Screen for ISG15-crossreactive deubiquitinases

Abstract

Background: The family of ubiquitin-like molecules (UbLs) comprises several members, each of which has sequence, structural, or functional similarity to ubiquitin. ISG15 is a homolog of ubiquitin in vertebrates and is strongly upregulated following induction by type I interferon. ISG15 can be covalently attached to proteins, analogous to ubiquitination and with actual support of ubiquitin conjugating factors. Specific proteases are able to reverse modification with ubiquitin or UbLs by hydrolyzing the covalent bond between their C-termini and substrate proteins. The tail regions of ubiquitin and ISG15 are identical and we therefore hypothesized that promiscuous deubiquitinating proteases (DUBs) might exist, capable of recognizing both ubiquitin and ISG15.

Results: We have cloned and expressed 22 human DUBs, representing the major clades of the USP protease family. Utilizing suicide inhibitors based on ubiquitin and ISG15, we have identified USP2, USP5 (IsoT1), USP13 (IsoT3), and USP14 as ISG15-reactive proteases, in addition to the bona fide ISG15-specific protease USP18 (UBP43). USP14 is a proteasome-associated DUB, and its ISG15 isopeptidase activity increases when complexed with the proteasome.

Conclusions: By evolutionary standards, ISG15 is a newcomer among the UbLs and it apparently not only utilizes the conjugating but also the deconjugating machinery of its more established relative ubiquitin. Functional overlap between these two posttranslational modifiers might therefore be more extensive than previously appreciated and explain the rather innocuous phenotype of ISG15 null mice.

Figure 1. Phylogram representation of DUBs.

Indicated with arrows are DUB homologs that were cloned and expressed by in vitro transcription/translation (IVT). Red arrows depict DUBs that bound to neither probe (UbVME, ISG15VS, or SUMO1VS), whereas black arrows indicate DUBs that formed covalent adducts with the indicated probes. Our screen represents the first biochemical proof for protease activity of USP13 and the Otubain-homolog CGI-77 (DUB homologs without publication record regarding biochemical function are marked with an asterisk). Otubain1 (OTU1) is an exception in that it binds to alkylhalide- or aldehyde-based probes, but not to the Michael acceptors employed in this study (data not shown).

Figure 2. Activity-based profiling of DUBs.

IVT was used to generate DUBs for profiling activity toward ISG15VS, UbVME and SUMO1VME probes. After incubation with the probes, samples were analyzed by SDS-PAGE, as shown for USP18 (A), USP14 (B), USP5 (C), USP13 (D), USP2 (E), and CGI-77 (F). Binding was inhibited by preincubation of the proteases with NEM. The SUMO protease SENP2 interacts specifically with SUMO1VME (G). Molecular weight in kDa is indicated on the left.

Figure 3. Unexpected apparent molecular mass of ISG15VS-DUB adducts is caused by unusual behavior in SDS-PAGE.

(A) Plot showing the ratio of observed versus expected ISG15VS-DUB adduct size of USP2, USP5, USP13, USP14 and USP18 in SDS-PAGE (8%) over the size of the unmodified DUBs. (B) Mutation of the catalytic cysteine residue to serine (C114S) in USP14 abolishes its reactivity toward UbVME and ISG15VS. When stored for longer periods at room temperature, the probes polymerize covalently, presumably by formation of secondary amine bonds between internal lysine residues and the reactive Michael acceptor at the C-terminus, thus resembling isopeptide-linked polyubiquitin. Such polymeric probes of UbVME likely caused the additional high-molecular mass adducts observed for USP14. Note that the smallest version of these adducts (a UbVME dimer) has a maximum electrophoretic mobility similar to that of the diubiquitin-like ISG15VS when complexed to USP14. The absence of any adducts in the C114S mutant of USP14 excludes the possibility of multiple binding sites for the probes. (C) Purified recombinant USP14 labels with UbVME and ISG15VS and results in the same abnormal mobility shift for ISG15VS-USP14 as seen in the IVT labeling experiments. (D) MALDI-TOF mass spectroscopic analysis of USP14 after incubation with ISG15VS. As described above for UbVME, ISG15VS also engages in internal polymerization. Molecular masses consistent with tri- and tetrameric ISG15VS are marked in this spectrogram by the numbers in superscript. Monovalently modified USP14 results in an adduct of predicted size, indicated with a red arrow. This complex is unique to the mixture containing both USP14 and ISG15VS, and is absent in the mass spectra of either component alone (data not shown).

Figure 4. Reactivity of USP14 toward ISG15VS is augmented by proteasomal association.

USP14 and USP5 were generated by IVT. Their activity toward ISG15VS and UbVME was analyzed in the presence of increasing concentrations of purified human 26S proteasomes. (A) Activity of USP14 toward UbVME and ISG15VS increases as a function of the concentration of added purified 26S proteasomes (in mg/ml). (B) The activity of USP5 remains unaffected. (C) Quantification of the radioactive signal of covalently modified USP5 and USP14. Binding affinity is depicted on the y-axis as percent in labeling intensity, determined by the ratio of labeled versus unlabeled USP5 or USP14. The ratio in the absence of exogenous proteasomes is defined as 100%.

Figure 5. Proteasome-associated USP14 has ISG15-specific isopeptidase activity.

(A) Scheme depicting the UbL-peptide conjugate used to assay isopeptidase activity. The biotinylated peptide heptamer is attached to either ISG15 or SUMO1 in isopeptide-linkage. Upon hydrolysis of the isopeptide bond by a specific DUB, the heptamer is released and the biotin signal lost. (B) Incubation of proteasome-enriched fraction (“5 hr pellet”) with UbL-peptide conjugate. After overnight incubation, the ISG15-peptide conjugate is completely cleaved, resulting in loss of the biotin-signal (significant proteolysis occurs already after one hour, data not shown). This activity is sensitive to NEM. Hydrolysis is not observed for the SUMO1-peptide conjugate. (C) Anti-HA immunoblot of HA-ISG15VS treated subcellular fractions. Based on previous identification and on electrophoretic mobility, USP5 is the dominant ISG15-reactive DUB in the five-hour supernatant, which is enriched for uncomplexed proteins of light and moderate size (red asterisk). The five-hour pellet represents heavy cytosolic complexes, in particular the 26S proteasome, and contains USP14 as the only ISG15-reactive DUB (blue asterisk).

Figure 6. In vivo probe-binding studies and analysis of subcellular DUB localization.

(A) Schemes of the EYFP fusion proteins: UCH-like Zinc Finger Motif (red), proteolytic UCH domain (blue, with the putative active-site cysteine depicted in yellow), ubiquitin-binding UBA-domain (light blue), ubiquitin-like domain (green), putative nuclear localization sequence (pink), EYFP fusion protein (yellow box). (B) In vivo binding of ubiquitin and ISG15 probes. Lysates obtained from 293T cells transfected with USP14_EYFP or USP14_C114S-EYFP were reacted with UbVME or ISG15VS and compared to untreated aliquots in an anti-YFP immunoblot. USP14_EYFP but not USP14_C114S-EYFP reacts with the probes. (C) Subcellular localization of DUBs analyzed via the distribution of C-terminal EYFP fusions. 24 hours post-transfection, 293T cells were fixed with paraformaldehyde and analyzed by confocal fluorescence microscopy. USP5_EYFP can be found throughout the cell (upper left). USP13_EYFP is expressed mainly within the nucleus, in a speckled pattern (upper right panel). USP14_EYFP is detected predominantly in the cytoplasm, with lower levels in the nucleus (lower left panel). USP3_EYFP and USP36_EYFP (lower right panel) are detected in the nucleus and nucleolus, respectively.

Figure 7. Structures of USP2 and USP14, modeled with ubiquitin and ISG15.

Both USP2 and USP14 accommodate ubiquitin (green) in a shallow groove, measuring approximately 25–30 Å in diameter (view from top shown in the upper panels, view from front shown in the lower panels). Based on these complexes, the C-terminal domain of ISG15 (red) can be modeled into the groove to replace ubiquitin. The N-terminal domain of ISG15 is not involved in the hydrolysis reaction, and is located outside of the catalytic core domain of both proteases. The N-termini of USP2 and USP14 are lacking in these representations and extend to the right side of the structure models.