Our interest in ISG15 originated in the course of experiments to elucidate the function of the NS1B protein of influenza B virus. We found that the NS1B protein binds ISG15 and inhibits its conjugation (6), indicating that ISG15 conjugation is likely to be an important part of the IFN-α/β-induced antiviral response. However, it was not evident how ISG15 conjugation might serve such a role. To address this issue and to elucidate the function of ISG15 conjugation, we first identified the E1 and E2 enzymes in the ISG15 conjugation pathway as Ube1L and UbcH8, respectively, both of which are induced by IFN-α/β (6, 7). These findings enabled us to develop a system for a proteomics-based identification of ISG15 target proteins, which is described in the present study.

Purification of ISG15 Conjugates. HeLa cells in each of five 150-mm culture dishes (total of 10⁸ cells) were transfected by using Fugene 6 (Roche) with three plasmids: 12 μg of pcDNA3-His₆-Flag-ISG15, 5 μg of pcDNA3-UbE1L, and 3 μg of pcDNA3-UbcH8. Transfection efficiency, as determined by using a GFP reporter plasmid, was ≈80%. Twenty-four hours posttransfection, 1,000 units/ml of human IFN-β (Berlex Biosciences, Richmond, CA) was added, and the cells were incubated for an additional 24 h. The untransfected control cells were incubated with 1,000 units/ml of human IFN-β for 24 h. Both sets of cells were lysed in a total of 5 ml of a buffer containing 8 M Urea, 50 mM Tris, pH 8.0, and 1.0% Triton X-100, followed by sonication for 40 sec with a Branson 450 Sonifier. Cell debris was removed by centrifugation at 18,000 × g for 30 min at 4°C. 2-Mercaptoethanol (β-ME) was then added to the resulting supernatant to a final concentration of 50 mM, and the mixture was incubated for 30 min at 4°C. Lysis buffer lacking Triton X-100 was then added to dilute the β-ME to 10 mM, and imidazole was added to 10 mM. The extracts were then subjected to the affinity selections described in Fig. 1B and the text and analyzed on a 10% SDS-polyacrylamide gel.

Target Verification Experiments. Several strategies for verifying ISG15 targets were used. In one strategy for HuP56 and MxA, immunoblots with anti-HuP56 antibody (provided by Ganes Sen, Cleveland Clinic, Cleveland) or anti-MxA antibody (provided by Otto Haller, University of Freiburg, Freiburg, Germany) were carried out on either whole-cell extract from IFN-β-treated HeLa cells or affinity-purified ISG15 conjugates. Because the tagged ISG15 used in these experiments was His₆-HA-ISG15, the purification procedure involved sequential affinity selection on anti-hemagglutinin (HA) antibody matrix (Covance, Richmond, CA) and Ni-NTA affinity matrix (Qiagen, Valencia, CA). In a second strategy for HuP56 and MxA, the whole-cell extract was immunoprecipitated with anti-HuP56, anti-MxA, or a control antibody, and the immunoprecipitate was immunoblotted with anti-ISG15 antibody or either anti-HuP56 or anti-MxA antibody. For the other ISG15 targets, cotransfection experiments were carried out by using ISG15 tagged with one epitope, and the target protein tagged with a different epitope, with or without cotransfection with plasmids expressing E1^ISG15 and UbcH8, as described in the text and the legends to Figs. Figs.2C2C and and33.

We carried out pilot purifications of ISG15 conjugates by using either His₆-HA-ISG15 or His₆–3xFLAG-ISG15 and found that a higher yield of purified conjugates was obtained by using the latter substrate. The scheme for the purification of His₆–3xFLAG-ISG15 conjugates from 10⁸ HeLa cells is outlined in Fig. 1B. The conjugates were purified by sequential affinity selections on Ni-NTA resin and anti-FLAG antibody matrix. Because efficient elution from the anti-FLAG antibody matrix required boiling in 1% SDS, which might also elute proteins that were nonspecifically bound to the matrix, we carried out a second Ni-NTA resin selection after the anti-FLAG column. The strongly denaturing conditions used in the two rounds of Ni-NTA selection would be expected to remove any proteins that were not themselves ISG15 modified but might have been in complexes with ISG15 protein targets. As a control, extracts from untransfected IFN-β-treated cells were subjected to the same purification procedure. Fig. 1C shows the Coomassie blue-stained SDS gel of the proteins purified from IFN-β-treated cells transfected with plasmids expressing His₆–3xFLAG-ISG15, E1^ISG15, and UbcH8 (lane 2) and from untransfected IFN-β-treated cells (lane 1). Both lanes contain heavy and light IgG chains, originating from SDS elution of the anti-FLAG affinity matrix. Lane 2 contains a large number of Coomassie-stained bands in the molecular weight region expected for ISG15 conjugates, whereas few or no bands were seen in this region of the gel in the control lane (lane 1).

This region of both gel lanes was cut into 11 slices, each of which was subjected to in-gel trypsin digestion (8). The resulting peptides were extracted from the gel and loaded onto a reverse-phase column. By using a gradient of increasing acetonitrile concentration, bound peptides were separated and eluted directly into a LTQ mass spectrometer. Once in the instrument, both MS (precursor mass) and tandem MS (sequencing) information was acquired for the eluting peptides. Data obtained in this manner were then used by the sequest search algorithm in conjunction with the human protein database to characterize the various peptides present (9). Identified peptides were subjected to stringent search standards, including high XCorr values. If a given protein was identified by two or more different peptides and was found only in the His₆–3xFLAG-ISG15 transfected cell sample (Fig. 1C, lane 2) and not in the control sample (lane 1), the protein was concluded to be a putative substrate of ISG15 and is tabulated in Table 1. This list of putative ISG15 targets contains 158 proteins. Proteins identified in both samples, or only in the control sample, were considered to be false positives and are tabulated in Table 2. Only 16 proteins are listed in this category, only seven of which are found solely in the control sample. Most of the proteins that were found in both samples were detected by a much larger number of peptides in the transfected sample compared with the control sample, indicating they are also likely to be ISG15 targets (e.g., the DHX15 splicing factor, p54nrb, and DNA topoisomerase1). However, we have not included these proteins in the list of putative ISG15 targets. Because there was very little overlap between the transfected and control samples, we can conclude that the purification procedure has removed most contaminating proteins lacking His₆–3xFLAG-ISG15 modification.

Twelve of the putative ISG15 target proteins are IFN-α/β-induced proteins, accounting at least in part for the increase in ISG15 conjugates resulting from IFN-β treatment. Nine of these IFN-α/β-induced proteins can be classified as antiviral proteins: PKR, MxA, GBP-1, HuP56, HuP54, HuP60, HuP58, RIG-I, and STAT1 (10). STAT1, the transcription factor required for the synthesis of IFN-α/β-induced mRNAs, was previously identified as an ISG15 target (4). We chose three of the IFN-α/β-induced antiviral proteins (HuP56, MxA, and RIG-I) and verified that they are ISG15-modified in IFN-β-treated cells. HuP56 binds to the eIF3 translation initiation factor and inhibits translation and virus replication (10), MxA is a large GTPase that inhibits the replication of a wide spectrum of viruses (10), and RIG-I is an RNA helicase that responds to RNA virus infection by triggering the activation of the transcription factors (IRF-3 and NF-κB) that are required for the production of IFN-β (11, 12).

HuP56 and MxA are abundant proteins, because they are highly induced in response to IFN-α/β (10). For this reason, we were able to establish that endogenous HuP56 and MxA are conjugated to ISG15 by the endogenous conjugation system. Suggestive evidence for conjugation was provided by direct immunoblotting of whole-cell extract from IFN-β-treated HeLa cells with anti-HuP56 or -MxA antibody. These immunoblots detected a minor HuP56-containing species that migrated at a position corresponding to a mono-ISG15 conjugate, and two minor MxA-containing species that migrated at positions corresponding to mono- and di-ISG15 conjugates, with the latter predominating (Fig. 2 A and B Left, lane 1). We took two approaches to definitively show that these bands corresponded to ISG15 conjugates. First, cells were transfected with a plasmid expressing His₆-HA-ISG15 (in the absence of plasmids expressing E1^ISG15 and UbcH8), and conjugates were purified by sequential affinity selection on anti-HA antibody and Ni-NTA affinity matrices. His₆-HA-ISG15-containing proteins were then immunoblotted with either anti-HuP56 (Fig. 2A Left, lane 3) or anti-MxA antibody (Fig. 2B Left, lane 3). These immunoblots detected a HuP56 protein species migrating at the expected position for HuP56-His₆-HA-ISG15 and two MxA species migrating at the positions expected for MxA- (His₆-HA-ISG15)₁ and MxA-(His₆-HA-ISG15)₂ (with the latter predominating), demonstrating that endogenous HuP56 and MxA are ISG15 modified. It should be noted that the His₆-HA-ISG15 conjugates migrated slightly slower than the conjugates containing endogenous untagged ISG15 (Fig. 2 A and B Left, compare lanes 1 and 3). In the presence of plasmid-expressed E1^ISG15 and UbcH8, increased amounts of the His₆-HA-ISG15 conjugates were produced, but the pattern of ISG15 conjugation was not changed by the enhanced expression of these two enzymes (Fig. 2 A and B Left, compare lanes 3 and 4).

In the second approach, untransfected IFN-β-treated cells were used to identify endogenous HuP56 and MxA conjugates containing only untagged endogenous ISG15. Whole-cell extract from IFN-β-treated HeLa cells was immunoprecipitated with either anti-HuP56 or -MxA antibody, followed by immunoblotting with anti-ISG15 antibody (Fig. 2 A and B Right). The ISG15 immunoblot detected a protein with the mobility expected for HuP56-ISG15 (Fig. 2A Right, lane 2), and an immunoblot with anti-HuP56 antibody confirmed that this species contains HuP56 (lane 3). This species was not present when the extract was immunoprecipitated with a control antibody (lanes 1 and 4). The MxA immunoblot detected a protein with the mobility expected for MxA-(ISG15)₂ (Fig. 2B Right, lane 2), and an immunoblot with anti-MxA antibody confirmed that this species contains MxA (lane 3). We conclude that endogenous HuP56 and MxA in IFN-β-treated HeLa cells are conjugated to endogenous ISG15 by the endogenous conjugation system.

For the validation of RIG-I, HeLa cells were transfected with plasmids expressing FLAG-RIG-I, E1^ISG15, UbcH8, and HA-ISG15, followed by IFN-β treatment (Fig. 2C). The control transfection left out the plasmid expressing HA-ISG15. Cell extracts were immunoprecipitated with anti-FLAG antibody and immunoblotted with either anti-FLAG antibody (lanes 1 and 2) or anti-HA antibody (lanes 3 and 4). RIG-I modified with one or more ISG15 molecules was detected in the cells expressing HA-ISG15 (lanes 2 and 4). The predominant conjugate migrated at the expected position of RIG-I-(ISG15)₃.

Most of the targets we identified by mass spectroscopy are constitutively expressed proteins. We chose five of these proteins and verified that they are ISG15 targets: TrxR1, an enzyme that protects against oxidative stress (13); maspin, a serpin that plays a role in normal development and malignant transformation of mammary glands (14); moesin, part of a three-protein complex that links plasma membrane proteins with the actin cytoskeleton (15); heat-shock protein 60 (Hsp60), a chaperonin that assists protein folding and protects proteins from denaturation after stress (16); and polypyrimidine tract-binding protein 1, an RNA-binding protein that regulates alternative splicing (17). The strategy for verification was similar to that used above for RIG-I, except that the target proteins were tagged with V5, and ISG15 was tagged with FLAG. The target protein was immunoprecipitated by using anti-V5 antibody, followed by immunoblots with anti-FLAG antibody to detect ISG15 conjugates (Fig. 3 A–E). Each of these five proteins was conjugated to one or more ISG15 molecules. The anti-V5 immunoprecipitate was also immunoblotted with anti-V5 antibody to verify that the target protein was ISG15 modified. As shown for TrxR1 and moesin (Fig. 3 A and B), a small fraction of these two proteins was ISG15 modified. Similar results were obtained for the other three proteins (data not shown). ISG15 conjugation of TrxR1 and moesin was also observed in the absence of plasmid-expressed E1^ISG15 and UbcH8 (Fig. 3 F and G), demonstrating that the endogenous ISG15 conjugation system modifies these two proteins. In the presence of plasmid-expressed E1^ISG15 and UbcH8, increased amounts of the ISG15 conjugates of TrxR1 and moesin were produced, but the pattern of ISG15 conjugation was not changed.

Three of the IFN-α/β-induced ISG15 targets identified by mass spectroscopy are Ub/Ubl conjugation enzymes. It is not surprising that two of these enzymes are UbE1L and UbcH8, the E1 and E2 enzymes in the ISG15 conjugation pathway. Proteomic analyses of SUMO- and Ub-conjugated proteins also identified several components of the conjugation systems for these proteins (20, 21). The third IFN-α/β-induced Ub/Ubl conjugation enzyme in our proteomic analysis was CEB1/Herc5, a HECT domain E3 ligase (22). CEB1/Herc5 may be a Ub E3 ligase that is ISG15 modified, analogous to the Nedd8 modification of cullins that activates cullin-based Ub ligases (23). Alternatively, CEB1/Herc5 may function as an E3 enzyme in ISG15 conjugation, particularly because at least some HECT domain E3 ligases have the ability to conjugate ISG15 to proteins in vitro (7). Our proteomic analysis also identified a constitutively expressed protein that functions in Ub conjugation, Ubc13, an E2 enzyme that catalyzes the synthesis of lysine-63 polyubiquitin chains in cooperation with Mms2 (24). Two SUMO-specific constitutively expressed enzymes were also identified, Pc2, a SUMO E3 ligase, and SENP1, a SUMO protease (18). These results suggest a possible crosstalk between the ISG15 pathway and the Ub and SUMO pathways at the level of conjugating/deconjugating enzymes. Interestingly, five of the identified ISG15 targets are also known SUMO targets (designated by ^** in Table 1).

The ISG15 targets identified by mass spectroscopy were categorized with respect to function and intracellular localization, as depicted in Fig. 4. Some ISG15 targets fall into more than one functional category. A relatively large number of the ISG15 targets are functionally or structurally associated with the cytoskeleton, including a number of major intermediate filament proteins such as vimentin. This likely explains a previous immunofluorescence study that localized ISG15 to intermediate filaments (25). In addition, two ISG15 targets, moesin and ezrin, are part of a tripartite complex that links the plasma membrane to the underlying actin cytoskeleton (15). Because this protein complex is essential for efficient cell-to-cell spread of Listeria and probably other bacteria (26), it is reasonable to hypothesize that ISG15 modification of one or more of the proteins in this complex may play a role in the host defense against these bacterial infections. Proteins that function in stress responses were also identified as targets, including key enzymes involved in oxidative stress, TrxR1, glutathione S-transferase (13), and several heat-shock proteins. A significant number of metabolic enzymes were identified as ISG15 targets by mass spectroscopy, although the significance of this finding in unclear.

Perhaps the most surprising finding was that a large number of ISG15 target proteins function in RNA processing or in chromatin remodeling/RNA polymerase II transcription. These proteins account for most of the nuclear targets of ISG15, which represent ≈30% of the total ISG15 targets. The chromatin remodeling/polymerase II transcription factors range from a component of the basal polII transcription machinery (TAFII150) to sequence-specific transcription factors (Smad4 and FUBP3), to chromatin remodeling factors (CHD-1, RBBP4, and Sin3a). The proteins that function in RNA processing include many splicing factors, which range from protein components of the essential U2 (DDX42, SAP49, and SAP62) and U5 snRNPs (U5–200 kD) to an alternative splicing factor (PTB-1), as well as several of the heterogeneous nuclear ribonucleoprotein proteins that are bound to nuclear premRNA before splicing (27). These results show that ISG15 conjugation impacts nuclear functions and may have a role in regulating transcription and premRNA splicing during the IFN-α/β response.