Involvement of UBE1L in ISG15 Conjugation during Retinoid-induced Differentiation of Acute Promyelocytic Leukemia*

Acute promyelocytic leukemia (APL) cases expressing the t(15,17) product, promyelocytic leukemia (PML)/reti-noic acid receptor (cid:1) (RAR (cid:1) ), have clinical remissions through leukemic cell differentiation after all- trans- retinoic acid (RA) treatment. This differentiation therapy propelled interest in uncovering molecular mechanisms for RA-dependent APL differentiation. We previously identified the ubiquitin-activating enzyme-E1-like protein ( UBE1L ) as an RA-regulated target gene in APL that triggers PML/RAR (cid:1) degradation and apoptosis. This study reports that conjugation of the ubiquitin-like species, interferon-stimulated gene, 15-kDa protein (ISG15), also occurs during RA-induced APL differentiation. Knock-down of UBE1L expression inhibited this conjugation. RA treatment of APL and other RA-responsive leukemic cells induced expression of UBE1L and ISG15 as well as intracellular ISG15 conjugates. Notably, ISG15 conjugation did not occur in RA-resistant NB4-R1 APL cells. Induction of UBE1L and ISG15 along with ISG15 conjugation in

The majority of acute promyelocytic leukemia (APL, FAB M3) 1 cases contain the t (15,17) rearrangement that results in expression of the oncogenic translocation product PML/RAR␣, as reviewed (1). RA treatment leads to clinical remissions in APL through leukemic cell differentiation (2)(3)(4). Expression of PML/RAR␣ is linked directly to the etiology of APL as shown through transgenic APL models, as reviewed (5), and PML/ RAR␣ expression is predictive of response to pharmacologic dosages of RA (6). PML/RAR␣ expression disrupts cellular signaling that regulates proliferation and differentiation through several mechanisms. PML/RAR␣ expression causes PML nuclear body disorganization (7)(8)(9) that interferes with the ability of PML to regulate proliferation and apoptosis (10). PML/RAR␣ also inhibits retinoid signaling by acting as a dominant-negative retinoid nuclear receptor (5). In the absence of ligand, PML/RAR␣ can dimerize (11,12) with the retinoid X receptor and occupy retinoic acid response elements. These dimers complex with nuclear co-repressors and inhibit transcription of retinoid target genes (13)(14)(15). Exposure of APL cells to pharmacologic RA dosages leads to release of corepressors, recruitment of coactivators, and, in turn, transcriptional activation (15)(16)(17). This reconstitution of retinoid signaling as well as reorganization of PML nuclear bodies restores terminal differentiation response to APL cells.
Prior work using expression profiling identified candidate retinoid target genes in APL, as reviewed previously (5). Microarray and biochemical analyses uncovered UBE1L as a regulated species during RA-induced differentiation of APL cells (18,19). Studies revealed that UBE1L expression was rapidly increased by RA treatment of RA-sensitive but not -resistant NB4 APL cell lines (19). A domain of the UBE1L promoter conferred RA transcriptional response using retinoic acid receptor (RAR) and UBE1L reporter co-transfection assays, whereas PML/RAR␣ co-transfection inhibited this response (19). Ectopic UBE1L expression triggered PML/RAR␣ degradation, and retroviral transduction of UBE1L into NB4 APL cells rapidly triggered leukemic cell death even in the absence of RA treatment. These findings directly implicated UBE1L as a retinoid target gene in APL (19).
UBE1L was identified as the activating enzyme for the ubiquitin-like protein interferon-stimulated gene, 15 kDa (ISG15) (20). Activating enzymes play critical roles in conjugation of ubiquitin as well as other ubiquitin-like proteins, as reviewed (21). Sequence homology between activating enzymes supports the hypothesis that this family of enzymes, which includes the activating enzymes for ubiquitin, small ubiquitin-related modifier (SUMO), and NEDD8, engage similar mechanisms during conjugate activation. ISG15, originally identified as an interferon-induced species (22), contains tandem ubiquitin homology domains and a carboxyl-terminal LRGG motif conserved between ubiquitin-like protein modifiers (23). ISG15 forms covalent conjugates with intracellular proteins following diverse signals such as type I IFN treatment (24) and bacterial (25) or viral challenges (20). The recent observation that serpin 2a (26), phospholipase 1␥C, JAK1, ERK1, and STAT1 (27) are targets for ISG15 conjugation indicated a role for ISG15 in pathways critical for signaling diverse second messenger pathways or cellular responses (28).
The reported induction of UBE1L expression during RAmediated APL differentiation (19) indicated that retinoid treatment might also promote ISG15 conjugation. The aim of this study was to explore comprehensively the coordinate regulation and potential physical association of UBE1L and ISG15 in vivo and then to examine the relationship between induction of ISG15 conjugation and RA-mediated leukemic cell differentiation. Retinoid-sensitive and -resistant leukemic cells were independently examined. The direct involvement of UBE1L in ISG15 conjugation was studied through loss of function experiments using small inhibitory RNA (siRNA) as well as small hairpin RNA (shRNA) to target UBE1L expression.
The findings reported here established that siRNA-or shRNA-mediated repression of UBE1L inhibits induction of ISG15 conjugation by RA or IFN. A role for UBE1L and ISG15 conjugation beyond leukemia is consistent with results from mRNA cell and tissue arrays where UBE1L and ISG15 were often detected as expressed in the same human cells and tissues. Yet, repression of UBE1L mRNA was evident in several human cancer cell lines. As will be discussed, these results build upon and extend prior work implicating UBE1L as a tumor suppressor. Taken together, these findings implicate this previously unrecognized, retinoid induction of UBE1L and ISG15 conjugation as having important functional roles in hematopoietic and non-hematopoietic cells.
Semiquantitative RT-PCR Assays-Logarithmically growing NB4-S1 and NB4-R1 APL cells were independently seeded in the described media at a density of 3 ϫ 10 5 cells/ml. U937 and THP-1 cells were each seeded in logarithmic growth phase at a density of 1 ϫ 10 6 cells/ml. These cell lines were cultured with either 1 M RA (Sigma, St. Louis, MO) or Me 2 SO (Sigma) as a vehicle control. Total cellular RNA was extracted using the RNeasy Protect Mini Kit (Qiagen, Valencia, CA) or TRI Reagent (Molecular Research Center, Cincinnati, OH). Contaminating DNA was removed using a DNA-free kit according to the manufacturer's procedures (Ambion, Austin, TX). The cDNA was synthesized from 5 g of total RNA primed with random hexamers and using Superscript II reverse transcriptase (RT) (Invitrogen, Carlsbad, CA). PCR amplifications were performed with Taq polymerase (Invitrogen), and expression of each species was compared with ␤-actin or glyceraldehyde 3-phosphate dehydrogenase (GAPDH) expression that served as a control. To detect UBE1L, the forward primer was 5Ј-AGGTGGCCA-AGAACTTGGTT-3Ј and the reverse primer was 5Ј-CACCACCTGGAA-GTCCAACA-3Ј (19). For ISG15, the forward primer was 5Ј-GGTGGA-CAAATGCGACGAAC-3Ј and the reverse primer was 5Ј-ATGCTGGTG-GAGGCCCTTAG-3Ј. For ␤-actin, the forward primer was 5Ј-GCGGG-AAATCGTGCGTGACA-3Ј and the reverse primer was 5Ј-AAGGAAG-GCTGGAAGAGTGC-3Ј. For GAPDH, the forward primer was 5Ј-GGA-GGTGAAGGTCGGAGTCA-3Ј and the reverse primer was 5Ј-GACAA-GCTTCCCGTTCTCAG-3Ј. These independent PCR assay reaction products were each size-fractionated by electrophoresis and transferred onto nylon filters (Nytran SuPerCharge, Schleicher & Schuell Bioscience, Inc., Keene, NH), and then selectively hybridized with appropriate radiolabeled probes, as needed to enhance detection of some species.
Immunoblot Analyses-Lysates were size-fractionated using SDS-PAGE before transfer to nitrocellulose membranes (Schleicher & Schuell Bioscience, Inc., Keene, NH) using previously described techniques (19). Independent primary antibodies used were a polyclonal rabbit antibody that recognized the amino terminus of UBE1L (19), a polyclonal goat antibody that recognized ␤-actin (Santa Cruz Biotechnology, Inc., Santa Cruz, CA), or a murine monoclonal antibody (Berkley Antibody Co., Berkley, CA) that recognized the 6-histidine tag (His 6 ). The primary murine monoclonal antibody that recognized ISG15 was provided by Dr. E. Borden (Cleveland Clinic Foundation). Horseradish peroxidase-linked anti-rabbit and anti-mouse IgG were each purchased (Amersham Biosciences, Piscataway, NJ) as was the antigoat IgG (Santa Cruz Biotechnology).
Pharmacologic Treatments-NB4-S1 and NB4-R1 APL cells were independently seeded as already described before treatments at the indicated dosages with RA (Sigma), and the same culture conditions were used for experiments with 9-cis-retinoic acid (Sigma), phorbol ester (12-O-tetradecanoylphorbol-13-acetate; provided by Dr. M. Sporn, Dartmouth Medical School), rosiglitazone (provided by Dr. M. Sporn), fenretinide (4-HPR) (BIOMOL, Plymouth Meeting, PA), IFN A/D␣ (Sigma), the triterpenoid CDDO (provided by Dr. M. Sporn), or with Me 2 SO that served as a vehicle control. Cells were lysed in a modified radioimmunoprecipitation assay buffer, and protein concentrations were determined using the Bradford assay. Samples were then subjected to immunoblot analyses using previously described procedures (19).
Tissue Arrays and Tissue Bank-Multiple Tissue Expression Arrays (Clontech, Palo Alto, CA) and Multiple Tissue Northern membranes (Clontech) were individually prehybridized in ExpressHyb hybridization solution (Clontech) at 65°C for 30 min. Membranes were hybridized overnight in prehybridization buffer at 65°C with a dCT 32 P-radiolabeled 1-kb EcoRI-NcoI UBE1L probe or a radiolabeled probe that contained the full coding region of ISG15 and was isolated from the pcDNA4 vector with EcoRI-XbaI restriction endonuclease digestions. After hybridization, membranes were stringently washed in 0.1ϫ SSC and 0.1% SDS at 50°C. The Multiple Tissue Expression Array was hybridized to a radiolabeled probe to ubiquitin (Clontech) as a loading control.
To confirm and extend these results from the mRNA to protein level of expression, immunoblot analyses for UBE1L were performed on protein isolated from paired normal lung tissues and malignant lung tissues obtained from a lung tumor bank developed at Dartmouth-Hitchcock Medical Center. The institutional review board reviewed and approved these studies.
Purification of the ISG15⅐UBE1L Complex-BEAS-2B cells were transfected using Effectene Transfection Reagent (Qiagen, Valencia, CA) according to the manufacturer's instructions. Forty-eight hours after transfection, cells were lysed in a buffer containing: 50 mM NaH 2 PO 4 , 300 mM NaCl, 10 mM imidazole, 0.05% Tween 20 at pH 8. After three freeze-thaw cycles and passage through a 27-gauge syringe to reduce viscosity, lysates were mixed with Ni 2ϩ -agarose beads that were prewashed with lysis buffer and blocked for 2 h at room temperature in lysis buffer containing 5 mg/ml bovine albumin. After 2 h of additional incubation, beads were washed with lysis buffer to a final stringency of 20 mM imidazole. Beads were then eluted with SDS loading dye, under reducing conditions. Eluates were subjected to SDS-PAGE and proteins were transferred to nitrocellulose membranes (Nytran SuPerCharge, Schleicher & Schuell Bioscience, Inc., Keene, NH), as previously described (19). Immunoblot analyses were performed using a polyclonal antibody that recognized the amino terminus of UBE1L and monoclonal antibodies that recognized the His 6 tag or ISG15, respectively.
SiRNA Targeting of UBE1L-Synthetic siRNAs were purchased (Dharmacon Research, Inc., Lafayette, CO) as annealed oligonucleotides and were resuspended in RNase-free H 2 O. The siRNA duplexes used independently targeted luciferase GL2 as a scrambled control against an unexpressed sequence (siRNA-Control) (31); ␤-actin (siRNA-Actin) (32); or UBE1L using the sequence AACUUGUUCUGAUGGGU-GUG (siRNA-UBE1L). HeLa S3 cells were transfected with siRNA twice in 24-h intervals using Oligofectamine-based methods (Invitrogen) and previously established techniques (31). Total RNA was then isolated, and semi-quantitative RT-PCR assays were performed using the described independent primers for UBE1L, ISG15, ␤-actin, or GAPDH. To induce ISG15 conjugation after siRNA transfection, cells were treated with or without IFN A/D ␣ (Sigma) for 24 h. Immunoblot analyses were performed using the described polyclonal antibodies that independently recognized UBE1L or ␤-actin and the murine monoclonal antibody that recognized ISG15.
shRNA Targeting of UBE1L-The shRNA construct that targeted bases 935-954 of the UBE1L sequence was subcloned into the pSuper-Retro (Oligoengine, Seattle, WA) retroviral vector. A pSuper-Retro vector with a scrambled sequence was used as a control in these experiments (provided by Dr. J. Curtin, Dartmouth Medical School). These retroviral vectors were independently transfected into GP2-293 packaging cells (Clontech) using LipofectAMINE 2000 reagent (Invitrogen) and the manufacturer's procedures. Two days after transfection, media were used for viral transduction of NB4-S1 cells in the presence of Polybrene (8 g/ml) (Sigma). After 24 h, cells were then cultured in fresh media, and 72 h after transduction, cells were selected with puromycin (5 g/ml) (Sigma) for 72 h to obtain stable transductants. Reduced basal UBE1L expression was detected in transductants by immunoblot analyses for at least 3 weeks after puromycin selection (data not shown). These transductants were then independently treated with or without RA, IFN A/D ␣, or vehicle control.

Induction of UBE1L and ISG15 Expression-
The retinoid regulation of UBE1L and ISG15 expression was independently examined in RA-sensitive NB4-S1 cells and in RA-resistant NB4-R1 APL cells as well as in the U937 and THP-1 (data not shown) leukemic cell lines before and after RA treatment. As shown by RT-PCR analyses in Fig. 1, UBE1L and ISG15 mRNA expression increased markedly at 48 h of RA treatment of NB4-S1 APL cells. Induction of ISG15 mRNA was most evident from 24 (data not shown) to 48 h after RA treatment of NB4-S1, U937, and THP-1 (data not shown) leukemic cells. Notably, induction of ISG15 and UBE1L mRNAs was blunted despite RA treatment of NB4-R1 APL cells, as shown in Fig. 1. RA treatment of U937 cells increased UBE1L and ISG15 mRNA expression with kinetics similar to that detected for NB4-S1 cells (data not shown). Similar results were seen following RA treatment of THP-1 cells (data not shown). These findings implicated a role for these species in the retinoid response of leukemic cells.
Induction of ISG15 Conjugation-Coordinate regulation of UBE1L and ISG15 mRNA expression following RA treatment of NB4-S1 but not NB4-R1 cells suggested the involvement of ISG15 conjugation in the differentiation program of APL cells. To explore this further, intracellular ISG15 conjugation was examined before and after RA treatment of NB4-S1 cells. Results were compared with findings obtained in the RA-resistant NB4-R1 cell line. Immunoblot analyses revealed that, before RA treatment, ISG15 was not readily detected in these cells. In contrast, following RA treatment, both unconjugated and conjugated ISG15 were prominently induced by 48 h, as displayed in Fig. 2. The pattern of ISG15 conjugates following RA treatment was similar to that observed following IFN-induced conjugation (data not shown). Conjugation of ISG15 occurred after RA treatment along with induction of UBE1L and ISG15 mRNAs. Notably, unconjugated and conjugated ISG15 species were not appreciably induced following RA treatment of NB4-R1 cells as shown in Fig. 2. To examine the pharmacologic regulation of UBE1L expression and ISG15 conjugation, RAsensitive and -resistant NB4 cells were treated independently with diverse antiproliferative and differentiation-inducing agents, including 9-cis-retinoic acid, a retinoid X receptor (rexinoid) agonist LG100268 (data not shown), a non-classic retinoid (4-HPR), a triterpenoid (CDDO), a peroxisome proliferator-activated receptor ␥ agonist (rosiglitazone), phorbol ester (12-O-tetradecanoylphorbol-13-acetate), or IFN A/D ␣. Of these pharmacologic agents, only 9-cis-retinoic acid and IFN A/D ␣ prominently induced UBE1L, ISG15 expression, and ISG15 conjugation. Induction of UBE1L and ISG15 conjugation was not appreciably observed following 9-cis-retinoic acid treatments of NB4-R1 cells (data not shown). However, IFN A/D ␣ treatment still augmented UBE1L expression and ISG15 conjugation, despite RA resistance of NB4-R1 cells (data not shown).
ISG15 and UBE1L Expression Profiles-ISG15 and UBE1L mRNA expression profiles were examined in human tissues and cell lines using tissue expression arrays and independent radiolabeled probes for ISG15 and UBE1L (Fig. 3). ISG15 mRNAs were widely detected across hematopoietic and nonhematopoietic tissues and cell lines, with elevated levels detected in specific adult and fetal tissues. UBE1L mRNA expression was most prominent in the small bowel, lymph node, spleen, fetal spleen, thymus, and lung as compared with other examined tissues and cell lines. Yet, repression of UBE1L mRNA expression was detected in brain tissues and several examined cancer cell lines (HL-60, HeLa S3, K562, Daudi, and A549). As shown in these arrays, ISG15 and UBE1L mRNAs were often detected in the same hematopoietic and non-hematopoietic cells and tissues. Similar loading of mRNA species in this array was independently confirmed by hybridization to a radiolabeled ubiquitin probe (data not shown). These expression patterns for ISG15 and UBE1L mRNAs were independently confirmed using Northern analyses of multiple human tissues (data not shown).
The observed repression of UBE1L mRNA in cancer cell lines was consistent with previous work that indicated repressed UBE1L mRNA levels in lung cancer cell lines (33). This repressed UBE1L expression in lung cancer cell lines was intriguing, because UBE1L maps to a region of chromosome 3p that is often deleted in lung cancers. This prior work was extended by examining UBE1L protein expression in a lung tissue bank containing paired normal and malignant lung tissues from the same surgical resections. Notably, immunoblot analyses for UBE1L expression revealed repression of this species in two of three examined lung cancers as compared with adjacent normal lung tissues (data not shown). These findings are consistent with prior work that implicated a tumor suppressive role for UBE1L in carcinogenesis (33,34).
Physical Association of ISG15 and UBE1L in Vivo-The demonstration of an in vitro association between UBE1L and ISG15 that disappeared under reducing conditions (20) provided strong evidence for a potential in vivo E1-modifier interaction. To investigate this possibility, UBE1L was co-transfected into BEAS-2B cells along with wild-type His-tagged ISG15 (His-ISG15) or a mutant His-tagged ISG15⌬GG (His-ISG15⌬GG) species that lacked the conserved carboxyl glycine residues required for activation of ubiquitin and ubiquitin-like proteins. Consistent with prior in vitro findings (20), Ni 2ϩ pull-down of His-ISG15 revealed a physical association between His-ISG15 and UBE1L in vivo, as depicted in Fig. 4. In marked contrast, this association was not detected after attempted pull-downs of the mutant His-ISG15⌬GG species, although His-ISG15 and His-ISG15⌬GG expression levels were similar (data not shown). Unlike the targets of ISG15 conjugation seen in Fig. 2, this association was sensitive to reducing conditions (data not shown). Independent UBE1L pull-down experiments were performed with His-tagged ISG15 and confirmed that ISG15 post-translational modification of UBE1L was not detected (data not shown). Taken together, these findings establish an in vivo association between UBE1L and ISG15.
Inhibitory RNA-mediated Repression of UBE1L-To explore further the role of UBE1L in ISG15 conjugation, UBE1L expression was knocked-down through use of siRNA. These experiments were independently conducted in the absence and presence of IFN A/D ␣ treatment. Because this pharmacologic agent prominently induced conjugation in HeLa S3 cells, immunoblot analyses were performed in these cells to assess intracellular ISG15 conjugation as well as to determine the functional role of UBE1L in this conjugation. The siRNA engineered to target only UBE1L reduced UBE1L mRNA expression without affecting actin, GAPDH, or ISG15 expression, as shown in Fig. 5. Targeting actin or transfection with a control siRNA sequence (pGL2-siRNA) did not affect basal UBE1L or ISG15 mRNA expression. These findings confirmed the ability of siRNA to target UBE1L selectively in vivo. IFN A/D ␣ treatment still induced UBE1L expression and ISG15 conjugation following transfection of a control siRNA, as shown in Fig. 5. Transfection of siRNA that preferentially targeted UBE1L inhibited IFN A/D ␣ induction of UBE1L and intracellular ISG15 conjugation, as confirmed in Fig. 5. These experimental results

FIG. 1. All-trans-retinoic acid (RA) induction of UBE1L and ISG15 expression occurs during terminal leukemic cell differentiation and is deregulated in RA-resistant APL cells.
A, semiquantitative RT-PCR analyses for UBE1L, ISG15, and actin expression for RA-treated and untreated NB4-S1 (left panel) and NB4-R1 (right panel) cells. These cells were exposed to 1 M RA (ϩ) or vehicle (Ϫ) for 48 h before harvesting for RT-PCR analyses. B, densitometric analyses of results displayed in A. The intensities of RA-mediated UBE1L and ISG15 induction as compared with vehicle-treated levels in the indicated APL cell lines were measured and normalized to Actin levels. C, RA-induced UBE1L and ISG15 expression in the non-APL U937 leukemic cell line. The semi-quantitative RT-PCR analysis of RA-treated versus untreated U937 cells is displayed. Cells were exposed to 1 M RA (ϩ) or vehicle (Ϫ) for 48 h before this analysis. Densitometric analyses of UBE1L and ISG15 expression in RA-treated as compared with untreated U937 cells was not displayed, because expression of these species was undetected in untreated cells.

FIG. 2. Induction of ISG15 conjugation and UBE1L expression by specific pharmacologic agents in differentiation sensitive but not resistant APL cells.
A, immunoblot analysis of RA-induced UBE1L expression and ISG15 conjugation in NB4-S1 cells. NB4-S1 cells were treated with 10 M RA or vehicle control for 0, 6, 24, or 48 h before immunoblot analysis. Similar results were observed for 1 M RA (data not shown). ISG15 conjugation is depicted in this figure as is unconjugated ISG15 (arrow). B, specific agents induced UBE1L expression and ISG15 conjugation in NB4 cells. NB4-S1 and NB4-R1 cells were The induction (ϩ) of UBE1L and ISG15 expression as well as the presence of ISG15 conjugates (ϩ) in NB4-S1 cells are depicted. Immunoblot analyses were performed to assess induction of UBE1L and ISG15 expression as well as the extent of ISG15 conjugation. C, induction of UBE1L expression and ISG15 conjugation by RA treatment was deregulated in RA-resistant NB4-R1 cells. NB4-R1 cells were treated with or without RA (10 M) for 48 h before examination. Immunoblot analyses were performed to assess expression of UBE1L and unconjugated ISG15 (arrow) as well as to determine the extent of ISG15 conjugation. Molecular weight markers are depicted. established the in vivo involvement of UBE1L in ISG15 conjugation following IFN A/D ␣ treatment.
To explore the impact of targeting UBE1L expression in APL cells, NB4-S1 cells were transduced with pSuper-retro vectors containing either an shRNA that targeted UBE1L expression (shRNA-UBE1L) or a control sequence (shRNA-Control). The results are displayed in Fig. 6. Consistent with the results in HeLa cells, IFN A/D ␣-mediated induction of ISG15 conjugation was markedly inhibited in shRNA-UBE1L-transduced NB4-S1 cells, as compared with shRNA-Control-transduced NB4-S1 APL cells, as shown in Fig. 6. As expected, UBE1L knock-down decreased IFN A/D ␣-induced conjugation of  ). B, densitometric analyses of relative expression levels for UBE1L and ISG15. Values were determined by densitometric analyses using ImageQuaNT (version 5.1). The differences between the highest and lowest values for each species were determined. These differences were divided equally and used to determine relative levels of expression spanning from "Low" to "High" for each indicated species. UBE1L and ISG15 expression profiles for the indicated tissues and cell lines were assessed.
ISG15. Levels of unconjugated ISG15 increased as compared with the control. The repression of UBE1L also inhibited RAinduced ISG15 conjugation. Knock-down of UBE1L levels had a greater inhibitory effect on IFN-induced ISG15 conjugation than RA-induced conjugation. This might be due to differences in the kinetics of ISG15 conjugation following treatment with these agents. In NB4-S1 cells, RA treatment induced ISG15 conjugation between 24 and 48 h after treatment. This conju- gation occurred rapidly and was associated with increased UBE1L levels, as shown in Fig. 2. The rapid and prominent induction of ISG15 conjugation following RA treatment was less susceptible to inhibition by UBE1L knock-down as was evident following IFN A/D ␣ treatment. In these targeting experiments, UBE1L expression and ISG15 conjugation were repressed but not extinguished. Accordingly, it was not surprising that RA-mediated NB4 APL cellular differentiation and PML/RAR␣ levels were not appreciably affected (data not shown). Taken together, these findings confirmed in APL cells that UBE1L was required for ISG15 conjugation in vivo following IFN or RA treatments.

DISCUSSION
This study reports that RA augments UBE1L and ISG15 expression as well as ISG15 conjugation during leukemic cell differentiation. ISG15 conjugation precedes terminal RA-induced differentiation of APL cells and is deregulated in RAresistant APL cells, implicating a direct role for this conjugation in the APL differentiation program. We previously reported that UBE1L, the activating enzyme for ISG15, is an RA-target gene in APL (19). Prior work established an association between UBE1L and ISG15 in vitro (20). That work was extended in this study by showing that an in vivo association also occurs between UBE1L and ISG15. This physical interaction requires the carboxyl-terminal glycines of ISG15, as shown in Fig. 4. The coordinate retinoid regulation of UBE1L and ISG15 mRNAs precedes (data not shown) induced ISG15 conjugation observed in APL cells. A role for UBE1L in ISG15 conjugation was found following retinoid or interferon treatments, as shown in Figs. 5 and 6. These findings likely have implications for physiologic and pharmacologic responses in hematopoietic and non-hematopoietic cells. ISG15 and UBE1L expression was found to overlap in multiple human tissues, but UBE1L was repressed, especially in several cancer cell lines Fig. 3. Taken together, these studies reveal the interaction of previously unrecognized retinoid-regulated proteins in the APL differentiation program. These findings also establish a direct role for UBE1L in ISG15 conjugation and reveal the consequences of reducing UBE1L expression on ISG15 conjugation.
Activation of ISG15 conjugation during retinoid-induced APL differentiation is a novel functional context for these regulated species. ISG15 conjugation has been implicated in the FIG. 4. ISG15 and UBE1L associate in vivo; this interaction required the carboxyl-terminal glycines of ISG15. A, a schematic diagram of His-ISG15 and His-ISG15⌬GG ( ϩ NH 2 , amino terminus; 6-his, histidine tag; Gly-Gly-COO Ϫ , diglycine motif; COOH Ϫ , carboxyl terminus). The His-ISG15⌬GG lacks the carboxyl-terminal diglycine motif conserved between conjugated, ubiquitin-like proteins. B, cotransfection of UBE1L and either His-ISG15 or His-ISG15⌬GG was followed by lysis and pull-down with Ni 2ϩ -agarose beads. Immunoblot analyses revealed that His-ISG15 interacts with UBE1L, whereas His-ISG15⌬GG does not. Co-transfection of pcDNA-his (His-pcDNA) with pSG5-UBE1L followed by incubation with Ni 2ϩ -agarose beads did not pull-down UBE1L protein, as shown in this figure.
FIG. 5. Targeting UBE1L inhibits IFN A/D ␣-induced ISG15 conjugation. A, HeLa S3 cells were independently transfected twice in 24-h intervals with siRNA targeting UBE1L (siRNA-UBE1L), ␤-actin (siRNA-Actin), or control siRNA (siRNA-Control) before semi-quantitative RT-PCR analyses were performed. As expected, targeting UBE1L and ␤-actin specifically knocked-down UBE1L and ␤-actin mRNA expression, respectively, but not GAPDH expression. Scrambled siRNA for an unexpressed species served as a control and did not affect expression of UBE1L or ␤-actin. B, targeting UBE1L mRNA inhibited IFN A/D ␣-induced ISG15 conjugation in HeLa S3 cells. HeLa S3 cells were transfected with siRNA-UBE1L or siRNA-Control before treatment with IFN A/D ␣ (100 units/ml, 24 h). Immunoblot analyses were performed to assess UBE1L, ISG15, and actin expression patterns as well as to determine the degree of ISG15 conjugation. The arrow depicts the position of unconjugated ISG15. C, densitometric analysis of results displayed in panel B. The intensities of UBE1L expression (left panel) and ISG15 conjugation (right panel) were measured and normalized relative to actin expression, which was unaffected by these targeting experiments. IFN A/D ␣-dependent ISG15 conjugation was diminished by siRNA-UBE1L transfection as compared with the siRNA-Control transfection experiment. A decrease in IFN A/D ␣-induction of UBE1L expression was also observed after siRNA-UBE1L transfection as compared with the displayed siRNA-Control transfection experiment. Molecular weight markers are depicted. cellular stress response (25), IFN response (24), and in response to viral infection (20). Induction of ISG15 conjugation in these settings implicates this conjugation process in these cellular responses. This report reveals that ISG15 conjugation also occurs during retinoid-induced leukemic cell differentiation and thereby implicates a role for this pathway in retinoid response in settings other than those previously described.
The cellular consequences of ISG15 conjugation are not fully understood. Biochemically related protein conjugation processes involve ubiquitination and sumoylation of target proteins. In the case of polyubiquitination, it is known that this modification can target the modified protein for proteasomal degradation. For ISG15 conjugation that is induced following retinoid treatment, this modification might function in a manner analogous to that observed for polyubiquitination. It has been shown that UBE1L triggered PML/RAR␣ degradation (19) and that the proteasome can affect ISG15 conjugates (35). Yet, the ISG15 conjugates that have been uncovered appear to be stabilized after modification (26,28). This is different than what has been observed following polyubiquitination indicating that ISG15 modification might serve functions that are different from those engaged by ubiquitination or sumoylation. Consistent with this view, recent evidence indicated a role for ISG15 conjugation in prolonging IFN signaling and promoting apoptosis (28). Finding that the signaling molecules STAT1, JAK1, and ERK1 are targets of ISG15 conjugation (28) indicated a potential link between ISG15 conjugation and cellular signaling pathways.
Prior work offers insights into the function of ISG15 conjugation in retinoid-induced leukemic cell differentiation. There is considerable prior evidence linking IFN and retinoid signaling in APL, as reviewed (36). Treatment of APL cells with RA and IFN promoted differentiation response of these cells (37). Perhaps this sensitization is accomplished by promoting induction of UBE1L and ISG15 proteins with a concomitant increase in ISG15 conjugation to critical signaling proteins. It is of interest that STAT1 was identified as a target of ISG15 conjugation (27). Targeting STAT1 inhibits RA-induced differentiation of U937 cells (38). Whether conjugation of ISG15 to STAT1 plays a mechanistic role in RA-mediated differentiation of promyelocytic leukemia is not yet known. The precise biologic actions engaged by retinoid-induced ISG15 conjugation during APL differentiation are the subject of future work. Whether PML/RAR␣ serves as a direct target for ISG15 conjugation should be determined.
The roles of UBE1L in physiology or pathophysiology are under active study. Prior work provides potential insights into UBE1L functions. The basal expression of UBE1L reported here in diverse human tissues and cells implicates physiologic roles for UBE1L in both hematopoietic and non-hematopoietic cells or tissues. In this regard, UBE1L was originally identified as a candidate tumor suppressor gene (33) with reduced expression in lung cancers (33,34). The chromosome 3p21 location of UBE1L is intriguing, because this region is often deleted in preneoplastic or neoplastic epithelial tissues (39). This finding raised the prospect that UBE1L silencing was important in multistep carcinogenesis. This prior work was confirmed and extended in this study. Prior work that showed repression of UBE1L mRNA in some lung cancer cell lines (33) was extended by showing a similar repression of UBE1L protein occurs in subsets of examined lung cancers as compared with adjacent normal lung tissues. These findings point to a potential tumorsuppressive role for UBE1L in lung cancers and perhaps other malignancies. The settings in which UBE1L induction occurs FIG. 6. Targeting UBE1L in NB4-S1 APL cells inhibits ISG15 conjugation following all-trans-retinoic acid (RA) treatment. NB4-S1 APL cells were independently transduced with a retroviral vector targeting UBE1L (shRNA-UBE1L) or a control vector (shRNA-Control). A, Western analyses confirmed marked repression of basal UBE1L protein levels in shRNA-UBE1L-transduced NB4-S1 cells (left panel) with densitometric analysis of UBE1L repression depicted (right panel). B, targeting UBE1L mRNA inhibited IFN A/D ␣and RA-induced ISG15 conjugation. NB4-S1 cells were treated independently with IFN A/D ␣-(100 units/ml) or RA (0.1 M) for 36 h. Immunoblot analyses were performed to assess independently UBE1L and ISG15 expression patterns. The position of unconjugated ISG15 is depicted by the arrow. Molecular weight markers are depicted. offer a view into the function of UBE1L. For example, this report found that UBE1L and ISG15 induction and ISG15 conjugation occurred early during retinoid-induced leukemic cell differentiation. Whether regulation of UBE1L is involved in other differentiation programs or in tumor suppression of epithelial cells should be investigated.
ISG15 expression alone is not sufficient for induction of conjugation. This report demonstrated the coordinate retinoid regulation of ISG15 and UBE1L and established the involvement of UBE1L in ISG15 conjugation. Because UBE1L repression can occur during epithelial carcinogenesis, this would in turn disrupt ISG15 conjugation and affect the functional consequences of this conjugation. Whether repression of UBE1L in specific epithelial cancers affects ISG15 conjugation in these settings is not yet known. Other pathophysiologic states are known to affect UBE1L function. For instance, the NS1B protein of type B influenza alters UBE1L function (20) by inhibiting the physical association between ISG15 and UBE1L and thereby disrupting ISG15 conjugation.
The coordinate induction of UBE1L and ISG15 following retinoid treatment, as shown in this study, strongly implicated a direct role for this association in regulating ISG15 conjugation. In turn, the disruption of UBE1L expression or activity would be expected to inhibit ISG15 conjugation, as established in this study. It is notable that silencing UBE1L would disrupt formation of ISG15 conjugates, thereby repressing this conjugation process. This was confirmed in the knock-down studies displayed in Figs. 5 and 6. It is not unexpected that repression of UBE1L by shRNA did not abrogate RA-dependent differentiation response in APL cells, because expression of UBE1L was not extinguished by this targeting. Perhaps the dual repression of UBE1L and ISG15 by this knock-down strategy would affect RA-response of APL cells.
Aside from an alteration in UBE1L expression or activity, other enzymes affect ISG15 conjugation. Notably, the ubiquitin-specific protease, 43 kDa (UBP43 or USP18) can remove ISG15 from target proteins (40). Loss of UBP43 was associated with increased ISG15 conjugation (40), sensitivity to IFN treatment (28), and even the onset of a neurologic disorder (41). Consistent with the findings reported here, UBP43 overexpression was previously shown to affect leukemic cell differentiation (42). E2 and E3 proteins are important in the conjugation of ubiquitin and other ubiquitin-like species. The E2 and E3 enzymes for ISG15 have not been reported, but these proteins could contribute to the regulation of ISG15 conjugation in diverse biologic settings.
In summary, this study uncovered that ISG15 conjugation and coordinate retinoid regulation of ISG15 and UBE1L occur during APL differentiation. UBE1L was also shown to be required for ISG15 conjugation in vivo. ISG15 and UBE1L were found to associate physically in vivo. This work extends prior work in this field by revealing that ISG15 conjugation is a marker of effective retinoid-induced APL differentiation, because RA-resistant APL cells do not undergo this modification. It is notable that regulation of ISG15 conjugation occurred after treatment with specific pharmacologic agents, indicating that this is a distinct program engaged by relatively few pharmacologic agents. Future work should determine the precise biologic role of ISG15 conjugation in retinoid-induced leukemic cell differentiation. Perhaps this previously unrecognized retinoid-induced conjugation process serves an important biologic or therapeutic role beyond leukemia. This view is consistent with the frequent finding of UBE1L and ISG15 expression in diverse cellular and tissue contexts. A potential tumor-suppressive role for UBE1L was also evident from the observed repression of UBE1L in certain cancer cell lines and cancers.
Whether pharmacologically engaging ISG15 conjugation through use of retinoids or other agents has therapeutic benefits should be determined in the future. In this regard, an agent that inhibits UBP43 activity should promote ISG15 conjugation and thereby confer the biologic consequences of this conjugation.