Precursor Processing of Pro-ISG15/UCRP, an Interferon-β-induced Ubiquitin-like Protein*

Induction of the 17-kDa ubiquitin-like protein ISG15/UCRP and its subsequent conjugation to cellular targets is the earliest response to type I interferons. The polypeptide is synthesized as a precursor containing a carboxyl-terminal extension whose correct processing is required for subsequent ligation of the exposed mature carboxyl terminus. Recombinant pro-ISG15 is processed in extracts of human lung fibroblasts by a constitutive 100-kDa enzyme whose activity is unaffected by type I interferon stimulation. The processing enzyme has been purified to apparent homogeneity by a combination of ion exchange and hydrophobic chromatography and found to be stimulated 12-fold by micromolar concentrations of ubiquitin. Analysis of the products of pro-ISG15 processing enzyme demonstrates specific cleavage exclusively at the Gly157-Gly158 peptide bond to generate a mature ISG15 carboxyl terminus. Irreversible inhibition of pro-ISG15 processing activity by thiol-specific alkylating agents and a pH rate dependence conforming to titration of a single group of pK a 8.1 indicate the 100-kDa enzyme is a thiol protease. Partial sequencing of a trypsin-derived peptide indicates the enzyme is either the human ortholog of yeast Ubp1 or a Ubp1-related protein. As yeast do not contain ISG15, these results suggest that a ubiquitin-specific enzyme was recruited for pro-ISG15/UCRP processing by adaptive divergence.

Type I interferons (interferons-␣ and -␤) induce a specific subset of proteins responsible for mediating the antiviral and broad pleiotropic effects of these cytokines, reviewed in Ref. 1. One of the earliest responses to type I interferons is the increased transcription of the ISG15 1 gene, which is controlled by a 5Ј interferon-stimulated response element (1,2). The sequence of the resulting 17.1-kDa ISG15 protein, also known as the ubiquitin cross-reactive protein (UCRP), is composed of two ubiquitin-like domains that retain the canonical LRLRGG sequence required for conjugation of ubiquitin to intracellular targets as the committed step for 26 S proteasome-dependent degradation (3)(4)(5). The biological effects of ISG15 are similarly mediated by its covalent conjugation to a small subset of cellular proteins through an enzyme pathway that is distinct from that of ubiquitin ligation (5,6). Both free and conjugated ISG15 pools are significantly induced following exposure to interferon-␣/␤; however, in contrast to the homeostatic mechanism that maintains a constant ratio of free to conjugated ubiquitin (7), the two pools of ISG15 exhibit a characteristic biphasic induction during which free polypeptide increases early during the interferon response and then undergoes increased conjugation from 12 to 72 h (5). The ISG15 polypeptide acts in trans to noncovalently bind associated target proteins and model chimeric constructs to intermediate filaments 2 ; in contrast, free ISG15 shows only marginal affinity for binding to these cytoskeletal components (8). Neither free nor conjugated ubiquitin levels are affected by interferon treatment, indicating that up-regulation of ISG15 ligation is a specific and independently regulated response (4,5). ISG15 was the first example of a small class of ubiquitin-like proteins that includes SUMO-1, 3 Nedd8, 4 and Ubl1 (reviewed in Refs. 9 and 10). These proteins exhibit significant sequence similarity to ubiquitin and, with the exception of ISG15, contain single ubiquitin domains (11). Although SUMO-1 and Nedd8 polypeptides are found throughout eukaryotes (9,10), search of the complete yeast genome has identified no candidate ISG15 ortholog, indicating that this regulatory pathway represents a relatively recent functional divergence. 5 Ubiquitin and ubiquitin-like proteins are synthesized as precursors bearing carboxyl-terminal extensions, which range in size from short polypeptides to complete protein domains (12). Expression of the carboxyl-terminal GTEPGGRS extension peptide of pro-ISG15 minimizes degradation of the nascent polypeptide, presumably through a kinetic folding effect. 2 Although the function(s) of the unrelated extension peptides on newly synthesized chains of ubiquitin and the other ubiquitin-like proteins is unknown, their presence requires processing of the precursor protein to expose the carboxyl-terminal glycine required in isopeptide bond formation to their respective targets. In addition, correct processing is required for a novel extracellular cytokine function for ISG15 involving induction of CD56 ϩ natural killer cell proliferation, augmentation of non-major histocompatibility complex-restricted cytotoxicity, and T-cell interferon-␤ induction (13). Processing of the pro-ISG15 carboxyl-terminal extension occurs rapidly without significant ac-cumulation of the precursor (4,14), as is generally characteristic for processing of the entire class.
A large family of enzymes has been described that act proteolytically at the carboxyl terminus of ubiquitin to generate the free polypeptide from fusion proteins, ubiquitin-protein conjugates, multiubiquitin chains that serve as degradation signals for 26 S proteasome specificity, and small molecule adducts arising by reaction of cellular nucleophiles with activated intermediates of ubiquitin (15). These enzymes collectively belong to the family of ubiquitin carboxyl-terminal hydrolases (UCH), which have been arbitrarily divided into two subfamilies based on molecular weight, sequence homology, and leaving group specificity (15,16). Family 1 is composed of low molecular mass (ϳ30 kDa) enzymes, of which three tissuespecific mammalian isozymes (L1, L2, and L3) have been described (15,17) that have related orthologs in yeast and Drosophila (16). These enzymes are believed to function in generating mature ubiquitin from the precursor fusion proteins and in salvaging ubiquitin carboxyl-terminal adducts, based on observations that this class prefers small leaving groups or extended polypeptide chains (15). Family 2 encompasses the ubiquitin-specific proteases and the deubiquitinating enzymes, which exhibit considerable heterogeneity in size (50 -300 kDa) and show little sequence homology except for conserved cysteine and histidine motifs containing the active site residues (16). Over 11 members of this family have been described in yeast with orthologs identified in mammals and Drosophila (16). This family of ubiquitin proteases functions to release free ubiquitin from ubiquitin-protein conjugates and their polyubiquitin degradation signals (16). The large number of UCH enzymes suggests a wide variety of cellular functions and/or potential roles with other ubiquitin-like proteins.
In the present work, we identify two pro-ISG15 processing activities in cell extracts of the human lung carcinoma line A549. A minor activity (Ͻ1%) is probably a member of the UCH isozyme family, based on molecular weight and competitive inhibition by free ubiquitin. In contrast, free ubiquitin is a positive allosteric effector of the major activity, which shares features of family 2 with respect to size and apparent catalytic mechanism. Partial sequencing following purification of the latter processing enzyme to apparent homogeneity suggests this activity is either the human ortholog of the yeast family 2 enzyme Ubp1 or a Ubp1-like protein. These observations suggest that an existing ubiquitin-specific enzyme has been recruited for pro-ISG15 processing by adaptive divergence.

MATERIALS AND METHODS
Bovine ubiquitin was purchased from Sigma and purified to apparent homogeneity (18). Human recombinant interferon-␤ containing a C17S mutation to enhance stability was supplied by Triton Biosciences.
Recombinant Pro-ISG15 Expression and Purification-Recombinant pro-ISG15 was expressed in Escherichia coli BL21 (DE3) cells harboring the pET11d-ISG17 plasmid. Purification of pro-ISG15 was carried out as described previously for mature recombinant ISG15 (6), with the exception that inclusion of CoCl 2 during isolation was omitted inasmuch as carboxypeptidase inactivation by cleavage of the carboxylterminal glycine dipeptide from ISG15 was precluded by the presence of the carboxyl-terminal extension peptide present on the precursor. The resulting recombinant pro-ISG15 was Ͼ99% pure by SDS-PAGE when visualized by silver staining. The concentration of pro-ISG15 was determined spectrophotometrically using the empirically determined ⑀ 280 nm of 0.79 ml/mg⅐cm for mature ISG15 (6) corrected for the molecular weight difference. Typical yields of pro-ISG15 were 2-3 mg/liter culture (A 600 nm ϭ1.0). Purified pro-ISG15 could be stored for several months at Ϫ80°C without a significant effect on processing activity and was unaffected by at least three freeze-thaw cycles.
Generation of the Pro-ISG15-Y Mutant-The pro-ISG15-Y mutant bearing an additional tyrosine residue after Ser 165 of the extension peptide was generated by polymerase chain reaction using a 19-mer forward primer encompassing the T7 promoter of the pET11d vector and a 39-mer reverse primer in which a tyrosine codon was engineered between the carboxyl-terminal serine of pro-ISG15 and the STOP codon followed by a BamHI restriction site. The resulting polymerase chain reaction product was digested with NcoI and BamHI, then ligated into NcoI/BamHI-digested pET11d. After transformation into E. coli DH5␣, the vector containing the mutant protein was isolated and the sequence confirmed by the dideoxy method (19). Expression and purification of the pro-ISG15-Y mutant was identical to that for the wild type protein, yielding 2 mg/liter culture (A 600 nm ϭ1.0) that was Ͼ99% pure by SDS-PAGE analysis and silver staining. The presence of the tyrosine residue in the purified mutant protein was confirmed by amino acid analysis by the Medical College of Wisconsin Protein and Nucleic Acid Shared Facility and laser desorption time of flight mass spectrometry. The absolute quantity of pro-ISG15-Y was determined spectrophotometrically using a calculated extinction coefficient of ⑀ 280 nm ϭ 0.85 ml/mg⅐cm. Control studies demonstrated that pro-ISG15-Y was processed at the same rate as the wild type pro-ISG15, indicating no significant structural differences between the two proteins. This has been independently confirmed by the quantitatively similar circular dichroism spectra for mature and precursor forms of ISG15.
The mutant protein was radiolabeled with IODOGEN (Pierce) as described previously (6), except that the iodination reaction was carried out for 15 min to preferentially label the additional carboxyl-terminal tyrosine residue rather than the two internal tyrosines present in ISG15. Specific activities ranged from 160 to 1600 dpm/pmol, with ϳ2% of the total pro-ISG15-Y bearing the radiolabel. Based on the sequence and secondary structure similarities between ubiquitin and ISG15, the carboxyl terminus of pro-ISG15 is assumed to be solvent-exposed whereas the two internal tyrosine residues are buried and thus less accessible to radioiodination. Product analysis after the processing of 125 I-pro-ISG15-Y revealed no detectable signal present in mature ISG15, confirming that the label was exclusively at the carboxyl-terminal tyrosine.
Activity Assay-The pro-ISG15 processing activity was monitored either by a SDS-PAGE gel-shift assay exploiting the molecular weight difference between pro-ISG15 and the mature polypeptide or by the release of the radiolabeled trichloroacetic acid-soluble carboxyl-terminal extension peptide from 125 I-pro-ISG15-Y. For the gel shift assay, reactions of 20 l final volume containing 50 mM Tris-HCl (pH 7.6), 1 mM DTT, and 5-25 M pro-ISG15 were incubated at 37°C in the presence of A549 10 5 ϫ g supernatant. At the indicated times, incubations were quenched by adding an equivalent volume of 2ϫ Laemmli sample buffer containing 4% (v/v) ␤-mercaptoethanol and boiled for 5 min. Unprocessed pro-ISG15 and the mature ISG15 product were resolved by 16% SDS-PAGE and then electrophoretically transferred to BA83 nitrocellulose (Schleicher & Schuell). Following sequential incubation of blots with affinity-purified anti-ISG15 or anti-ubiquitin antibodies (10 g/ml) and 125 I-Protein A (2 ϫ 10 5 cpm/ml), immunospecifically bound antibody was visualized by autoradiography as described previously (5,20). Product ISG15 was quantitated by excising the bands from the dried blot, determining the associated radioactivity by ␥ counting, and comparing the latter to that obtained with parallel recombinant ISG15 standards.
The addition of a tyrosine residue at the carboxyl terminus of pro-ISG15 allowed the processing reaction to be followed by the appearance of trichloroacetic acid-soluble radioactivity representing the cleaved peptide bearing the labeled tyrosine residue. Incubations were carried out as described for the gel shift assay but were quenched by addition of 0.2 ml each of 5 mg/ml carrier bovine serum albumin and 20% (w/v) ice-cold trichloroacetic acid. Quenched reactions were incubated on ice for 10 min and then centrifuged at 14,000 ϫ g for 10 min. The supernatant was quantitatively transferred to another tube and the pellet washed with 0.2 ml of ice-cold 2% (w/v) trichloroacetic acid, which was combined with the supernatant. Both the pellet and the supernatant were counted. Radioactivity present in the supernatant was corrected for label present in a 0-min blank that generally represented Ͻ10% of the total radioactivity. Subsequent studies to be described below demonstrated that the major pro-ISG15 processing activity exhibited a linear dependence on substrate concentration below 0.6 M and was stimulated by the presence of free ubiquitin. Therefore, a unit of processing activity was defined as the quantity required to produce 1 pmol/min mature ISG15 at pH 7.5 and 37°C in the presence of 5 M pro-ISG15, 25 M ubiquitin, and 1 mM DTT.
Cell Culture-Human A549 lung carcinoma cells were obtained from the American Type Culture Collection and grown in Dulbecco's modified Eagle's medium (high glucose) supplemented with 2 mM L-glutamine and 10% (v/v) fetal calf serum. Monolayer cultures were maintained in a humidified 5% CO 2 atmosphere at 37°C. Cells were seeded at a density of 5 ϫ 10 5 cells/ml with one change of medium 24 h after seeding and were harvested at confluence (approximately 48 h after seeding) without further manipulation. Cells were rinsed with phosphate-buffered saline and then scraped into 50 mM Tris-HCl (pH 7.6) containing 0.25 M sucrose and 1 mM DTT at a concentration of ϳ10 8 cells/ml, followed by sonication for 10 s. The broken cells were centrifuged at 4°C and 10 5 ϫ g for 90 min to obtain the cytosolic fraction, which was flash-frozen in small aliquots and stored at Ϫ80°C.
FPLC Column Chromatography-Cell cytosol (15-50 mg of cellular protein) was applied to a Mono Q HR 10/10 anion exchange column equilibrated at 4°C in 50 mM Tris-HCl (pH 7.6) containing 1 mM DTT at a flow rate of 2 ml/min. Bound proteins were eluted at 1 ml/min with a linear 0 -0.5 M NaCl gradient (10 mM/ml). Mono Q fractions representing the major processing activity were pooled and adjusted to 1 M (NH 4 ) 2 SO 4 and then applied to 2.5 ϫ 36-cm preparative Phenyl-Sepharose Fast Flow FPLC column (2 ml/min) equilibrated at 4°C in 20 mM sodium phosphate buffer (pH 7.6) containing 1 M ammonium sulfate and 1 mM DTT. Bound protein was eluted with a linear 1-0 M ammonium sulfate gradient (Ϫ5 mM/min) at a flow rate of 2 ml/min. Fractions containing pro-ISG15 processing activity were pooled and dialyzed at 4°C against 50 mM Tris-HCl (pH 7.5) containing 1 mM DTT. The enzyme was stable for several months when flash-frozen and stored at Ϫ80°C but lost activity with successive freeze-thaw cycles.
HPLC Analysis of the Peptide Product-Processing reactions were carried out in 50 mM Tris-HCl (pH 7.6) containing 1 mM DTT and were quenched by addition of 0.2 ml each of 5 mg/ml bovine serum albumin and 20% (v/v) trifluoroacetic acid and then allowed to set on ice for 10 min. Control studies demonstrated that trifluoroacetic acid quantitatively precipitated unprocessed 125 I-pro-ISG15-Y and the mature ISG15 product. Following centrifugation of quenched samples for 10 min at 14,000 ϫ g, the supernatant containing the processed carboxyl-terminal extension peptide was quantitatively removed and the trifluoroacetic acid volatilized in vacuo. The resulting pellet was resuspended in 1% (v/v) aqueous trifluoroacetic acid and applied to a 4 ϫ 250-mm Bio-Sil ODS-5S reverse phase C18 HPLC column (Bio-Rad) equilibrated in 1% (v/v) aqueous trifluoroacetic acid (0.7 ml/min). Bound peptide was eluted with a linear 0 -60% acetonitrile gradient (2.1%/ml) and was monitored at 215 nm.
Microsequencing of the Pro-ISG15 Processing Enzyme-The 100-kDa pro-ISG15 processing enzyme obtained after Phenyl-Sepharose chromatography ( Fig. 5) was resolved from minor contaminants by 8% (w/v) SDS-PAGE, after which the sample was electrophoretically transferred to Immobilon-CD membrane and visualized by fluorescamine staining. The 100-kDa band was excised from the blot and then extracted from the Immobilon membrane by incubation in 70% (v/v) formic acid. Following vacuum centrifugation, the dry protein was dissolved in 0.1 M ammonium bicarbonate (pH 8.0) and then digested for 2 h at room temperature with 1% by weight of TPCK-treated sequencing grade trypsin. A second aliquot of trypsin was added, and the incubation was allowed to continue an additional 2 h. The resulting sample was lyophilized, washed twice with water, and dissolved in 20 l of 5% (v/v) aqueous acetonitrile. The peptide mixture was resolved by reverse phase chromatography using a Hewlett Packard 1090 HPLC fitted with a 2 ϫ 150-mm UltraSphere (Beckman) column equilibrated in 0.1% (v/v) aqueous trifluoroacetic acid (40°C). Sample was applied at 0.2 ml/min and bound peptide detected at 215 nm following elution with a linear 5-65% acetonitrile gradient. Fractions containing a well resolved, apparently homogeneous peak were pooled and dried in vacuo, after which the resulting peptide was sequenced on a Beckman/Porton LF3000 peptide sequencer.

RESULTS
Pro-ISG15 Processing Activity Is Not Induced by Interferon-␤-Incubation of recombinant pro-ISG15 with an A549 postribosomal supernatant obtained from uninduced confluent cultures results in the appearance of an anti-ISG15 immunoreactive band that comigrated with authentic mature ISG15 when analyzed by Western blotting as described under "Materials and Methods" (Fig. 1). Endogenous free ISG15 present in the cell extract is not detected at the exposure time chosen for the autoradiogram of Fig. 1. An absolute rate for pro-ISG15 processing of 4.2 pmol/min/10 6 cells was calculated by quantitating mature ISG15 associated 125 I radioactivity by ␥ counting and comparing to signal obtained from an authentic recombinant mature ISG15 standard (Fig. 1, left lane). That equivalent amounts of mature and precursor forms of ISG15 yield similar autoradiographic intensities (Fig. 1, left lanes) indicates that the presence of the carboxyl-terminal extension peptide does not significantly affect binding of the affinity purified rabbit polyclonal antibodies.
Similar pro-ISG15 processing is observed for A549 extracts harvested after 24 h of treatment in the presence of 10 3 IU/ml interferon-␤, a concentration previously shown to yield maximum rates of ISG15 induction (4) (Fig. 1). Elevated levels of endogenous mature ISG15 are evident in the blank sample (lanes B) obtained prior to addition of exogenous pro-ISG15 and in the 0-min lane quenched prior to addition of the substrate (Fig. 1). Quantitation of the absolute amounts of mature ISG15 in the interferon-␤-induced incubations and correction for that introduced with the A549 extract yielded a linear rate for pro-ISG15 processing of 4.4 pmol/min/10 6 cells. These results indicate that pro-ISG15 processing activity is constitutively present in uninduced cells and is not significantly affected by interferon treatment, confirming earlier qualitative results from pulse-chase experiments (3).
When incubations similar to those of Fig. 1 are extended to longer times, a progressive loss of pro-ISG15 and appearance of the mature product is observed when probed with anti-ISG15 antibody ( Fig. 2A). Approximately 5% of the initial pro-ISG15 remained at the longest times examined (Fig. 2B). The residual unreacted pro-ISG15 was not due to inactivation of the processing activity inasmuch as no further loss of pro-ISG15 was observed when an additional aliquot of fresh A549 cell extract was added (data not shown). Inability of the processing enzyme to convert the remaining fraction of substrate likely represents denaturation of pro-ISG15, consistent with the complete absence of processing by fresh extracts in the presence of heat denatured pro-ISG15 (data not shown). Both the decline in pro-ISG15 and appearance of the mature product displayed rigorous first order kinetics over 10 half-lives (solid lines in Fig.  2B) for which first order rate constants of 0.53 h Ϫ1 and 0.58 h Ϫ1 , respectively, were determined from nonlinear leastsquares fitting of the data in Fig. 2B. Quantitation of the Western blot showed that appearance of mature ISG15 accounted for 78% of the pro-ISG15 lost, the remaining 22% likely resulting from incomplete degradation of the polypeptide by the extracts.
Identical samples from the upper panel of Fig. 2A were also probed with affinity-purified anti-ubiquitin antibodies, which displayed a reduced recognition for pro-ISG15 compared with the mature polypeptide. Previous work has shown that the anti-ubiquitin antibodies are directed against an epitope(s) in the carboxyl terminus of the polypeptide, accounting for the ability of these immunoprobes to discriminate between free and conjugated ubiquitin (20). As the carboxyl-terminal LRLRGG is conserved between ubiquitin and ISG15, it is reasonable that the anti-ubiquitin antibodies can readily detect mature ISG15. As the anti-ubiquitin antibody provided a significantly improved signal to noise ratio for Western blot quantitation of the mature ISG15 band when poorly resolved from pro-ISG15, these antibodies were used in subsequent gel shift assays.
Purification of the Pro-ISG15 Processing Activity-Resolution of the A549 post-ribosomal supernatant on a Mono Q HR10/10 anion exchange FPLC column results in two peaks of activity (Fig. 3). The smaller peak consistently represented 10% of the recovered activity eluted at 0.18 M NaCl (Peak 1), whereas the major peak of activity eluted at 0.27 M NaCl (Peak 2). Separate control studies confirmed that the processing activities were unaffected by NaCl concentrations present in the fractions (data not shown). The activity in peak 1 eluted at a native molecular mass of 31 kDa when subsequently resolved by Superose 12 gel filtration chromatography (data not shown); in addition, processing activity by peak 1 was competitively inhibited by 50 M ubiquitin (data not shown), a concentration approximating that of the free polypeptide in this cell line (5). These two observations suggested that the activity contained in peak 1 represented a member of the UCH family 1; however, this activity was not human UCH-L3 inasmuch as an authentic recombinant sample of this enzyme failed to act on pro-ISG15 (data not shown). The relative native molecular mass of the activity in peak 2 proved to be 100 kDa by Superose 12 chromatography (data not shown). At higher concentrations of sample, the activity eluted at apparent molecular masses that were approximately multiples of the 100-kDa mass found at low protein concentrations, suggesting that the enzyme undergoes a weak concentration-dependent oligomerization.
We consistently observed a significant loss in total recovered activity when cytosolic A549 extracts were resolved by Mono Q chromatography. This was not due to instability of the activity, inasmuch as control samples not loaded onto the column retained complete activity. Nor was the loss the result of denaturing interactions with the column, inasmuch as all activity was recovered when the sample was loaded in high NaCl to block binding. This technical problem was resolved when we observed that 5 M free ubiquitin significantly stimulated the processing activity (Fig. 4), in contrast to the expected competitive inhibition by free ubiquitin if the activity were a ubiquitin-specific enzyme. The significant loss in activity during Mono Q FPLC is explained by resolution of the processing activity from free endogenous ubiquitin, which appears in the unbound fraction at pH 7.5. In the gel shift assay of Fig. 4, approximately the same stimulation of pro-ISG15 processing is observed at 5 and 50 M, indicating that the effect of ubiquitin is saturated at the lower concentration. The experiment does not distinguish whether ubiquitin acts as an allosteric activator of the processing enzyme or stabilizes ISG15 against proteolytic degradation in the extract. The latter alternative is less likely, as exogenous ISG15 was relatively stable in the cell extracts when followed by quantitative Western blotting; in addition, bovine serum albumin had no effect on the rate of pro-ISG15 processing, suggesting a specific effect of free ubiquitin (Fig. 4).
Fractions encompassing peak 2 were pooled and adjusted to 1 M ammonium sulfate then applied to a 2.5 ϫ 36-cm preparative Phenyl-Sepharose Fast Flow column equilibrated with 25 mM sodium phosphate buffer (pH 7.5) containing 1 M ammonium sulfate and 1 mM DTT. Bound proteins were eluted from the column with a 1-0 M negative linear ammonium sulfate gradient (Ϫ5.5 mM/ml) at a flow rate of 2 ml/min. The 125 I-pro-ISG15-Y processing activity eluted as a single symmetric peak at 0.22 M ammonium sulfate. Resolution by SDS-PAGE of fractions spanning the 125 I-pro-ISG15-Y processing activity revealed a single band of 100-kDa relative molecular mass by Coomassie Blue staining (Fig. 5A). Fig. 5B compares the initial rates for formation of mature ISG15 versus the quantity of 100-kDa protein determined densitometrically (arbitrary units). That the 100-kDa protein band corresponds to the pro-ISG15 processing activity is indicated by the excellent correspondence between activity and protein, confirmed by the constant relative specific activity across the protein peak (Fig. 5C). Fractions from the Phenyl-Sepharose activity peak were pooled and concentrated, then dialyzed against 50 mM Tris-HCl (pH 7.5) containing 1 mM DTT. The pooled Phenyl-Sepharose activity represented a 250-fold purification with a 65% yield relative to the post-ribosomal supernatant. Table I summarizes the three-step purification of the pro-ISG15 processing activity to apparent homogeneity.
Pro-ISG15 Processing Occurs at the Predicted Site-The data of Fig. 2 suggest processing of pro-ISG15 occurs at the correct site between glycine 157, representing the carboxyl terminus of the mature polypeptide, and glycine 158 of the carboxyl-terminal octapeptide extension; however, the size discrimination by SDS-PAGE was not adequate to confirm this cleavage position unambiguously. In a larger processing reaction similar to that of Fig. 2 but containing tracer levels of 125 I-pro-ISG15-Y to monitor product formation and recovery, quantitation revealed that 2 nmol of total pro-ISG15 was processed in 30 min at 37°C using 60 units of the partially purified enzyme from the Mono Q HR10/10 column. The trifluoroacetic acid-soluble superna- tant from this incubation was resolved by C18 reverse phase chromatography, as described under "Materials and Methods." Control reactions conducted in the absence of either processing enzyme (Fig. 6A) or pro-ISG15/ 125 I-pro-ISG15-Y (data not shown) yielded comparable HPLC elution profiles. Quantitation of the acid-soluble radioactivity indicated complete recovery of the 2 nmol of processed extension peptide, half of which was applied to the HPLC column. The resulting elution profiled showed a new peak at 23 min (Fig. 6B). The peak of radioactivity eluted slightly earlier than the new absorbance peak at 23 min (data not shown), presumably due to differences in binding between unlabeled and iodopeptide. The absolute amount of 125 I-nonapeptide present was below the limit of detection in trace B.
That the new peak eluting at 23 min was the expected octapeptide product was demonstrated by coelution with 1 nmol of synthetic unlabeled octapeptide (Fig. 6C). In addition to retention of elution position (Fig. 6, insets), the integrated peak area in panel C agreed well with that expected for the sum of processed peptide and exogenous standard. In addition, the predicted molecular mass for the HPLC-purified peptide peak was confirmed by mass spectrometry (data not shown). Therefore, the 100-kDa processing enzyme catalyzes the specific cleavage of the extension peptide to yield mature ISG15, confirmed by the ability of the resulting ISG15 to support ubiquitin activating enzyme-catalyzed ATP:PP i exchange (data not shown).
Characterization of the Processing Enzyme-The kinetics of the purified enzyme was quantitated by following the processing of radiolabeled pro-ISG15-Y. Preliminary isotope dilution experiments confirmed that 125 I-pro-ISG15-Y was identical to wild type pro-ISG15 in rates of processing (data not shown). At pH 7.5 the initial velocity for 125 I-pro-ISG15-Y processing was linear to 0.6 mM polypeptide, the highest concentration tested (data not shown), which sets a lower limit for K m of ϳ2 mM based on the sensitivity of the assay to detect deviations from linearity with respect to substrate concentration that would be associated with saturable binding. Addition of 50 M ubiquitin increased V m /K m 12-fold (5.9 ϫ 10 Ϫ4 min Ϫ1 ) without altering the linearity of the concentration dependence (data not shown).
In the presence of 25 M each of 125 I-ISG15-Y and ubiquitin, the initial rate of processing measured by the formation of trichloroacetic acid-soluble radioactivity increased nonlinearly from pH 6 to 9 (Fig. 7, closed circles) and then decreased abruptly at higher pH values (Fig. 7, open circles). The sharp decrease in initial rate above pH 9 likely results from alkaline denaturation of the processing enzyme inasmuch as the loss of activity was irreversible (data not shown). Below pH 9, the pH dependence for initial rates of processing could be best fit to a single titration of pK a 8.1 using a non-linear curve fitting algorithm (GraFit, Erithacus Software, Ltd.). Similar pH dependence for processing was observed in the absence of added ubiquitin (data not shown), indicating that the pK a 8.1 titration observed in Fig. 7 was not contributed by the polypeptide.
The effect of selected protease inhibitors on 125 I-pro-ISG15-Y processing was determined at pH 7.5 (37°C) by measuring the pseudo first order rate constants for inactivation. With the exception of EDTA and EGTA, all other inhibitors listed in Table II exhibited first order kinetics for the quantitative inactivation of processing activity when corrected for the slow rate of spontaneous inactivation in the absence of inhibitors. Iodoacetic acid, iodoacetamide, and TLCK exhibited the greatest rates for the irreversible inactivation of 125 I-pro-ISG15-Y processing (Table II). Differences in reactivities among the thiol-selective alkylating reagents probably reflect steric constraints and the active site microenvironment. In contrast, the serine protease-selective inhibitor PMSF exhibited an ϳ20-fold lower rate constant for inactivation relative to iodoacetamide. The relative difference in rates of reaction between the thioland serine-selective inhibitors suggests the processing enzyme is a thiol protease, although the data do not rule out modification of an essential but noncatalytic sulfhydryl group. Measurable inhibition was also exhibited by the chymotrypsin and trypsin inhibitors TPCK and TLCK, respectively, suggesting the processing enzyme has a reactive nucleophile that reacts with the carbonyl groups present on both inhibitors. The ϳ2fold greater effect of TLCK may reflect the lysyl residue of the inhibitor mimicking arginine 155 of pro-ISG15, similar to the inhibition of ubiquitin activating enzyme by TLCK. 5 Neither EDTA or EGTA had any measurable effect on processing activity at concentrations of 10 mM, suggesting the absence of a metal requirement for octapeptide cleavage.
Microsequencing of the Pro-ISG15 Processing Enzyme-We were unable directly to sequence the apparently homogeneous pro-ISG15 processing enzyme from Fig. 5, presumably due to a blocked amino terminus; however, sequencing from trypsindigested immobilized protein following SDS-PAGE resolution, as described under "Materials and Methods," was successful. One well resolved major peak from reverse phase resolution of the resulting trypsin-digested peptides was chosen for partial sequencing. Although the amount of peptide (1.4 pmol) was near the limit of detection, unambiguous peaks could be assigned for 8 of 11 cycles (0.8 pmol signal in final cycle). Comparison to the GenBank and SwissProt data bases yielded significant homology to only two proteins: the 100-kDa Saccharomyces cerevisiae Ubp1 and the 33-kDa predicted product of human clone KIAA0161 (GenBank accession no. D79983) (Fig.  8). As processing activity correlated with the 100-kDa polypeptide band identified by SDS-PAGE in Fig. 5, the predicted molecular mass of the KIAA0161 gene product precludes it as the candidate processing enzyme in the absence of possible sequencing artifacts. In contrast, agreement between the partial sequence of the pro-ISG15 processing enzyme and Ubp1 is consistent with the relative molecular mass of 100 kDa (Fig. 5); the pH dependence for processing, which suggests the observed pK a 8.1 corresponds to the active site cysteine present in Ubp1 (Fig. 7); and the inhibitor profile, which also suggests the activity is a cysteine protease (Table II). The partial peptide sequence of the processing enzyme corresponds to an inserted segment within the His box motif common to all class II ubiquitin-specific proteases that is unique to yeast Ubp1 (Fig. 8). Therefore, the pro-ISG15 processing enzyme is either the human ortholog of yeast Ubp1or a Ubp1-related protein. DISCUSSION The constitutive and interferon-␤-induced biological effects of ISG15 require correct processing of the octapeptide extension from the precursor protein in order to expose the mature carboxyl terminus required for subsequent conjugation to intracellular targets and for extracellular cytokine activity (5,13). Processing of carboxyl-terminal extensions from nascent precursors is a requirement shared by all ubiquitin and ubiq-uitin-like proteins identified to date (11), although the enzyme(s) responsible for such activation has not been identified. We were initially interested in distinguishing whether pro-ISG15 processing occurred through existing ubiquitin-specific enzyme(s), as an alternate substrate, or through a pro-ISG15specific enzyme activity.
The present studies reveal two pro-ISG15 processing activities in A549 cell extracts (Fig. 3), both of which appear to be ubiquitin-specific enzymes for which pro-ISG15 serves as an alternate substrate. A minor component, representing 10% of total recovered activity after Mono Q anion exchange chromatography, likely results from the action of a UCH. This conclusion is supported by the native molecular mass of 31 kDa determined by gel filtration chromatography and competitive inhibition of pro-ISG15 processing by exogenous free ubiquitin. However, the activity is not human UCH-L3 as an authentic recombinant sample of this enzyme failed to process pro-ISG15 yet rapidly cleaved tyrosine 77 from a 125 I-UbY77 construct (data not shown). Instead, our attention focused on the major component, comprising 90% of the total recovered activity following Mono Q chromatography (Fig. 3). Although the latter activity exhibited trace inhibition at low ubiquitin concentrations (data not shown), the effect was masked by the marked stimulation of processing activity by free ubiquitin at physiological concentrations (Fig. 4). Stimulation of pro-ISG15 processing by free ubiquitin accounted for the marked loss of activity during Mono Q FPLC chromatography; in addition, when activity was adjusted for the effect of ubiquitin, this enzyme accounts for greater than 99% of total pro-ISG15 processing activity in A549 extracts.
The major processing activity was purified to apparent homogeneity and quantitatively correlated with a protein band of 100 kDa (Fig. 5). Partial microsequencing of this enzyme yielded a peptide sequence showing significant similarity to S. cerevisiae Ubp1 (GenBank accession no. M63484) when physical and catalytic characteristics of the family 2 ubiquitin hydrolase are compared with those of the pro-ISG15 processing enzyme. The peptide sequence obtained in Fig. 8 corresponds to FIG. 5. Phenyl-Sepharose resolution of the pro-ISG15 processing enzyme. Fractions containing peak 2 from the Mono Q HR 5/10 anion exchange column were pooled and resolved by Phenyl-Sepharose Fast Flow hydrophobic chromatography as described under "Materials and Methods." Panel A, aliquots of each fraction encompassing the peak of processing activity were resolved by SDS-PAGE on 10% (w/v) gels and visualized by Coomassie Blue staining to reveal a single band of 100 kDa. Panel B, enzyme activity (solid circles) is plotted versus the band density in arbitrary units (open circles) for the 100-kDa Coomassie-stained band of panel A. Panel C, plot of relative specific activity across the peak of 100-kDa processing activity. an insert within the otherwise conserved His box of the family 2 ubiquitin-specific proteases that is unique to yeast Ubp1 (21); therefore, it is probable that the pro-ISG15 processing enzyme corresponds to the human ortholog of yeast Ubp1 or a Ubp1related enzyme. This assignment is supported by the molecular weight correlation between Ubp1 and the processing enzyme, a pH rate profile corresponding to titration of a single group of pK a 8.1 presumably corresponding to the active site cysteine of the ubiquitin-specific proteases (Fig. 7), and sensitivity to inactivation by thiol selective alkylating reagents (Table II). Ubiquitin-specific proteases also contain a conserved histidine that serves as a general base catalyst to extract the proton from the nucleophilic cysteine during the catalytic cycle (16). Preliminary experiments show that pro-ISG15 processing is blocked by preincubation with diethylpyrocarbonate (data not shown), a histidine-specific modifying reagent. 5 Although the properties of the pro-ISG15 processing enzyme resemble those of another family 2 isopeptidase, isopeptidase T (22)(23)(24), an authentic sample of isopeptidase T was unable to process 125 Ipro-ISG15 even though it effectively cleaved free lysine 48linked polyubiquitin chains (data not shown). This observation, together with the absence of the partial sequence (Fig. 8) within isopeptidase T, precludes this ubiquitin-specific protease as the processing enzyme.
Yeast Ubp1 is a non-essential enzyme for which null mutants exhibit a decreased rate of ubiquitin/26 S proteasome-dependent degradation of ␤-galactosidase polyubiquitinated conjugates (21). As ISG15 is absent from yeast, the normal role for Ubp1 must be as an isopeptidase involved in polyubiquitin chain metabolism. Interestingly, yeast Ubp1 is not active against pro-ISG15 because the latter is not processed in cellfree yeast extracts or when transiently expressed from an appropriate plasmid. 5 This suggests that, if the processing enzyme is the human ortholog of yeast Ubp1, activity against pro-ISG15 must represent an adaptive divergence in specificity to accommodate a low affinity alternate substrate. Such adaptation of an existing activity may reflect the relatively recent emergence of ISG15-dependent regulation compared with the more ancient systems of SUMO-1 and NEDD8. Interestingly, Li and Hochstrasser (25) have very recently shown that processing of the SUMO-1 precursor in yeast occurs through a FIG. 6. Processing of pro-ISG15 occurs at the correct site. The trifluoroacetic acid-soluble supernatant from a processing activity similar to that of Fig. 2 was applied to a C18 reverse phase HPLC column and eluted with a linear acetonitrile gradient as described under "Materials and Methods." Panel A, elution profile for a control incubation in the absence of added pro-ISG15 (arrow denotes the elution position of the authentic carboxyl-terminal extension peptide). Panel B, elution profile for half of the processing reaction containing pro-ISG15. Panel C, elution profile for an aliquot of the processing reaction identical to that of panel B to which had been added 1 nmol of the carboxyl-terminal extension octapeptide. Insets to each panel represent enlargements of the regions in which the octapeptide extension elutes.   8. Partial sequence of the pro-ISG15 processing enzyme. A partial sequence for the pro-ISG15 processing enzyme was determined as described under "Materials and Methods." Sequence analysis reveals only to significant matches corresponding to the 100-kDa S. cerevisiae Ubp1 (21) and human KIAA0161 clone. specific enzyme apparently not required in the ubiquitin system. This interpretation is consistent with the low affinity of the processing enzyme for pro-ISG15 revealed by the linear concentration dependence for the latter.
Although substrate recognition by the processing enzyme is of relatively low affinity, it must require specific interactions because denatured pro-ISG15 is not a substrate ( Fig. 2 and Footnote 5). This conclusion is also supported by the observation that only one major activity, when adjusted for the stimulatory effect of ubiquitin, is active with pro-ISG15 despite the large number of constitutive UCH and Ubp isozymes present in this and other eukaryotic cell lines (16). The rate of ISG15 synthesis in interferon-␤-induced A549 cells can be estimated as 0.02 pmol/min per 10 6 cells based on earlier steady state quantitation of total ISG15 accumulation after 10 h of induction (10 pmol/10 6 cells) (5), a cell volume of 4 l/10 6 cells (26), and the reasonable assumption that ISG15 accumulation is linear. Therefore, the rate of pro-ISG15 processing in A549 extracts from Fig. 1 is Ͼ100-fold faster than the estimated rate of pro-ISG15 synthesis, consistent with the absence of pro-ISG15 accumulation within intact cells (3,4).
Several lines of evidence indicate that the processing enzyme cleaves the precursor at the correct Gly 157 -Gly 158 bond to generate active mature polypeptide and the carboxyl-terminal octapeptide extension. The product of this reaction, when resolved by SDS-PAGE, results in migration of ISG15 with a relative molecular weight identical to an authentic mature polypeptide standard (Figs. 1, 2, and 4). More important, the putative mature ISG15 product is capable of supporting ubiquitin activating enzyme-catalyzed ATP:PP i exchange (data not shown), unlike des-Gly-Gly-ISG15 in which the carboxyl-terminal glycine dipeptide is absent (6). In addition, the carboxylterminal extension peptide released during incubation with purified processing enzyme coelutes with an authentic synthetic peptide by C18 reverse phase HPLC (Fig. 6) and corresponds to the expected product by mass spectrometry (data not shown).
The most unexpected property of the pro-ISG15 processing enzyme is its marked stimulation by free ubiquitin (Fig. 4). Similar stimulation of a ubiquitin-specific protease by free ubiquitin has been observed for isopeptidase T, a 100-kDa family 2 ubiquitin-specific protease involved in the disassembly of free polyubiquitin chains released following degradation of the target protein by the 26 S proteasome (22)(23)(24). Detailed kinetic studies of isopeptidase T show that the effect of free ubiquitin is in stabilizing a catalytically competent conformation of the enzyme (23,24). It is possible that an analogous conformational effect accounts for the stimulation of pro-ISG15 processing, and we are currently expanding the kinetic analysis of pro-ISG15 processing to examine this question.
The present studies are the first to identify the cellular processing activity responsible for generating active ISG15 from the carboxyl-terminal blocked precursor. Identification of the pro-ISG15 processing enzyme as the human ortholog of yeast Ubp1 or a Ubp1-like protein indicates that with the emergence of the ISG15 regulatory pathway, the critical step of precursor processing was assumed by an existing ubiquitinspecific enzyme, perhaps by selective mutation of selected residues to accommodate the new low affinity alternate substrate.