubiquitin in Lys11 linkages in these samples (Extended Data Fig. 8f, 1.1 pmol per mg of protein). Prolonged fasting in m2 resulted in a decrease to 72 fmol of ubiquitin attached to glycogen per mg total protein, indicating a dynamic Ub-glycogen system.

Extended Data Fig. 8. Absolute quantification of ClipTag-glucosaccharides in mouse liver.

a. Schematic for MS-based absolute quantification of ClipTag-glucosaccharide species. b. Isotopic envelope for detected ClipTag-glucosaccharide species. In vitro generated and α-amylase digested FITC-labelled ¹²C-ubiquitinated glycogen and ¹³C-ubiquitinated -maltose, -maltotriose, and -maltotetraose were processed via NoPro-clipping. MS1 spectra of the expected masses for ClipTag-maltose, -maltotriose, and -maltotetraose confirm the expected ¹³C shift of 4 carbon atoms (¹³C4) from the C-terminal GG of ubiquitin. c. Annotated MS2 spectral comparison between ¹³C4-labelled ClipTag-maltotriose from ¹³C-ubiquitinated maltotriose (top) and unlabelled ClipTag-maltotriose from FITC-labelled ubiquitinated ¹²C-glycogen (bottom), both in vitro catalysed and processed via NoPro-clipping. Mass range 0 – 900 m/z is shown. d. Same comparison as in c but for ClipTag-maltotetraose. Mass range 0–850 m/z is shown. e. Absolute quantification of total ubiquitin from mice livers under fed and 6 h fasted conditions (n = 5 mice per condition). The mean signal of two unmodified peptides summed with the overlapping K63 peptide is reported. f. Absolute quantification of ubiquitin chain linkages from mice livers under fed and 6 h fasted conditions. ¹³C¹⁵N standards for all linkage peptides (M1, K6, K11, K27, K29, K33, K48, and K48) and three unmodified ubiquitin peptides (to determine the total amount of ubiquitin present) were spiked into samples at 5 fmol and used to determine absolute quantities via parallel reaction monitoring LC-MS. Data represent n = 5 mice from both conditions. g. Annotated MS2 spectral comparison between ClipTag-maltotriose from human skeletal muscle and ClipTag-maltotriose from in vitro generated FITC-labelled ubiquitinated glycogen.

Source data

Finally, ubiquitinated glycogen is present in human skeletal muscle biopsies, as evidenced by the detection of the ClipTag-maltotriose species (Fig. 4d and Extended Data Fig. 8g). This suggests that patient tissue banks could provide a useful resource to analyse changes in glycogen ubiquitination by NoPro-clipping.

Ub-metabolites in cells and tissue

Above, we describe targeted mass spectrometry approaches to study ubiquitinated saccharides and ubiquitinated glycogen in a cellular context. NoPro-clipping could in principle also be used in untargeted, exploratory workflows, to identify other ubiquitinated non-proteinaceous species. To this end, we over-expressed HOIL-1L* in Huh7 cells as above and pre-treated lysates with α-amylase and a deglycosylation mix prior to processing via NoPro-clipping. Our analysis identified 94,027 extracted MS1 m/z features after pre-processing. These were subsequently filtered to retain features that (1) were significantly increased after BpJOS treatment (compared with no Ub-clipping); (2) depended on the addition of ClipTag peptide to the sortase reaction (compared with the omission of ClipTag peptide); and (3) were enriched in cell lysates relative to identically processed blanks (Fig. 5a). We found that 185 m/z features met these criteria and included ClipTag-labelled maltotriose, tetraose and pentaose species (Fig. 5a and Extended Data Fig. 9a). The presence of ubiquitinated glycogen (that is, detection of ClipTag-glucosaccharides) in our sample is consistent with the targeted workflows discussed above (Figs. 2–4), and indicated that untargeted NoPro-clipping could identify non-protein ubiquitinated substrates.

Fig. 5. Untargeted NoPro-clipping identifies ubiquitinated glycerol and ubiquitinated spermine.

a, MS1 m/z features detected in Huh7 cells over-expressing HOIL-1L*. m/z features highlighted in orange are significantly (two-sided moderated t-test with Benjamini–Hochberg adjustment, adjusted P value < 0.05) increased (>2log₂) after BpJOS treatment, sortase labelling and are absent from processing blanks. Lysates were deglycosylated and treated with α-amylase. The presence of ClipTag-glucosaccharides (purple) was dependent on HOIL-1L* expression, and corresponded to species characterized in Figs. 2–4. b, Untargeted m/z features identified in a were retained if they were also elevated (>2 log₂) in wild-type Huh7 cells (Extended Data Fig. 9b). Orange m/z features contained ClipTag a2, b2 and b3 fragment ions in their MS2 spectra (Fig. 2d), and were confirmed as M⁰ parental species. The green feature corresponds to a confirmed identified peptide species (Extended Data Fig. 9c). c, Putative structure and chemical formula of ubiquitinated glycerol and ubiquitinated spermine. d, Top, annotated chemical structure of ClipTag-glycerol. Bottom, annotated MS2 spectral comparison between ClipTag-glycerol from endogenous ubiquitinated glycerol (green background) and from in vitro-generated ubiquitinated glycerol (grey background). e, EICs of ClipTag-¹²C-glycerol and ClipTag-¹³C3-glycerol from Huh7 cells cultured in unlabelled or ¹³C-uniformly labelled glucose (¹³C6-glucose) followed by HOIL-1L* overexpression and processing by NoPro-clipping, as depicted in Extended Data Fig. 5c. Also see Extended Data Figs. 9g–i and 10a–c. f, Top, annotated chemical structure of ClipTag-spermine. Bottom, annotated MS2 spectral comparison between ClipTag-spermine from endogenous ubiquitinated spermine (green background) and from in vitro-generated ubiquitinated spermine (grey background). Also see Extended Data Fig. 10d–h and Methods). g, Quantification of ClipTag-spermine from Huh7 cells treated for 3 days with DMSO or DMSO plus 500 µM difluoromethylornithine (DMFO), an inhibitor of ornithine decarboxylase that participates in the biosynthesis of spermine. Data represent n = 3 technical replicates.

Source data

Extended Data Fig. 9. Untargeted NoPro-clipping validation and identification of Ub-glycerol.

a. Integrated peak area for ClipTag-glucosaccharide species (Fig. 5a, purple features, compare Fig. 2c) in the untargeted NoPro-clipping workflow. Data are from n = 3 independent samples and used for validation of the method. Interestingly, HOIL-1L* overexpression is necessary and sufficient to detect ubiquitinated glucosaccharide species in Huh7 cells. b. Untargeted NoPro-clipping performed on untreated Huh7 cell lines. Following pre-filtering of all MS1 peak groups (see Methods), the remaining m/z features were filtered according to the criteria used in Fig. 5a. Orange data points represent m/z features that were significantly (two-sided moderated t test with Benjamini-Hochberg adjustment, adjusted p value < 0.05) increased (>2 log₂) upon BpJOS treatment, sortase labelling and not present in cell-free processing blanks. Lysates were deglycosylated and α-amylase treated (see Methods). c. Annotated MS2 spectrum of a GG-containing peptide (green feature in Fig. 5b) identified via the untargeted NoPro-clipping workflow. Peptidic species were deprioritised. d,e. Integrated peak area for m/z features 567.784 (Δmass of 92.0458) and 415.581 (Δmass of 202.2159) identified as a putative ubiquitin substrate, across conditions analysed in the untargeted NoPro-clipping workflow. Data are from n = 3 independent samples. f. Structure of ubiquitinated glycerol and ClipTag-glycerol. The Δmass of 92.0458 calculated for the endogenously identified m/z feature 567.784 (z = 2) corresponds to glycerol (the exact m/z of ClipTag-glycerol z = 2 is 567.785). g. In vitro generated ubiquitinated glycerol was processed via NoPro-clipping. EICs of the expected m/z for ClipTag-glycerol are presented +/− BpJOS, +/− ClipTag. Matching retention times between the endogenous m/z 567.784 feature, and in vitro-generated ClipTag-glycerol further confirmed the presence of ubiquitinated glycerol in cells. h. Modulation of the endogenous ClipTag-glycerol peak by exogenous addition of increasing quantities of in vitro ubiquitinated glycerol independently NoPro-clipped. Numbers indicate the volumetric ratio of NoPro-clipped in vitro ubiquitinated glycerol to Huh7 cell lysate processed identically (i.e. 0 = no in vitro addition, 5 = 5:1 addition of in vitro to cell lysate). i. Isotopic envelope of ClipTag-glycerol under unlabelled and ¹³C6-glucose labelling of Huh7 cells, followed by HOIL-1L* overexpression (for schematic see Extended Data Fig. 5c). j. Annotated MS2 spectral comparison between (top) uniformly ¹³C-labelled ClipTag-glycerol, which emerged in ¹³C6-glucose-labelled Huh7 cells and (bottom) unlabelled ClipTag-glycerol from unlabelled Huh7 cells, both under HOIL-1L* overexpression. A Δmass of 3 Da for y-ions corresponds to metabolic labelling of three carbon atoms.

Source data

We generated an analogous dataset for Huh7 cells not over-expressing HOIL-1L* (Extended Data Fig. 9b). To identify putative non-proteinaceous ubiquitinated species, m/z features present in both unmodified and HOIL-1L* overexpression datasets that met the three criteria (described above) were selected for further MS2 fragmentation (Fig. 5b). Six m/z features contained diagnostic a and b ions from the ClipTag N-terminal region and were identified as M⁰ parental species on the basis of their MS1 spectra. The ClipTag peptide mass was subtracted from the neutral mass for each species, to arrive at the mass of the ubiquitinated substrate (termed Δmass;  Methods). MS2 spectral annotation identified several peptidic species (for example, Extended Data Fig. 9c) which were deprioritized.

Two notable m/z features had Δmasses of 92.0458 and 202.2159 (Fig. 5b). Both species were significantly increased in abundance in Huh7 cell lysates that were processed with BpJOS and sortase-labelled, and whereas the latter species did not significantly change after HOIL-1L* expression, the former species was decreased by two orders of magnitude following HOIL-1L* expression (Extended Data Fig. 9d,e). Metabolite database searching returned putative matches to glycerol (Δmass = 92.05 Da) and spermine (Δmass = 202.22) (Fig. 5c).

An in vitro standard of ubiquitinated glycerol was generated (Methods) that, after NoPro-clipping, produced an identical m/z feature with the same retention time as the endogenous species (Extended Data Fig. 9f,g) and matching MS2 spectra (Fig. 5d). In vitro-ubiquitinated glycerol, when spiked into lysates, increased the ClipTag-glycerol MS1 peak intensity (Extended Data Fig. 9h). Notably, ubiquitinated glycerol could be metabolically labelled. In HOIL-1L*-expressing cells cultured in the presence of ¹³C6-glucose (see Extended Data Fig. 5), a fraction of the ClipTag-glycerol mass was increased by 3 Da (Fig. 5e and Extended Data Fig. 9i,j). Using the same quantification strategy as for ubiquitinated glycogen (Extended Data Fig. 8), we detected and quantified ubiquitinated glycerol to be 35 fmol per mg of protein in untreated Huh7 cells (Extended Data Fig. 10a,b), and were also able to detect ubiquitinated glycerol in mouse liver lysates (Extended Data Fig. 10c). We thus conclude that the ubiquitinated glycerol that we detected is a bona fide endogenous species generated by cellular glucose metabolism.

Extended Data Fig. 10. Identification and quantification of Ub-glycerol and Ub-spermine.

a. Annotated MS2 spectral comparison between ¹³C4-labelled ClipTag-glycerol from in vitro generated ¹³C-ubiquitinated glycerol and unlabelled ClipTag-glycerol endogenously present in Huh7 cells. b. Absolute quantification of ClipTag-glycerol present in Huh7 cells using in vitro generated isotope-labelled ubiquitinated glycerol. c. Quantification of ClipTag-glycerol detected in mouse liver (n = 5 mice). d. Structure of ubiquitinated spermine and ClipTag-spermine. The Δmass of 202.2159 calculated for the endogenously identified m/z feature 415.581 (z = 3) corresponds to spermine (the exact m/z of ClipTag-spermine z = 3 is 415.582). e. EICs of an in vitro ubiquitinated spermine standard (see Methods) processed via NoPro-clipping, with and without BpJOS treatment and with and without sortase labelling. Matching retention times between the endogenous m/z 415.582 feature, and in vitro-generated ClipTag-spermine further confirmed the presence of ubiquitinated spermine in cells. f. Annotated MS2 spectral comparison between ¹³C4-labelled ClipTag-spermine from in vitro generated ¹³C-ubiquitinated spermine and unlabelled ClipTag-spermine endogenously present in Huh7 cells. g. Absolute quantification of ClipTag-spermine present in Huh7 cells using in vitro generated isotope-labelled ubiquitinated spermine. h. Quantification of ClipTag-spermine detected in mouse liver (n = 5 mice).

Source data

Similarly, we confirmed the presence of ubiquitinated spermine via NoPro-clipping of in vitro-generated ubiquitinated spermine (Extended Data Fig. 10d,e and Methods). MS2 spectra of in vitro-ubiquitinated spermine and the endogenous cellular species (corresponding to a Δmass of 202.2159) were analogous (Fig. 5f). As for ubiquitinated glycerol, we detected and quantified ubiquitinated spermine at 16 fmol per mg of total protein in unmodified Huh7 cells, and readily detected the species in mouse liver lysates (Extended Data Fig. 10f–h). Spermine can be sourced exogenously by cells but is also synthesized via the polyamine biosynthetic pathway⁴⁸. Cells incubated with difluoromethylornithine (DFMO)—an inhibitor of ornithine decarboxylase—led to a reduction in ubiquitinated spermine compared with untreated cells (Fig. 5g), revealing spermine as an endogenous substrate of ubiquitination in normal tissue culture cells.

Discussion

Here we present a new layer of ubiquitin biology, which we access via a new mass spectrometry workflow. We reveal a high abundance and diversity of ubiquitinated non-proteinaceous material. Our method utilizes Ub-clippases^20,22 and sortase²³ to enable selective peptide labelling of clippase products and relies on the breakdown of ubiquitinated macromolecules into small molecule, metabolite-like species. We focus on glucosaccharides, which are simplified in lysates by pre-treatment with α-amylase (for glycogen breakdown), however, NoPro-clipping could also be used with different catabolic enzymes to study ubiquitination of diverse macromolecules, such as polynucleotides, glycans or lipids. Our method further requires Ub-clippases to generate GG remnants on non-protein molecules. Although we have not encountered a ubiquitinated standard that is resistant to BpJOS or Lb^pro*, it is possible that not all macromolecules are equally susceptible to Ub-clipping, and that we may miss species, especially with untargeted workflows. BpJOS targets both ubiquitin and NEDD8 to generate identical GG remnants²⁰; although there is currently no evidence of a role for NEDD8 in non-proteinaceous modification, future studies will be needed to distinguish between these modifications. Finally, sortase labelling was a key step in transforming GG-modified biomolecules into peptidic species that were amenable to high-sensitivity nanoflow mass spectrometry workflows and produced information-rich MS2 fragmentation. However, species annotation is complicated by the lack of software and databases to search for peptide–metabolite hybrids, and we relied on in vitro standards to unambiguously identify ubiquitinated metabolites.

We found high amounts of ubiquitin attached to glucose within glycogen, which we detected by NoPro-clipping, in glycogen-containing tissues analysed from mice. Our data suggest that during fasting, whereas glycogen is actively depleted, the amount of ubiquitinated glycogen increases, indicating that ubiquitin responds to metabolic cues and partakes in glycogen breakdown (Fig. 4b). Indeed, under 6 h fasting conditions, around 1% of all ubiquitin in mouse liver was attached to glycogen, making this the fourth most abundant ubiquitin conjugate after K48 and K63 ubiquitin chain linkages (Extended Data Fig. 8), and histone ubiquitination. Together, this ubiquitin-glucose linkage is both highly abundant and dynamic.

Functionally, our data indicate that ubiquitination can mediate trafficking of glycogen to lysosomes for degradation. The common route for ubiquitin-dependent lysosomal delivery involves the autophagy machinery and our preliminary data may indeed support a mechanism of ubiquitin-dependent glycophagy, which needs more detailed exploration. A ubiquitin-independent glycophagy pathway has been reported⁴⁹, and whether or how these pathways intersect will require further study. Alternatively or in addition, the appearance of ubiquitinated glycogen foci may be conceptually similar to RNA stress granules or RNA processing bodies, and may enable longer term storage of glycogen particles for delayed catabolism.

Our data suggest that ubiquitin is a previously unrecognized component of glycogen metabolism, which presents intriguing possibilities for physiological modulation of cellular glycogen. To this end, we demonstrated that induced glycogen ubiquitination coincides with decreased glycogen in cell lines (by inducing co-localization of a capable E3 ligase, such as HOIL-1L, to glycogen) (Figs. 2 and 3). Loss of, or mutations in, HOIL-1L (and other components of LUBAC, such as HOIP, SHARPIN or OTULIN) causes inflammation in animal models and humans, but also leads to an amylopectinosis phenotype^42–44. We show here regulation of ubiquitinated glycogen when HOIP or OTULIN are defective, corroborating the notion that LUBAC and OTULIN loss of function cause GSDs⁴². We previously reported that liver-specific deletion of the M1-polyubiquitin-specific deubiquitinase OTULIN leads to the loss of glycogen and severe liver inflammation⁵⁰. Together with other recent data⁴⁵,⁵¹, these findings suggest a role for M1-linked polyubiquitin in glycogen metabolism. It is tempting to speculate that M1-polyubiquitinated glycogen (rather than protein ubiquitination) could initiate inflammatory signalling.

Our data indicate ubiquitination of ‘normal’ glycogen in the liver of healthy mice, which increases during fasting, but also illustrates substantial differences in the ubiquitination of abnormal, unbranched polyglucosan (for example, when GBE1 function is lost). Ubiquitination of polyglucosan could function as part of a glycogen-directed quality control pathway. Of note, the ubiquitination events for each pathway may be distinct, and involve specific components (E3 ligases, deubiquitinases and ubiquitin signals), and cellular contexts. In these varied scenarios, ubiquitination of glycogen may alter glycogen metabolism by switching degradation pathways or catabolic flux, or may even protect glycogen from cytosolic catabolism.

Questions remain regarding (1) the types of ubiquitin chains that are attached to glycogen; (2) the proteins that recognize ubiquitinated glycogen; and (3) whether induced ubiquitination can clear soluble and insoluble glycogen in physiological settings. The last would open new possibilities for pharmacological treatment of GSDs and potentially other metabolic disorders.

Applying NoPro-clipping in an untargeted, exploratory manner revealed several ubiquitinated metabolites in cell lysates. We identified two of these, ubiquitinated (Ub)-glycerol and Ub-spermine, which we confirmed by creating in vitro standards that behave identically to the endogenous species. We also validated the endogenous origin of these ubiquitinated non-proteinaceous substrates by detecting both species in cells and mouse organs, by metabolic labelling, and by inhibition of a biosynthetic pathway. The synthesis of either species is unclear. Although it is possible that Ub-glycerol and/or Ub-spermine are products of to-be-identified E3 ligase reactions, these species might also arise via spontaneous discharging of E1 or E2 thioesters or pre-primed E3–E2–Ub complexes. Indeed, early work by Pickart & Rose had generated Ub-spermidine in vitro to characterize (and name) the ubiquitin C-terminal hydrolases⁵², which are capable of, and in fact may have evolved, to hydrolyze small molecules from the ubiquitin C terminus⁵³. Functionally, both spermine and glycerol are important cellular metabolites with dedicated biosynthetic pathways, yet, to our knowledge, their direct ubiquitination has not featured in conjunction with any of their roles. Further studies will address whether ubiquitination of these metabolites is modulated by physiological cues.

With the many roles of ubiquitination in human-disease settings and the emergence of ubiquitin-mediated targeted protein degradation as a therapeutic modality, there is a demand to understand the entire range of ubiquitin modifications, across cellular contexts of health and disease. NoPro-clipping provides a foundational tool for exploring this unchartered realm of ubiquitin biology.

Methods

Cloning

All cloning was performed using In-Fusion Cloning by Takara Bio. The sequences of N-terminally 2×Strep-tagged 3C protease and of N-terminally 2×Strep-tagged BpJOS²⁰ for bacterial expression were cloned into the pOPINK vector backbone⁵⁴ and the antibiotic resistance was switched from kanamycin to ampicillin. The sequence of USP10^368–798 for bacterial expression was cloned into the pOPINB vector⁵⁴, incorporating an N-terminal His-tag and 3C cleavage site. The sequence of USP21^200–565 and of sortase A-7M⁵⁵ for bacterial expression were cloned into the pOPINS vector⁵⁴, incorporating an N-terminal His-tag and SUMO cleavage site. For stable expression in mammalian cells, the sequences of mCherry–FRB–HOIL-1L and of mCherry–PYR–HOIL-1L were cloned into the lentiviral vector pFU-hygromycin⁵⁶, and the sequences of FKBP–GFP–CBM20 and of ABI–GFP–CBM20 into the lentiviral pFU-puromycin vector (pFU-hygromycin vector with antibiotic resistance switched to puromycin). ABI and PYR sequences have been described³⁴. The sequences of human 3×Flag–GYS1, 3×Flag–GYS2 and 3×Flag–HOIL-1L* (ref. ²⁶) were cloned into the pcDNA3.1 vector (Invitrogen) for transient expression. All sequences used in mammalian cells were codon optimized for human expression.

Expression and purification of proteins used in HOIL-1L-based ubiquitination reactions

Proteins used for HOIL-1L-based in vitro ubiquitination reactions (E1, UBE2L3, HOIL-1L, blocked M1 di-ubiquitin, ubiquitin, His-tagged ubiquitin and ubiquitin(S20C)) were expressed and purified as described²⁶.

Expression and purification of 2×Strep–3C-protease, BpJOS and Lb^pro*

2×Strep-tagged 3C protease and 2×Strep-tagged BpJOS were expressed in Escherichia coli BL21 cells (DE3). Bacterial cultures were grown in S-broth at 37 °C supplemented with carbenicillin until OD₆₀₀ = 0.6. Cultures were cooled to 18 °C and protein expression induced with 0.3 mM IPTG (Golden Biotechnology) and collected by centrifugation after 16 h. Cell pellets were resuspended in 50 mM Tris-HCl (pH 8.5), 300 mM NaCl, 10 mM DTT and 1 mM PMSF and lysed by adding lysozyme and DNaseI. Lysates were sonicated and cleared by centrifugation. Equilibrated Strep-Tactin XT resin (iba, 2-5030-002) was added for purification and incubated for 1.5 h. The resin was washed with 25 mM Tris-HCl pH 8.5, 500 mM NaCl and 5 mM DTT in a gravity flow column. 2×Strep-tagged 3C protease was eluted by incubating the resin in 50 mM Tris-HCl pH 8.5, 300 mM NaCl, 10 mM DTT and 50 mM biotin for 5 min. BpJOS was eluted by adding 2×Strep-tagged 3C protease (cleaving off the Strep-tag) for 2 h. Both proteins were further purified by size-exclusion chromatography using the AKTA pure chromatography system (Cytiva) and a HiLoad 16/600 Superdex 75 pg column (Cytiva). Lb^pro* was expressed and purified as described²².

Expression and purification of USP10^368–798, USP21^200–565 and sortase

His-3C-tagged USP10^368–798, His–SUMO-tagged USP21^200–565 and His–SUMO-tagged sortase A-7M were expressed in E. coli BL21 cells (DE3). Bacterial cultures were grown in 2× YT medium supplemented with kanamycin until OD₆₀₀ = 0.8. Protein expression was induced with 0.4 mM IPTG (Golden Biotechnology) and cultures maintained at 18 °C for 16 h. Cells were collected by centrifugation and pellets resuspended in lysis buffer. The lysis buffer for USP10^368–798 was 50 mM HEPES (pH 7.2), 300 mM NaCl, 30 mM imidazole, 2 mM beta-mercaptoethanol (βME), the lysis buffer for USP21^200–565 and for sortase was 50 mM Tris-HCl (pH 8.0), 300 mM NaCl and 2 mM βME. Cells were lysed by adding lysozyme, DNaseI and 1× cOmplete EDTA-free protease inhibitor cocktail (Roche), followed by sonication. Lysates were cleared by centrifugation and equilibrated Ni-NTA His-Bind resin (Merck, 70666-5) was added and incubated in a gravity flow column until the resin settled. The resin was washed with lysis buffer and protein was eluted by incubating the resin in lysis buffer containing 300 mM imidazole for 1 min. The His–SUMO tag was cleaved off by adding 6 µg ml⁻¹ His–SENP1 (ref. ^57), and the His–3C tag was cleaved off by adding 5 µg ml⁻¹ His–3C protease. Overnight dialysis was performed in 50 mM HEPES pH 6.8, 300 mM NaCl and 2 mM βME for USP10^368–798, and 50 mM Tris pH 8.0, 150 mM NaCl and 2 mM βME for all other proteins. USP10^368–798 was further purified by cation exchange using the AKTA pure chromatography system (Cytiva) and a RESOURCE S 6 ml column (Cytiva), with buffer A (50 mM HEPES pH 6.8 and 2 mM βME) and buffer B (HEPES pH 6.8, 1 M NaCl and 2 mM βME). USP10^368–798 and all other proteins were further purified by size-exclusion chromatography using the AKTA pure chromatography system (Cytiva) and a HiLoad 16/600 Superdex 75 pg column (Cytiva).

Generation of FITC-labelled ubiquitin

FITC-labelled ubiquitin was generated by labelling the S20C ubiquitin mutant with fluorescein-5-maleimide (Thermo 62245) per the manufacturer’s instructions. Ub S20C (2.3 mM in 240 µl buffer) was reduced using 0.5 mM TCEP for 15 min at room temperature. Reduced Ub S20C was diluted to 2 ml (~180 µM final concentration) in reaction buffer (25 mM HEPES pH 7.0, 100 mM NaCl). A fourfold molar excess of fluorescein-5-maleimide (10 mM stock in DMSO) was added dropwise to the diluted protein solution and incubated for 4 h at room temperature. The reaction was stopped with βME (10× molar excess over dye). FITC-ubiquitin was purified via size-exclusion chromatography using the AKTA pure chromatography system (Cytiva) and a HiLoad 16/600 Superdex 75 pg column (Cytiva) equilibrated in 10 mM HEPES (pH 7.9), 100 mM NaCl.

Generation and purification of ubiquitinated non-proteins

To generate ubiquitinated non-proteins in vitro (except for ubiquitinated glycogen and spermine), a reaction mix was made that contained 200 nM E1, 2 μM UBE2L3, 5 μM HOIL-1L, 10 μM blocked M1 di-ubiquitin, 10 μM ubiquitin, 10 mM MgCl₂, 0.5 mM TCEP and 20 mM of the respective non-protein substrate. If the ubiquitinated non-protein was later purified, His-tagged ubiquitin instead of native ubiquitin was used. Reactions were initiated by adding 10 mM ATP (pH 7) and incubated overnight at 30 °C. His-tagged ubiquitinated non-proteins were further purified according to ref. ²⁶. To generate ubiquitinated spermine, a reaction mix was made that contained 100 nM E1, 2 μM UBE2L3, 5 µM NEDD4 (ref. ²²), 10 μM native ubiquitin, 50 mM Tris pH 8, 100 mM NaCl, 10 mM MgCl₂, 0.6 mM DTT, 5% (v/v) glycerol and 20 mM spermine. 10 mM ATP (pH 7) was added to initiate the reaction, and samples were incubated overnight at 30 °C. To generate ubiquitinated glycogen, a reaction mixture was made that contained 1 µM of E1, 2 μM UBE2L3, 5 μM HOIL-1L, 10 μM blocked M1 di-ubiquitin, 10 µM FITC-labelled ubiquitin, 10 mM MgCl₂, 0.5 mM TCEP and 23 mg ml⁻¹ glycogen from bovine liver (Sigma). Reactions were initiated by adding 10 mM ATP (pH 7) and incubated overnight at 30 °C.

Generation of ¹³C-labelled ubiquitinated standards

Ubiquitin with an N-terminal His tag²⁵ was expressed in BL21 E. coli cells (DE3) grown in M9 minimal medium supplemented with 2 mg ml⁻¹ of ¹³C6-glucose (Cambridge Isotopes) as the sole carbon source, and was purified according to ref. ²⁶. Maltose, maltotriose, maltotetraose, spermine and glycerol were ubiquitinated with ¹³C-labelled His-tagged ubiquitin as described above. The concentration of each standard was determined by measuring the absorbance at 280 nm on a NanoDrop One (ThermoFisher Scientific) in triplicate. Concentrations were calculated using the theoretical extinction coefficient derived from ProtParam (https://web.expasy.org/protparam/).

Chemical synthesis of depsi ClipTag peptide

Biotin-YALPRT[G]G was synthesized by manual solid phase peptide synthesis. The Fmoc-protected TG depsi peptide was synthesized as reported²⁷. The 2-chlorotrityl chloride resin (1.46 mmol g⁻¹ loading) (Sigma) was swollen in DCM for 30 min. Fmoc-protected Gly (3 equivalents) was dissolved in DMF (5 ml) and DIPEA (6 equivalents) added. The solution was added to the swollen resin and the mixture left to roll overnight. The resin was drained, filtered and washed with DMF (3 × 5 ml) and DCM (3 × 5 ml) and DMF (5 ml). Fmoc deprotection was carried out with treatment of the resin with 4-methylpiperidine (20% in DMF + 0.1 M Oxyma) and rolled for 5 min at ambient temperature. This process was repeated 5 times before the resin was washed with DMF (5 × 5 ml × 2 min). Further amino acid incorporation dissolved 5 ml of the respective amino acid (4 equivalents of 1.0 M in DMF) with DIC (1 ml, 1 M in DMF) and Oxyma (1 ml, 1 M in DMF) and the resin rolled overnight. The Fmoc deprotection was repeated and peptide elongation continued sequentially to give the Fmoc-YALPRT[G]G sequence. The resin was treated with biotin (3 equivalents) in DMF 0.2 M (with DMSO) and HATU (3 equivalents) and DIPEA (5 equivalents) left to roll overnight. The resin was treated with resin cleavage cocktail (95:2.5:2.5 trifluoroacetic acid (TFA)/TIPS/H₂O) at 40 °C for 40 min. After cleavage, the filtrate was precipitated with ice-cold diethylether (35 ml) and then pelleted in the centrifuge (4 min, 3,000 rpm). The supernatant was decanted off and the pellet was washed with fresh diethylether (further two times) to give the crude peptide as a white powder. Purification by prep-HPLC afforded Biotin-YALPRT[G]G as a white powder (26.7 mg, 10%). LC–MS (ESI): 1.244 min, m/z: 531.4 [M + 2H]²⁺. LC purity: 100% (214 nm). ¹H NMR (300 MHz, DMSO) d 10.02 (s, 1H), 8.23 – 7.82 (m, 6H), 7.29 (s, 1H), 7.02 (d, J = 8.3 Hz, 3H), 6.62 (d, J = 8.1 Hz, 2H), 6.38 (d, J = 7.7 Hz, 2H), 4.65 – 4.24 (m, 9H), 4.20 –4.04 (m, 2H), 3.69 – 3.51 (m, 4H), 3.04 (t, J = 6.4 Hz, 2H), 2.93 – 2.77 (m, 3H), 2.65 – 2.52 (m, 2H), 2.10 – 1.74 (m, 7H), 1.70 – 1.31 (m, 10H), 1.26 – 1.09 (m, 5H), 1.05 (d, J = 6.3 Hz, 3H), 0.94 – 0.80 (m, 6H).

Gel electrophoresis and detection

To visualize ubiquitinated maltoheptaose (and deubiquitinases added to ubiquitinated maltoheptaose), samples in NuPAGE LDS Sample Buffer (Invitrogen) were loaded onto a NuPAGE 4–12% Bis-Tris gel (Invitrogen) and electrophoresis was performed until the dye reached the end of the gel. Gels were incubated in InstantBlue Coomassie Protein Stain (abcam) until protein bands appeared and were then washed with water. A ChemiDoc (v2.3.0.37 Bio-Rad) was used to acquire images. Uncropped gels are presented in Supplementary Fig. 9.

To visualize ubiquitinated glycogen, FITC-labelled ubiquitinated glycogen reaction mixtures in NuPAGE LDS Sample Buffer (Invitrogen) were loaded onto a NuPAGE 4–12% Bis-Tris gel (Invitrogen) and electrophoresis was performed until the dye reached the end of the gel. Fluorescent ubiquitin images were acquired on a ChemiDoc (Bio-Rad) using blue illumination and the 532/28 emission filter. Afterwards, gels were PAS stained following the Pierce Glycoprotein Staining Kit (Thermo Fisher Scientific, 24562), or both PAS and Coomassie stained. Gels were then photographed, uncropped gels are presented in Supplementary Figs. 9 and 13.

In vitro cleavage assay of ubiquitinated glycogen

FITC-ubiquitinated (S20C) bovine glycogen 5 mg ml⁻¹ were treated with 250 nM USP10 catalytic domain (amino acids 368–798) or OTULIN (full-length⁴⁷). Reactions were performed in 1× DPBS, 4 mM DTT at 37 °C. Reactions were quenched by removing 20 μl reaction into 3× NuPAGE LDS Sample Buffer (Invitrogen) for each time point before loading on SDS–PAGE. Gels were visualized as described above. Uncropped gels are presented in Supplementary Fig. 13.

Cell culture

Cells were cultured at 37 °C in a humidified 5% CO₂ atmosphere. Huh7, HepG2, HeLa and HEK293T cells were maintained in Dulbecco’s Modified Eagle Medium (DMEM, Gibco, 11885) supplemented with 10% (v/v) FBS (Bovogen, SFBS-F), 1× GlutaMAX (Gibco), 1 mM sodium pyruvate (Gibco) and 1,000 U ml⁻¹ penicillin/streptomycin (Gibco). SKOV3 cells were maintained in McCoy’s 5a (Modified) Medium (Gibco, 16600) supplemented with 15% (v/v) FBS, 1× GlutaMAX and 1,000 U ml⁻¹ penicillin/streptomycin. MEWO cells were maintained in Minimum Essential Medium (MEM, Gibco, 10370) supplemented with 10% (v/v) FBS, 1× GlutaMAX, 1 mM sodium pyruvate and 1,000 U ml⁻¹ penicillin/streptomycin. Huh7 and HepG2 cell lines were acquired from the American Type Culture Collection (ATCC) and authenticated by CellBank Australia via Short Tandem Repeat (STR) profiling. SKOV3, MEWO, HeLa and Hek293T cells were verified by the supplier, either the European Collection of Authenticated Cell Cultures (ECACC) or the American Type Culture Collection (ATCC).

Lentivirus production and generation of stable cell lines

For lentiviral production, 8 million HEK293T cells were seeded into 12 ml DMEM in 10 cm dishes. The following day, cells were transfected with a 1 ml mixture of 10 μg target vector, 7.5 μg psPAX2 packaging vector (Addgene #12260), 3 μg VSV.g packaging vector (Addgene #14888) and 82 μg ml⁻¹ PEI in Opti-MEM (Thermo Fisher). After 24 h, the transfection medium was discarded and replaced with 8 ml fresh DMEM. After another 24 h, the virus containing supernatant was collected, centrifuged (300g, 5 min, room temperature), filtered through a 0.45-μm syringe filter and stored at −80 °C until transduction. Target cells were seeded into 2 ml of medium (DMEM for Huh7 and HeLa cells, McCoy’s 5a for SKOV3 cells) in 6-well plates and transduced at 80% confluency by replacing the supernatant with 1 ml fresh medium and 1 ml virus containing supernatant. Polybrene was added at a concentration of 8 μg ml⁻¹ and cells incubated for 8 h, after which the supernatant was replaced with 2 ml of fresh medium. After 1–2 days, 50 mg ml⁻¹ hygromycin and/or 10 mg ml⁻¹ puromycin was added for 7 days or until non-transduced control cells died, to select for cells that had integrated the recombinant DNA. For stably expressed constructs (from pFU plasmids), GFP and/or mCherry positive cells were then sorted via fluorescence-activated cell sorting, pooled and recovered in fresh medium.

CRISPR genomic disruption

OTULIN and PYGL bulk knockout cell lines were generated by nucleofection using the SE Cell Line 4D-Nucleofector X Kit (V4XC-1032) as per the manufacturer’s instructions. In brief, targeting guide sequences were designed using the Synthego online CRISPR–Cas9 design tool (https://design.synthego.com/#/), Hs_OTULIN_sg2 GAGAAACCUCUUCAUAGCCC, Hs_OTULIN_sg11 GCUCAUGGCCUGGAACAGCG, Hs_PYGL_sgNA1 GUUGAAUGUAGCCUGCUUCU, Hs_PYGL_sgNA2 UACGGCAUUCGGUAUGAAUA and 2 × 10⁵ WT SKOV3 cells per condition were collected (400g, 5 min), washed once in PBS and reisolated. Cells were resuspended in premixed nucleofection buffer (SF solution and supplement mixed according to manufacturer’s instructions) and combined with RNP complexes made by combining 0.45 µl of each sgRNA with 1 µl of Alt-R S.p. Cas9 Nuclease V3 (IDT, 23829947) to achieve a final ratio of 1:3, Cas9:sgRNA. Cells were nucleofected in a 4D Nucleofection X Unit (Lonza) using the FE-132 program. Immediately after nucleofection, 100 µl of pre-warmed complete McCoy’s 5a medium was added, cells were incubated at 37 °C for 5–10 min and transferred to 24-well plates for recovery and expansion.

GBE1^−/− cells were generated by transient CRISPR–Cas9-mediated gene targeting as previously described⁵⁸. Guide sequences were designed using the CHOPCHOP online tool. Oligonucleotide sequences CACCGGTACTTGAAGCCCTACGCCG and AAACCGGCGTAGGGCTTCAAGTACC were annealed, phosphorylated (NEB T4 PNK) and ligated into px458 (containing Cas9 and GFP)⁵⁹. Correct guide sequence insertion was confirmed by sequencing and the plasmid was transfected into SKOV3 cells using Lipofectamine 3000 (Invitrogen). GFP-positive cells were single-cell sorted into 96-well plates and surviving clones were expanded and successful knockout was confirmed by immunoblotting.

Transient transfections for protein overexpression

Cells were transfected at ~80% confluency in antibiotic-free medium. For cells grown in 200 µl in chambered coverslips (microscopy experiments), a 25 µl mixture of 0.56 µl Lipofectamine 3000 (Thermo Fisher), 0.25 µg target vector and 0.5 µl P3000 (Thermo Fisher) in Opti-MEM (Thermo Fisher) was prepared. For cells grown in 10 ml in 10-cm dishes, a 1.25 ml mixture of 28 µl Lipofectamine 3000, 12.5 µg target vector and 25 µl P3000 in Opti-MEM was prepared. For cells grown in 20 ml in 15-cm dishes, a 3.75 ml mixture of 84 µl Lipofectamine 3000, 37.5 µg target vector and 75 µl P3000 in Opti-MEM was prepared. Transfection mixtures were vortexed, incubated for 15 min and dropwise added to the cell medium. Cells were incubated for 24 h at 37 °C until either further treatment commenced, or cells were collected.

siRNA-mediated gene silencing of HOIP

siRNA targeting HOIP (Thermo Fisher Scientific, 4392420, siRNA ID s30108, 5′−3′ sequence sense: GCAUGAACGACCCAGAAUAtt, antisense: UAUUCUGGGUCGUUCAUGCgt) was used as recommended by the manufacturer and a negative control siRNA (4390843, Thermo Fisher) was included as a control. In brief, SKOV3 GBE1^−/− cells were plated into 15 cm dishes in McCoy’s 5a medium in the absence of antibiotics, left to adhere overnight and transfected at approximately 70% confluency using a mixture of (1) 1,500 μl Opti-MEM (Thermo Fisher) and 90 μl Lipofectamine RNAiMAX (Thermo Fisher) with (2) 1,500 μl Opti-MEM and 50 nmol l⁻¹ siRNA per plate. The mixture was diluted 1 in 5 in cell culture medium and cells were grown for 48 h.

¹³C6-glucose labelling of Huh7 cells and HOIL-1L* overexpression

Huh7 cells were split into two and cultured for 10 days in either regular DMEM (Gibco, 11885) or DMEM without glucose (Gibco, 11966025) supplemented with 1 g ml⁻¹ of ¹³C6-glucose (Cambridge Isotopes). Both media were further supplemented with 10% (v/v) FBS (Bovogen, SFBS-F), 1× GlutaMAX (Gibco), 1 mM sodium pyruvate (Gibco) and 1,000 U ml⁻¹ penicillin/streptomycin (Gibco). During the 10 days, cells were split two times. Cells were then seeded in triplicates into 20 ml of their respective medium (containing glucose or ¹³C6-glucose) and grown in 15-cm dishes until ~80% confluency. 3× Flag–HOIL-1L* was transiently over-expressed for 24 h in all six plates and cells processed via NoPro-clipping.

Mouse handling

All mouse experiments were approved by The University of Melbourne Animal Ethics Committee (21123) and fulfill the guidelines set out by the National Health and Medical Research Council of Australia for the care and use of animals for scientific purposes. 8- to 11-week-old male C57Bl/6 mice were obtained from the Walter and Eliza Hall Institute animal facility (Kew, Melbourne, Victoria, Australia). All tissue collections were performed under isoflurane anaesthesia, with tissues (skeletal muscles, liver, lung, heart and brain) being immediately frozen in liquid nitrogen. Sample size was determined on the basis of previous experience and samples were randomized and blinded prior to processing and analysis.

For fasting experiments, food was withdrawn from mice at the end of the night feeding cycle (06:00), with continued ad libitum access to water. Fasted liver samples were obtained after fasting for 6 h, 24 h or 48 h. For refeeding experiments, mice were first fasted for 24 h and then provided access to food for 3 h or 24 h (both fasting and refeeding starting at 06:00). Control mice were maintained on the ad libitum diet, with liver samples also being obtained at the end of the night feeding cycle (06:00).

Human sample handling

Archived de-identified frozen skeletal muscle biopsies from the Victorian Neuromuscular Laboratory Service were selected and processed with approval from Alfred Health Ethics Committee (267/25). This work was performed in accordance with all relevant guidelines and regulations. An informed consent exemption was approved by Alfred Health Ethics Committee (267/25).

Confocal microscopy

For confocal experiments visualizing CBM20, HOIL-1L and ubiquitin co-localization, Huh7 cells were grown on chambered coverslips (µ-Slide 8 Well, ibidi) in 300 µl DMEM until ~80% confluency. Following experiment-specific treatments, the medium was removed and plates gently washed with PBS three times. Cells were fixed by incubating them in pre-warmed 4% (w/v) paraformaldehyde (PFA) for 20 min at 37 °C. After three more washing steps with PBS, cells were permeabilized by adding 0.2% (v/v) Triton in PBS for 20 min. Plates were again washed with PBS three times and blocking was performed with 1% (w/v) BSA in PBS for 20 min. Cells were immunostained for ubiquitin by incubation with anti-ubiquitin antibody FK2 (1:200 dilution in PBS + 1% BSA, Merck, 04-263) for 1 h. Following three PBS washing steps, Alexa Fluor 467 donkey anti-mouse antibody (1:1,000 dilution in PBS, Thermo Fisher, A-31571) was added for 30 min. Plates were washed with distilled water three times and ibidi mounting medium with DAPI (ibidi, 50011) added to cover the surface. Images were acquired at 40× magnification on a Zeiss LSM 980 with Airyscan 2 confocal microscope under oil immersion controlled via the ZEN (v3.7.97.03000) software package. Images were processed using the software Fiji/ImageJ2 (v2.14.0/1.54 h).

For confocal experiments quantifying CBM20, HOIL-1L, ubiquitin and LAMP1 or EEA1 co-localization, Huh7, SKOV3 and HeLa cells were seeded at 80k cells per ml in 8-well ibidi chamber slides prior to treatment. Cells were pre-treated with 6 µM Tak243 for 30 min prior to dimerizer addition. Dimerizer was added at a concentration of 500 nM for 4 h. For 24 h treatments, Huh7 and HeLa cells were seeded at 60k cells per ml in 8-well ibidi chamber slides prior to treatment. Bafilomycin A1 was added at a concentration of 300 nM, and dimerizer at a concentration of 500 nM. After respective treatment times, cells were fixed in 4% (v/v) PFA/PBS for 15 min, followed by three PBS washes. Permeabilization was done with 0.1% (v/v) Triton X-100/PBS for 10 min. The cells were then blocked with 3% (v/v) goat serum in 0.1% (v/v) Triton X-100/PBS for 15 min. Primary antibodies were added for 1.5 h at following concentrations: anti-GFP (Thermo Fisher, A10262) 1:500, anti-ubiquitin (FK2, Merck, 04-263) 1:500, anti-LAMP1 (Cell Signaling, 9091) 1:500, and anti-EEA1 (Cell Signaling, 3288) 1:500. Antibodies were washed with three PBS washes. Alexa Fluor secondary conjugated antibodies (Alexa Fluor 405 goat anti-mouse antibody, Thermo Fisher, A-31553; Alexa Fluor 646 goat anti-rabbit antibody, Thermo Fisher, A-21245; Alexa Fluor 488 goat anti-chicken antibody, Thermo Fisher, A-11039; or Alexa Fluor 594 goat anti-chicken antibody, Thermo Fisher, A-11042) were added at a concentration of 1:500 for 1.5 h. Three additional washes were performed with PBS, and chamber slides were maintained at 4 °C in PBS until images were acquired. Images were acquired with the Leica Stellaris 8 confocal microscope controlled with the LAS X (v4.8.2.29567) software package. Acquisition was achieved with 40× lens with 2× manual zoom in 3D by optical sectioning with a z-stack size of 4.2 µm, and a scan speed of 700 Hz. For co-localization quantification 3 images were taken as technical replicates per sample at random stage position, with 10–15 cells per field of view for the 4 h treatments, and 20–25 cells per field of view for the 24 h treatments. Three biological replicates were acquired per treatment condition. For quantification of co-localization of different channels, a macro for Fiji/ImageJ (v2.9.0) was designed that processes multi-channel z-stack images to quantify pixel co-localization in an automated manner. In brief, binary masks of all four channels were generated separately for every image in a z-stack. Universal suitable thresholds for every channel were determined and applied, followed by counting the amount of thresholded pixels colocalizing with every thresholded pixels of other channels respectively. Respective values were normalized to the untreated condition. Resulting data is represented as the sum of co-localization of indicated channels in all z-planes of one image. Uncropped microscopy images and exemplary images used for quantification are presented in Supplementary Figs. 2–8.

Live-cell imaging of CBM20 foci formation

Huh7 cells stably expressing FKBP–GFP–CBM20 and mCherry–FRB–HOIL-1L were grown on chambered glas bottom coverslips (µ-Slide 8 Well high Glass Bottom, ibidi 80807) in 300 µl DMEM until ~80% confluency. One hour before imaging, SiR-lysosome stain (spirochrome, sc012) was added to a final concentration of 1 µM. Immediately before imaging, rapamycin was added to a final concentration of 500 nM. Imaging was performed using the Zeiss Lattice Light Sheet 7 (LLS7, Zeiss). Time-lapse imaging was acquired using light sheets (488 nm, 561 nm and 640 nm) of 30 μm length with a thickness of 700 nm created at the sample plane via a 13.3 × 0.44 NA objective. Fluorescence emission was collected via a 44.83 × 1 NA detection objective. Aberration correction was set to a value of 170 to minimize aberrations as determined by imaging the Point Spread Function using 100 nm fluorescent microspheres at the coverslip of a glass bottom chamber slide. Data were collected with a frame rate of 10 ms and a y-step interval of 400 nm. Data were collected at a rate of 1 volume every 10 min for 24 h. A volume of 250 μm × 250 μm × 30 μm was acquired for each time point. Light was split to two cameras to minimize bleed through from the SiR-Lysosome signal into the mCherry–FRB–HOIL-1L channel. Signals for FKBP–GFP–CBM20 and mCherry–FRB–HOIL-1L were split to camera 1 via a LP 640 beam splitter and then further filtered using a dual band pass 495–550 nm and 570–620 nm filter. Signal for SiR-lysosome was sent to camera 2 through the LP 640 splitter and further filtered using a multi-band stop filter 405/488/561/640. Data are presented as maximum intensity projection time-lapse images and videos. The live-cell imaging video and still images are present in Supplementary Video 1 and Supplementary Fig. 1.

Fluorescence recovery after photobleaching

Huh7 cells stably expressing FKBP–GFP–CBM20 and mCherry–FRB–HOIL-1L or SKOV3 cells stably expressing ABI–GFP–CBM20 and mCherry–PYR–HOIL-1L were grown on chambered coverslips (µ-Slide 8 Well, ibidi) in either 300 µl DMEM or 300 µl McCoy’s 5a, respectively, until ~80% confluency. Depending on the condition, either GYS1 was transiently over-expressed for 24 h as described above or dimerizer (either rapamycin or mandipropamid) was added to a final concentration of 500 nM for 24 h, to induce CBM20 foci formation. All fluorescence recovery after photobleaching (FRAP) experiments were performed using a Zeiss 980 laser scanning confocal microscope. All images were captured via a Plan-Apochromat 40×/1.3 NA oil objective lens across 512 ×512 pixels with a pixel size of 83 nm and a pixel dwell time of 420 ns. The 488 nm and 594 nm laser lines were directed via a 488/594 nm main beam splitter and collected on a spectral array GaAsP-PMT detector with detection bands of 499–596 nm and 601–714 nm, corresponding to GFP and mCherry respectively. For FRAP experiments, 5 pre-bleach images were taken followed by a bleach step using the 405 nm, 450 nm, and 488 nm laser lines all set to 50%. A 2 µm² circular region was defined for bleaching and was followed by a 2-min time-series with a 0.5 s time interval. FRAP curves were recorded for multiple sub-regions within single cells. FRAP was performed on the FKBP–GFP–CBM20 signal only.

FRAP analysis was performed using the TrackMate⁶⁰ plugin in Fiji/Imagej⁶¹. Small sub-regions were extracted corresponding to the bleached regions of the Huh7 and SKOV3. The GFP-positive glycogen sub-region, corresponding to the bleached area, was segmented using the StarDist⁶² segmentation method. Segmented objects were tracked using an advanced Kalman tracking algorithm and the mean pixel intensity within the segmented region was measured through time. FRAP curves were normalized to the mean pre-bleach intensity and the minimum recorded intensity immediately post-bleaching, normalizing the curves between 0 and 1. FRAP curves were then averaged together and plotted as the mean with 95% confidence intervals.

Transmission electron microscopy

Cells were grown on Ibidi 35 mm dishes until 90%, medium was removed from each well and immediately replaced with 1 ml Karnovsky’s fixative, composed of 2% PFA, 2.5% glutaraldehyde, and 0.1 M cacodylate buffer in dPBS. Each Ibidi dish was thoroughly wrapped with parafilm, and cells were fixed for 2 h at room temperature.

Fixative was aspirated from all dishes, then cells were washed three times for 10 min with 0.1 M sodium cacodylate buffer at room temperature. Cacodylate buffer was removed and replaced with 2% osmium tetroxide and 1.5% potassium ferrocyanide solution, then incubated in a Biowave (Pelco) (150 W; vacuum on) three times in 2-min cycles. Cells were washed with sterile MilliQ water, and incubated in the Biowave (150 W; vacuum off) twice for 40 s, then once for 10 min at room temperature on a shaker. Each sample was then osmicated with 2% osmium tetroxide in MilliQ water, with the same Biowave incubation and wash steps performed as above. Cell samples were then dehydrated with a gradual ethanol series (30%, 50%, 70%, 90%, 100%, 100%) followed by two additional dehydrations with 100% anhydrous acetone, with a 40-s Biowave (150 W; vacuum off) incubation between each series. Cells were then infiltrated with Hard Epon resin dilutions (33%, 66%) prepared in 100% anhydrous acetone, followed by three 100% Hard Epon infiltrations, and incubated in the Biowave (350 W; vacuum on) for 3 min after each infiltration. After the final infiltration, the remaining Hard Epon was poured out of each dish, leaving a thin (<0.5 mm) layer of resin over the cells, which was polymerized in a 60 °C oven for 48 h.

After resin embedding, the polymer centre was cut and split into smaller (~2 mm²) pieces, which were mounted atop separate resin blocks. A Leica UC7 Microtome was used to reduce each sample area to a symmetrical trapezium (~0.5 mm²). Ultra-thin trimming was performed using a diamond knife (Ultra 45° Diatome), to obtain 70 nm slices and mounted onto copper grids. Grid sections were post-stained with 1% uranyl acetate for 3 min, washed 5 times in water, post-stained with 1% lead citrate for 3 min, and washed 5 times with water. Grids were then imaged on a JEOL JEM-1400Plus TEM at 80 kV.

Western blotting

Cells were washed with ice-cold PBS, scraped into ice-cold PBS, pelleted and resuspended in ice-cold RIPA lysis buffer containing 50 mM Tris-HCl (pH 7.5), 1% (v/v) NP-40, 0.25% (w/v) sodium deoxycholate, 0.1% (w/v) SDS, 150 mM NaCl, 1 mM EDTA, 10% (v/v) glycerol and 1× cOmplete EDTA-free protease inhibitor cocktail (Roche). Protein concentrations were measured using the Pierce BCA Protein Assay Kit (Thermo Fisher) and samples normalized to equal protein amounts using RIPA buffer. One part of lysate was combined with three parts of 4× NuPAGE LDS Sample Buffer (Invitrogen). Samples were loaded onto NuPAGE 4–12% Bis-Tris gels (Invitrogen) and electrophoresis performed until the dye reached the end of the gel. Proteins were transferred to PVDF membranes (Bio-Rad) using a Trans-Blot Turbo Transfer System (Bio-Rad). Standard western blot analysis was performed using primary antibodies against HOIL-1L (1:1,000, Novus Biologicals, NBP2-27105), FKBP12 (1:1,000, Invitrogen, PA1-026A), GFP (1:1,000, Invitrogen, A10262), Flag (1:1,000, Sigma, F1804), OTULIN (1:1,000, Cell Signaling, 14127), PYGL (1:1,000, Sigma, HPA000962), GBE1 (1:1,000, abcam, ab180596), HOIP (1:1,000, R&D Systems, MAB8039) and vinculin (1:1,000, Invitrogen, 700062). β-Actin-HRP (1:20,000, Santa Cruz Biotechnology, sc-47778 HRP), goat anti-rabbit-HRP (1:5,000, SouthernBiotech, 4010-5), donkey anti-chicken-HRP (1:5,000, Thermo Fisher, SA1-72004) or goat anti-mouse-HRP (1:5,000, SouthernBiotech, 1010-05) were used for enhanced chemiluminescence detection (ECL) with Clarity Western ECL substrate (Bio-Rad). A ChemiDoc (Bio-Rad) was used to acquire images. Uncropped blots are presented in Supplementary Figs. 10–13.

NoPro-clipping

Tissue culture cells were washed with ice-cold PBS and collected into ice-cold PBS by scraping. Cells were pelleted by centrifugation, supernatants discarded, and pellets were stored at −80 °C until lysis. Frozen cell pellets were resuspended in 450 μl ice-cold NoPro-clipping buffer (50 mM ammonium formate, 50 mM NaCl, 10 mM chloroacetamide, 10 mM N-ethylmaleimide, 5 mM EDTA, 5% glycerol (v/v), 1× protease inhibitor cocktail (Roche), and either 0.5% (v/v) NP-40 or 0.5% (w/v) N-dodecyl β-D-maltoside, pH 6.5). For SKOV3 cells in the bafilomycin A1 experiment (see Fig. 3e), samples were normalized to 1 mg total protein in 500 μl with NoPro-clipping buffer after total protein concentrations had been measured using the Pierce BCA Protein Assay Kit (Thermo Fisher). For experiments validating and quantifying ubiquitinated glycerol and spermine, glycerol was omitted from the lysis buffer and lysates were immediately desalted.

Frozen mouse and human tissue samples were homogenized in ice-cold NoPro-clipping lysis buffer supplemented with 5 mM EGTA. Protein concentrations were measured using the Pierce BCA Protein Assay Kit (Thermo Fisher) and samples were normalized to 1 mg total protein in 500 μl with NoPro-clipping buffer.

For some experiments, in vitro-ubiquitinated non-protein standards were added to the lysates. If measured on their own, in vitro-ubiquitinated non-protein standards were diluted in 500 µL 100 mM ammonium acetate (pH 6).

All samples were sonicated twice for 20 s (Omni Sonic Ruptor 400). If ubiquitinated glycogen was analysed, samples were incubated with 6 U ml⁻¹ α-amylase (Merck, A1031-1KU) at 37 °C for 2 h (750 rpm).

ZEBA desalting columns (7 kDa MWCO, 2 ml volume, Thermo Fisher) were pre-equilibrated with 100 mM ammonium acetate (pH 6) and centrifuged for 2 min (1,000g) 3 times. Samples were centrifuged (21,000g, 10 min, 4 °C), the clarified lysate transferred to ZEBA desalting columns and centrifuged at 1,000g for 2 min. Eluates were transferred to 1.5 ml Protein Lo-Bind microfuge tubes (Eppendorf) and incubated with either 100 nM BpJOS at 37 °C for 1 h (750 rpm) or 10 μM Lb^pro* at 37 °C for 16 h on a rotating platform at 750 rpm. 3 kDa MWCO columns (Amicon) were first pre-washed with 0.1 M NaOH and then pre-equilibrated with 100 mM ammonium acetate (pH 6) twice. Samples were transferred to the 3 kDa MWCO columns and GG-containing small molecules were collected by centrifugation (14,000g, 30 min, 4 °C). Samples were transferred into Protein Lo-bind tubes (Eppendorf), snap-frozen and lyophilized overnight. Dried samples were then stored at −80 °C until they were acquired via microflow LC–MS or further processed via sortase-mediated ClipTag labelling.

Sortase-mediated ClipTag labelling

Samples were reconstituted in 28 μl 50 mM ammonium acetate (pH 6) and incubated with 2.5 μM sortase enzyme and 25 μM depsi ClipTag peptide for 1 h at 37 °C. Samples were then quenched with 0.25% (v/v) TFA in 2% (v/v) acetonitrile. SDB stage tips (GL Sciences) were activated with 25 μl buffer B (80% (v/v) acetonitrile, 0.1% (v/v) TFA) and equilibrated with 25 µl buffer A (2% (v/v) acetonitrile, 0.1% (v/v) TFA), with centrifugation following each addition (3,000g, 2 min). Quenched samples were then loaded on to the tips and centrifugation was repeated, followed by one wash with 25 µl buffer A. Samples were eluted directly into level 3 SureStart 0.3 ml MS vials (Thermo) with 50 µl buffer C (80 % (v/v) acetonitrile, 0.1% (v/v) formic acid). Samples were dried via vacuum centrifugation and stored at −80 °C until nanoflow LC–MS acquisition.

Untargeted NoPro-clipping sample preparation

For untargeted NoPro-clipping, one 10-cm dish of Huh7 cells (either over-expressing HOIL-1L* or transfected with empty vector, for 24 h) was used per replicate and processed as described above. A ‘no-cell’ processing blank was treated identically to cell lysate conditions to enable identification and filtering of m/z features not originating from the Huh7 cells. To the cell lysates not over-expressing HOIL-1L, 195 µM in vitro-generated His-ubiquitinated ¹³C6-glucose was added. At the point of BpJOS treatment, each replicate was split into two equal samples and incubated at 37 °C for 1 h in the presence or absence of 100 nM BpJOS. Prior to sortase labelling, all samples were equally divided into two samples and incubated with 2.5 μM of sortase for 1 h at 37 °C in the presence or absence of 25 μM depsi ClipTag peptide.

Mass spectrometry data acquisition microflow workflow

NoPro-clipped samples were reconstituted in 70% (v/v) acetonitrile and injected onto an Ultimate 3000 HPLC connected to a Q Exactive Quadrupole-Orbitrap mass spectrometer (ThermoFisher), controlled with Xcalibur 4.2.47. Samples were separated over a ZIC-pHILIC column (2.1 mm internal diameter) heated to 45 °C and resolved with an acetonitrile/ammonium acetate gradient consisting of solvent A (100% (v/v) acetonitrile) and solvent B (20 mM ammonium acetate). The flow rate was 0.2 ml min⁻¹. The gradient started at 90% A for 2 min before reducing to 30% A over 18 min, held at 30% A for 4 min, before returning to 90% A for 0.5 min and re-equilibration for 5.5 min. MS1 scans were acquired across 100–1,500 m/z in both positive and negative mode at a resolution of 70,000. The automatic gain control (AGC) target was set to 3 × 10⁶ and the maximum injection time to 50 ms. For parallel reaction monitoring, multiple normalized collision energies were used (NCE, 10, 20, 26 or 30), with a quadruple isolation window of 1.2 m/z. The AGC target was set to 2 × 10⁵ and the maximum injection time to 20 ms. Xcablibr (v4.7.69.37 Thermo) software was used to operate the mass spectrometer and acquire the data.

Mass spectrometry data acquisition nanoflow workflow

NoPro-clipped and ClipTag-labelled samples were reconstituted in 2% (v/v) acetonitrile, 0.1% (v/v) formic acid and loaded on a 15 cm IonOpticks Aurora column maintained at 50 °C. A Neo Vanquish LC system (ThermoFisher) was directly coupled to an Eclipse mass spectrometer (ThermoFisher) controlled with Xcalibur 4.7.69.37, and ClipTag-labelled samples were separated with a binary buffer system of buffer A (0.1% (v/v) formic acid) and buffer B (99.9% (v/v) acetonitrile plus 0.1% (v/v) formic acid), at a flow rate of 400 nl min⁻¹. The gradient started at 2% B and increased to 5% over 1 min before increasing to 17% B over 14 min, 25% B over 10 min, and 34% B over 5 min. The column was flushed by increasing B to 85% in 0.1 min and held for 2.5 min at a flow of 800 nl min⁻¹. The column was re-equilibrated via pressure control (1,500 bar maximum) for 4.2 column volumes. The mass spectrometer was operated in positive polarity mode with a capillary temperature of 275 °C and 1,800 V. Data were either acquired as MS1 only or targeted parallel reaction monitoring. MS1-only mode consisted of scanning across 350–1,650 m/z with an AGC target of 1.2 × 10⁶ and a maximum injection time of 60 ms (R = 120,000). Parallel reaction monitoring was achieved with an MS1 scan across 350–1,650 m/z with an AGC target of 4 × 10⁵. Pre-defined MS1 parental masses were isolated using a 0.8 quadrupole isolation and fragmented at 28 (normalized collision energy) via HCD. MS2 scans were acquired at a resolution of 120,000 with an AGC target of 2.5 × 10⁵ and maximum injection time of 246 ms.

Intact mass spectrometry acquisition

Ubiquitinated maltose was monitored via intact mass spectrometry following with or without BpJOS clipping, addition of the ClipTag peptide and then incubation with sortase. Samples were diluted with 50% (v/v) acetonitrile, 0.1% (v/v) formic acid and directly infused into a Q Exactive mass spectrometer (ThermoFisher). Samples were ionized at 1,700 V at 250 °C using a HESI-II source at a flow rate of 10 μl min⁻¹. Data were collected at a resolution of 140,000 for 1 min, the spectra were averaged and subsequently deconvoluted using Freestyle 1.8 SP2 QF (Thermo). The spectra of Ub-maltose, UbΔGG, unreacted ClipTag peptide, and ClipTag-maltose were exported for annotation.

Absolute quantification of Ub linkages

Absolute quantification (AQUA) mass spectrometry was performed on mouse liver under a fed or 6 h fasted state. In brief, 20 µg of protein was reduced (10 mM TCEP) and alkylated (40 mM chloroacetamide) for 20 min at 55 °C and then digested with 2 µg of trypsin with micro-S-trap columns following the manufacturers specifications. Peptides were desalting with SDB stage tips (GL Sciences). Peptides were reconstituted in 2% (v/v) acetonitrile/0.1% (v/v) formic acid containing a mixture of Ub AQUA peptides (5 fmol ml⁻¹ final; Cell Signaling) and acquired on an Orbitrap Eclipse (Thermo Fisher Scientific) coupled to a Vanquish Neo. Peptides were separated using a 30 min gradient (solvent A, 0.1% formic acid; solvent B, 99.9% (v/v) acetonitrile/0.1% (v/v) formic acid) on a C18 analytical column (IonOpticks, Aurora Series Emitter Column, 15 cm × 75 mm). Mass spectrometry acquisition consisted of a SureQuant method scanning for each ¹³C¹⁵N standard, with 4/5 MS2 ions observed triggering ¹²C¹⁴N peptide MS2 fragmentation. Data were processed using the Skyline Daily and the area of each MS2 fragment combined and used to quantify the abundance of light peptide to the heavy synthetic peptide standard. Absolute quantification of total ubiquitin was determined using the mean from two unmodified peptides (TLSDYNIQK and ESTLHLVLR) summed with the K63 linkage peptide.

Glycogen quantification

Glycogen content from mouse tissue was determined by boiling 100 μl of samples homogenized in NoPro-clipping buffer (see NoPro-clipping) for 10 min. Boiled lysates were centrifuged at 12,000g for 5 min and the clarified supernatant was desalted using a ZEBA desalting column (7 kDa MWCO, 0.5 ml volume, Thermo Fisher). Liver samples were diluted 1/30, skeletal muscle 1/20, and all other tissues were kept undiluted. A colorimetric glycogen content kit (Abcam ab282931) was used and the absorbance of samples was measured at 590 nm with a ClarioStar plate reader (BMG Labtech). This method of glycogen extraction was selected to ensure consistency with handling/processing for Ub-glycogen quantification. Our extraction method may not capture the entire pool of insoluble glycogen present from cells and tissue. Hence the measurements here represent the soluble content and some proportion of the insoluble glycogen present and is referred to as the ‘detectable’ glycogen pool. Complete glycogen recovery requires harsh acid or base extraction, boiling the homogenate and precipitation with organic solvents⁶³. Milder extraction methods require pH 8 buffered extraction⁶⁴, which, like the harsher extraction alternatives, will hydrolyze ubiquitinated glycogen. Here we estimate the amount of glucosyl units per mg of protein to be 0.6 μmol versus 2.9 μmol (ref. ⁶⁵).

Glycogen content from cell lines was determined by washing cells four times with ice-cold PBS prior to collecting them into ice-cold PBS by scraping. One volume of 0.3 M HCl was added and cells were incubated for 5 min at room temperature. Samples were pH-adjusted with 1 volume of Tris (pH 5), and glycogen was measured using the Glycogen-Glo Assay (Promega, J5051). Luminescence was quantified using the 580/80 emission filter on a ClarioStar plate reader (BMG Labtech). For both glycogen quantification methods, negative controls in which samples were not incubated with glucoamylase, were measured to obtain residual glucose quantities. For experiments with glucose removal, the glycogen phosphorylase inhibitor (CAS: 648926-15-2) was added at 1 µM for 1 h prior to glucose removal.

Absolute quantification of ClipTag-species

To quantify ubiquitinated glycogen in mouse livers, samples homogenized in NoPro-clipping buffer (see ‘NoPro-clipping’) were split into three, containing 1 mg total protein each. The volume of each sample was adjusted to 500 µl with fresh NoPro-clipping buffer and the initial steps of the NoPro-clipping workflow—up to the ZEBA desalting step—were performed as described above. Just before samples were added to the desalting column, 1,000 fmol, 100 fmol or 10 fmol of Ub-maltose, Ub-maltotriose and Ub-maltotetraose, prepared with ¹³C-His-tagged ubiquitin (see above), were spiked into the three samples. Samples were mixed and loaded onto equilibrated ZEBA desalting columns and the NoPro-clipping workflow continued as described before. After data acquisition, linear standard curves were calculated for the concentration and measured peak areas of each ¹³C-labelled standard. Concentrations of ClipTag-maltose, -maltotriose and -maltotetraose with endogenous origin were inferred from measured peak areas using the standard curves.

The absolute quantities of endogenous ClipTag-maltose, -maltotriose, and -maltotetraose were added, which gave an estimate of at least 1.35 pmol ubiquitinated glycosyl units of glycogen per mg of total protein for 6 h fasted mice and at least 72 fmol for 24 h fasted mice. The fed control amount was determined to be 0.89 pmol ubiquitinated glycogen per mg of total protein by using the mean relative ratio of ClipTag-maltotriose measured between both cohorts.

Absolute quantification of ClipTag-glycerol and ClipTag-spermine in Huh7 cells were determined in a similar way as described above by spiking ¹³C-labelled ubiquitinated standards into cell lysates and processing via NoPro-clipping. The ClipTag-glycerol peak partially co-elutes with a lower-mass species, but its M + 3 isotope only differed in mass by 0.002. To ensure accuracy, the first half of the peak to apex was used for quantification to avoid co-isolation and fragmentation of this M + 3 species.

Targeted and untargeted mass spectrometry data analysis

EICs and MS2 spectra were visualized in FreeStyle (v1.8 SP2 QF1 ThermoFisher) and exported as comma-separated values.

For targeted mass spectrometry, parallel reaction monitoring data were analysed in Skyline (v21.2.0.369)⁶⁶ and peak area intensities for MS2 diagnostic ions exported. Structures were visualized with ChemDraw (v23.11.3) and used to annotate MS2 fragment ions. The following MS2 ions were used for each species; 631.3 for ¹²C-ClipTag-maltose, -triose, and -tetraose, 635.3 for the ¹³C4-labelled ClipTag-maltose, -triose, and -tetraose (used for absolute quantification), 561.3 and 564.3 for the ¹²C- and ¹³C3-ClipTag-glycerol, and 391.3 for ClipTag-spermine.

For untargeted mass spectrometry analysis, raw LC–MS files were centroided and converted to mzXML using msConvert (v3.0.24169-fe51146)⁶⁷. MS1 peak picking and grouping was performed in MAVEN (v2.10.13)⁶⁸. Peak extraction was performed with a 5ppm mass tolerance and minimum highest-peak intensity of 10,000. Minimum signal/noise ratio was set to 100 and minimum number of ‘good’ peaks per group set to three.

Downstream analysis and visualization were performed in R (v4.50) via R Studio. In brief, untargeted ClipTag analysis consisted of filtering MS1 m/z features that were observed in all replicates of the ‘BpJOS treated and ClipTag-labelled’ condition and had a coefficient of variation of <0.5. Data was further processed using the DEP R package (v1.32)⁶⁹. QRILC imputation was performed prior to differential expression analysis via limma⁷⁰. m/z features were retained if the log₂ fold-change was >2 for condition ‘BpJOS present and sortase present’ compared to (1) ‘BpJOS present and sortase absent’; (2) ‘BpJOS absent and sortase present’; (3) ‘BpJOS absent and sortase absent’; and (4) ‘BpJOS present and sortase present, with no cells included’. Finally, only m/z features that had an adjusted P value <0.05 for contrasts (1) to (3) were retained in the final dataset.

Data visualization

R 4.50, RStudio 2025.05.0 and the R packages tidyverse 2.0.0, ggplot 3.52 and ggrepel 0.9.6 together with Adobe Illustrator 28.5 were used to generate figures.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Online content

Any methods, additional references, Nature Portfolio reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at 10.1038/s41586-026-10548-x.

Author contributions

D.K., S.A.C. and M.J. conceived and designed the study and interpreted all data, and M.J. and S.A.C. designed workflows and performed most experiments. C.A.G. and P.G. contributed to the experimental design and C.A.G. performed all mouse studies up to tissue lysis. C.K. performed and analysed microscopy-based lysosomal co-localization experiments. A.C. generated ubiquitinated non-protein standards, purified proteins and performed independent repeats of experiments. L.F.F. created knockout cell lines. M.L.K. performed mandipropamid-dependent glycogen assays. R.A. performed electron microscopy. P.S. performed siRNA-mediated gene silencing experiments. X.S.W. and B.C.L. generated ubiquitinated non-protein standards. M.P. and L.T. cloned and tested plasmids for mammalian expression. S.M.D. and M.B. synthesized the depsi ClipTag peptide, T.H. and K.H. provided the BpJOS plasmid and protocols, and S.R.S. expressed and optimized BpJOS for Ub-clipping. N.D.G. performed FRAP and live-cell imaging experiments. C.A.M. provided human skeletal muscle biopsies. Y.S. performed preliminary linear ubiquitination experiments and B.C.L. provided critical reagents. D.K. and P.G. acquired funding. M.J., S.A.C. and D.K. wrote the manuscript.

Peer review

Peer review information

Nature thanks the anonymous reviewers for their contribution to the peer review of this work.

Data availability

All data generated during and/or analysed in this study are included in this published Article and its Supplementary Information. Uncropped versions of all gels and blots, and additional images for the cell biology studies, are provided in the Supplementary Information file. Source data for Figs. 1–5 and Extended Data Figs. 1–10 are provided with this paper. Untargeted mass spectrometry data have been deposited with the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD064299. Source data are provided with this paper.

Code availability

No specific scripts were developed in this study. The use of R packages is described in the relevant data analysis sections.

Competing interests

D.K. is founder, shareholder and scientific advisory board member of Entact Bio and Proxima Bio, and founder and scientific advisory board member of Ternarx. B.C.L. is founder and scientific lead at Ternarx. The Walter and Eliza Hall Institute has filed a provisional patent (AU) 2025905184 with D.K., S.A.C. and M.J. as inventors in relation to modulating glycogen via induced ubiquitination. The other authors declare no competing interests.

Extended data

is available for this paper at 10.1038/s41586-026-10548-x.

Supplementary information

The online version contains supplementary material available at 10.1038/s41586-026-10548-x.