Interferon Regulatory Factor 1 Mediates the Interferon- g Induction of the Human Immunoproteasome Subunit Multicatalytic Endopeptidase Complex-like 1*

Proteasomes generate antigenic peptides from intra-cellular proteins for presentation to the immune system by the major histocompatibility complex class I molecules. The antiviral cytokine IFN- g alters the catalytic specificity of proteasomes by inducing the synthesis of an alternative set of three proteolytically active proteasome subunits. We have analyzed the mechanism of IFN- g induction for the IFN- g -induced subunit multicatalytic endopeptidase complex-like 1 (MECL1). The human MECL1 promoter contains two interferon-stim-ulated response elements (ISREs), generally known to bind members of the interferon regulatory factor (IRF) family. The importance of these elements for IFN- g induction of MECL1 was addressed by transfecting an endothelial cell line with MECL1 promoter constructs. By deletions and mutations of the ISRE sequences, we demonstrated that both ISREs were needed for full IFN- g induction of the reporter gene. The second (down-stream) ISRE was essential for both IFN- g -induced and basal transcriptional activity of the promoter. In electrophoretic mobility shift assays, anti-IRF-1 antibodies supershifted an IFN- g -induced protein binding specifically to both ISRE sequences, whereas IRF-2 bound the second ISRE before induction. Co-transfection of IRF-1 resulted in induced MECL1 promoter activity in the absence of IFN- g . These data indicate that the IFN- g induction of human MECL1 is mediated by IFN- g -induced IRF-1.

Antigenic peptides are continuously generated from intracellular proteins and presented to the immune system by major histocompatibility complex class I molecules on the cell surface (1). Proteasomes have a central role in antigen processing as the multicatalytic complex that degrades the bulk of cytosolic proteins (2)(3)(4). IFN-␥, a cytokine up-regulating the antigen presentation during viral infection, induces the synthesis of an alternative set of proteolytically active proteasome subunits from a low constitutive level (5). The three IFN-␥-induced subunits, low molecular mass polypeptide (LMP) 1 7, LMP2 and multicatalytic endopeptidase complex-like 1 (MECL1) (also called LMP10), replace three active constitutive subunits, X, Y, and Z, respectively, forming immunoproteasomes (6 -13). Incorporation of the IFN-␥-induced subunits modifies the peptidase activity of the proteasome (10, 14 -17) in a way that may favor generation of antigenic peptides (18 -20).
IFN-␥ induction of genes is mainly mediated directly by signal transducers and activators of transcription (Stat1), a rapidly activated transcription factor binding IFN-␥ activated sequence (21), or by Stat1-induced interferon regulatory factor 1 (IRF-1), a member of the IRF family known to bind interferon-stimulated response elements (ISREs) (22,23). Conserved ISRE sequences have been found in the promoter of all the three IFN-␥-induced proteasome subunit genes (24 -27).
The aim of the present study was to examine the role of the ISRE sequences for constitutive and IFN-␥-induced expression of the human proteasome subunit MECL1 and to investigate the transcription factor mediating the IFN-␥ induction.

EXPERIMENTAL PROCEDURES
Cell Lines-The human endothelial cell line ECV304 (ATCC) was grown in medium 199 with 10% fetal calf serum (both from Life Technologies, Inc.). The human B cell lines Reh (acute lymphoblastic lymphoma; pre-B cell) and IM9 (multiple myeloma; IgG secreting) (ATCC) were grown in RPMI 1640 medium (Life Technologies, Inc.) with 10% fetal calf serum.
Promoter Constructs-The genomic region encompassing the MECL1 gene (also called PSMB10) has previously been cloned and sequenced by our group (25) (GenBank TM accession no. X71874). Using a subclone containing the MECL1 promoter as template, promoter fragments were amplified by Pfu DNA polymerase (Stratagene) in polymerase chain reactions (PCRs) with the forward primers 2051, 2162, 2180, 2219, and 2235, respectively, and the reverse primer 2334rev (Fig. 1). The PCR products were cloned into the SmaI site of the luciferase reporter vector pGL3 Basic (Promega), and the sequence was confirmed by DNA sequencing.
Constructs with mutated ISREs (28,29) were made using 2162mut and 2219mut forward primers (see Table I for sequences) and 2334rev primer in PCR as described above. The resulting PCR products were either cloned directly into pGL3 Basic or used as reverse primers in new PCRs to incorporate the mutations into larger constructs. The mutations were confirmed by DNA sequencing.
Transfections and Detection of Luciferase Reporter Activity-ECV304 cells (60 -90% confluent) were transfected in 35-mm wells using 6 l of LipofectAMINE reagent (Life Technologies, Inc.) and 1.1 g of DNA (1 g of vector construct and 0.1 g of pRL-TK vector (Promega) as internal standard). In co-transfection experiments, 0.33 g of each vector construct was used. Empty pcDNAI vector was added to a total amount of 1 g of DNA when none or just one IRF expression vector was used. The cells were transfected as recommended, using 5 h of incubation of cells with DNA-LipofectAMINE complexes in serum-free medium and replacement of the transfection mixture with fresh complete medium 22 h after the start of transfection. The cells were harvested 7 h later.
The effects of IRF-1 and/or IRF-2 on MECL1 expression were studied in transfection experiments using the respective cDNA in pcDNAI, kindly provided by Prof. T. Maniatis.
For investigating the time course of induction, IFN-␥ was added to the cell medium at a final concentration of 200 units/ml at 1, 2, 3, 4, 5, 6, 7, 8, and 24 h prior to cell lysis. When indicated in other transfections, the cells were always stimulated with 200 units/ml IFN-␥ for 24 h before harvesting.
The cells were lysed and assayed for firefly and Renilla (sea pansy) luciferase activities using the Dual-Luciferase reporter assay system (Promega). The luciferase activities were determined as recommended for the Dual-Luciferase system using a Lumat LB 9507 (Berthold) luminometer with manual mixing of reagents or an automatic MicroLu-matPlus luminometer (EG&G Berthold). The firefly luciferase activity was normalized to the Renilla luciferase activity of the same sample, and the mean was calculated from the parallels. From the mean values of each independent run, the overall mean and its standard error (S.E.) were calculated.
Northern Blot-ECV304 cells were stimulated with 200 units/ml IFN-␥ for different time periods (1, 2, 3, 4, 6, 8, and 24 h) prior to lysis. mRNA was isolated directly from the cells using magnetic oligo(dT) Dynabeads (Dynal) and the buffers recommended by the manufacturers. Cells were grown to confluence in 9-cm plates and lysed directly in 1 ml of lysis buffer. Cell extract from each plate was run through a 23 gauge syringe to shear the DNA and mixed with 300 g beads. Isolated mRNA was run on a denaturing agarose gel and blotted. The filter was probed with MECL1 cDNA (27) and ␤-actin.
Electrophoretic Mobility Shift Assay (EMSA)-Nuclear extracts were prepared by the method of Andrews and Faller (30) with minor modifications. Nuclear proteins were extracted from unstimulated cells and from cells stimulated with 200 units/ml IFN-␥ for 24 h. When investigating the time course of induction, IFN-␥ was added to the cell medium at a final concentration of 200 units/ml at 0.5, 1, 2, 4, and 8 h prior to cell lysis. Cycloheximide was added to a final concentration of 50 g/ml, 15 min before addition of IFN-␥. Cells were washed in cold phosphatebuffered saline, and the ECV304 cells were scraped into phosphatebuffered saline. Pellets containing 1-5 ϫ 10 7 cells were resuspended in 400 l of Buffer A (10 mM Hepes, pH 7.9, at 0°C, 10 mM KCl, 1.5 mM MgCl 2 , 2 mM dithiothreitol, 0.2 mM phenylmethylsulfonyl fluoride, 1 g/ml aprotinin, 1 g/ml antipain, 1 g/ml chymostatin, 2.5 g/ml leupeptin, and 0.25 g/ml phosphoramidon). The cells were incubated for 5-15 min on ice. Pelleted nuclei were resuspended in 50 l of Buffer B (10 mM Hepes, pH 7.9, at 0°C, 420 mM NaCl, 1.5 mM MgCl 2 , 2 mM dithiothreitol, 0.2 mM EDTA, and protease inhibitors as for Buffer A) and incubated for 20 -30 min on ice with regular mixing. The nuclear lysates were cleared by centrifugation, frozen in aliquots in liquid nitrogen, and stored at Ϫ80°C.
Two double-stranded oligonucleotides from the MECL1 promoter, oligo A, containing ISRE1, and oligo E, containing ISRE2 (see Table I), were end-labeled with [␥-32 P]ATP using T4 polynucleotide kinase and purified on polyacrylamide gels. EMSA reactions were performed in volumes of 20 l with a final EMSA buffer of 10 mM Hepes, pH 7.9, at 0°C, 40 mM KCl, 1 mM dithiothreitol, 1 mM EDTA, and 5% glycerol. Nuclear extracts (15-20 g of protein) were incubated in EMSA buffer with 1 l of labeled oligonucleotide (20,000 cpm; 5-10 fmol) for 15 min at room temperature. In competition and supershift reactions, 1 l of competitors (2 pmol) (Table I) or 1 l of antibodies (1 g, or 1:10 dilution of antisera in phosphate-buffered saline) were incubated with the extract in EMSA buffer for 15 min at room temperature before the labeled oligonucleotide was added. Polyclonal antibodies against Stat1, IRF-1, IRF-2, IFN-stimulated gene factor-3␥, IFN consensus sequence-binding protein, and IFN consensus sequence-binding protein in adult T-cell leukemia cell lines or activated T cells were obtained from Santa Cruz Biotechnology. Rabbit antisera against IRF-3 and IRF-7 were kindly provided by Dr. Takahashi Fujita (The Tokyo Metropolitan Institute of Medical Science) and Dr. Luwen Zhang (University of North Carolina), respectively. The reactions were run on 4% polyacrylamide gels in 0.5ϫ TBE (45 mM Tris-borate, 1 mM EDTA, pH 8.0) for 80 min at 100 V. The gels were vacuum dried, and the signals detected by phosphor screen autoradiography.

RESULTS
A Lag Period in the IFN-␥ Induction of the MECL1 Promoter-The human MECL1/PSMB10 promoter (25,27) and the promoter of the murine MECL1 gene homologue Lmp10/ Psmb10 (26,31) show high sequence similarity in a 250-base pair region upstream of the translational start site, above which the murine sequence contains a transposable element (31). We therefore limited our analysis of the human promoter to this region of homology, and started at position 2051 of the genomic fragment (25) (Fig. 1). A 284-base pair promoter fragment was cloned into a luciferase reporter vector. In order to determine the IFN-␥ responsiveness of this construct, called 2051, we transfected ECV304 cells, in which MECL1 is known to be IFN-␥ inducible (27). The time course of IFN-␥ induction was examined by stimulation for different time periods.
The promoter segment proved to contain the elements necessary for IFN-␥ induction of the reporter gene (Fig. 2). A lag period of around 3 h was observed before the IFN-␥ stimulation led to a rise in luciferase activity. Five-fold induction was reached after 7 h of stimulation by IFN-␥ (Fig. 2). Stimulation periods between 8 and 24 h and between 24 and 60 h resulted in small fluctuations of luciferase activity above the level of 5-fold induction (results not shown).
To compare the time course of the IFN-␥ induction of the reporter construct to the induction of MECL1 mRNA, ECV304 cells were stimulated for different time periods, and their mRNA was analyzed by Northern blotting. The amount of MECL1 mRNA increases after 3 h of stimulation ( Fig. 2B), similar to the lag period observed for induction of the reporter construct ( Fig. 2A). Thus, the time course of promoter induction parallels that of the MECL1 mRNA induction, suggesting that elements in the promoter segment chosen for analysis can account for the induction of MECL1 mRNA. The increase of MECL1 mRNA seen from 8 to 24 h of stimulation ( Fig. 2B) was not accompanied by an increase of the expression of the reporter, possibly because the luciferase mRNA and/or protein had a shorter half-life and then did not accumulate.
The Two ISREs Are Important for IFN-␥ Inducibility-The presence of a lag period suggested that the induction was not due to the phosphorylation-activated transcription factor Stat1 directly, but rather mediated by a factor downstream of Stat1 in the IFN-␥ signaling pathway, such as IRF-1. In the promoter region selected for study, there are two ISRE sequences with the consensus binding site for members of the IRF family (23). These elements are conserved in the murine Psmb10 promoter and have been shown by deletions to be involved in IFN-␥ induction of murine MECL1 (31). To determine the importance of the two ISRE sequences in regulation of human MECL1, 5Ј deletion constructs were made based on the positions of the ISREs (Fig. 3). The constructs were named 2162, 2180, 2219, and 2235 after their starting point in the sequence (see Fig. 1). The potential of the promoter to function in the opposite direction was tested by including a construct with the 2051 fragment in reverse orientation (2051rev). The luciferase activity of  (27) is numbered according to the genomic sequence (25). The two ISRE sequences are boxed. The primers used for generating the promoter constructs are marked by long arrows and numbered after the position of their 5Ј end. Small arrows indicate the previously mapped transcriptional start sites (27).
The IFN-␥-induced activity of the reverse promoter, 2051rev, was higher than the 2051 construct, suggesting that the involved promoter elements can up-regulate transcription in both directions. The fragment starting at the first ISRE (2162) had 77% IFN-␥ inducibility relative to the 2051 construct. This reduction might reflect an upstream element contributing to the induction, or the sequence upstream of ISRE1 may be important for optimal binding to the element. Upon deletion of the ISRE1 element (2180), the induction was reduced to 47% of maximum, clearly indicating a role for ISRE1 in the IFN-␥ induction. In the next construct (2219), where the segment between the ISREs was deleted, 38% of the induction was still maintained. Interestingly, when the ISRE2 element was deleted (2235), the luciferase activity in IFN-␥-stimulated cells dropped to the background level.
The basal level of transcription in unstimulated cells transfected with the 2051 construct was 17% of the IFN-␥-induced level, confirming the 5-6-fold induction seen after 24 h of IFN-␥ stimulation (see Fig. 2). The constitutive level of luciferase activity was unaffected by deletions until deletion of ISRE2 (construct 2235), when the constitutive activity, like the IFN-␥-induced activity, fell to the level of vector alone.
Dual Role for ISRE2 in Regulation of Transcriptional Activity-The results obtained with the deletion constructs of the MECL1 promoter suggested a role for ISRE2 both in IFN-␥induced and basal transcriptional activity, whereas ISRE1 seemed to enhance the level of IFN-␥ induction and not the level of constitutive transcription. To confirm these results, we examined in transfection experiments the role of the two ISREs by mutating bases essential for IRF binding (23), while keeping the length of the constructs intact. In constructs 2051 and 2162, the first element, the second element, and both were mutated, and in construct 2219, ISRE2 was mutated (Fig. 4).
When ISRE1 was mutated, the induction was reduced to 38% in 2051 m1 and 35% in 2162 m1 (Fig. 4), close to the level of induction for the construct 2219 devoid of promoter sequence upstream of ISRE2 (28% in Fig. 4; 38% in Fig. 3), confirming the importance of ISRE1 for IFN-␥ induction. Comparing the 2162 construct with 2162 m1, ISRE1 seemed to confer a 2-fold induction upon stimulation with IFN-␥. Moreover, the similar transcriptional activities of the 2051 m1 and 2162 m1 constructs suggested that a possible third element upstream of ISRE1, conceivably accounting for the difference in activity between the 2051 and 2162 constructs (see Figs. 3 and 4), would depend on ISRE1 function for its contribution. For the basal transcriptional activity of the constructs, mutation of ISRE1 had only minor effects.
Construct 2219, containing ISRE2 only, showed 2-fold induction upon IFN-␥ stimulation (Figs. 3 and 4). In contrast to mutation of ISRE1, mutation of ISRE2 essentially abolished all transcription, both IFN-␥-induced and constitutive, regardless of the length of the construct. Similarly, constructs in which both ISRE1 and ISRE2 were mutated showed no transcriptional activity above the level of vector alone. Therefore, whereas ISRE1 seemed to be necessary for maximal IFN-␥ induction, ISRE2 appeared to have a dual role, mediating both basal and IFN-␥-induced transcription.

IRF-1 Binds Both ISREs in Extracts from IFN-␥-stimulated ECV304 Cells-
The transfection experiments showed that the two ISREs in the MECL1 promoter were instrumental for the IFN-␥ induction of MECL1 in the endothelial cell line. The transcription factors binding these elements are therefore likely to mediate the induction of MECL1. To characterize these proteins, oligonucleotide probes containing ISRE1 and ISRE2 were made (see Table I) and incubated with nuclear extracts from unstimulated and IFN-␥-stimulated ECV304 cells (Fig. 5).
In extracts from IFN-␥-stimulated ECV304 cells, the ISRE1 probe (labeled oligo A) gave one strong band and two weaker bands (Fig. 5, upper panel, left). The specificity of the binding was examined by including various cold competitor oligonucleotides in excess (see Table I). All the complexes formed with  probe A were competed out by the corresponding cold oligo A and were not competed out by the mutated oligonucleotide (oligo B), indicating that all the bound proteins needed the unmutated sequence to bind the probe (see Table I for the sequences). A shorter oligonucleotide containing ISRE1, oligo C, competed out the two lower complexes, the lowermost not completely. Again, mutation resulted in no competition (oligo D). Interestingly, oligo E containing ISRE2 competed out the strong lower complex and competed better than its short form, oligo G. The same complex was competed out by the consensus ISRE (32, 33) (oligo I) but not by the mutated consensus ISRE (32, 33) (oligo J). Therefore, the protein component of the lower complex seemed to bind specifically to the ISRE sequence and to require several bases upstream of the consensus sequence for efficient binding. The upper complex appeared to be specific for the long ISRE1 oligo and may represent proteins binding outside the ISRE sequence, although they seemed to depend on an intact ISRE sequence for binding.
The ISRE2 probe (labeled oligo E) gave two protein complexes with the nuclear extract from IFN-␥-stimulated ECV304 cells (Fig. 5, upper panel, right). The competition pattern was similar to that seen for ISRE1. Both complexes were competed out by the cold oligo E, but not with the mutated version (oligo F). The upper complex was not competed out by the shorter ISRE2 oligo G and therefore seemed to be specific for the longer ISRE2 oligonucleotide. Both ISRE1 (oligo A) and consensus ISRE (oligo I) competed for the lower complex, which thereby appeared to be specific for the ISRE sequence. Neither probe gave any shift in the absence of protein extract (Fig. 5, upper  panel, left and right).  Table I for  probe sequences. Interestingly, in nuclear extracts from unstimulated ECV304 cells (Fig. 5, middle panel), the shifted complexes were similar to the complexes of the IFN-␥-stimulated cells. However, the signal of the lower complex of both probes (middle panel, left and right) was weaker than observed for stimulated cells (middle panel, left and right), indicating that this nuclear protein was induced by IFN-␥. Again, the competition patterns showed the lower complex of both probes to be ISRE-specific. Thus, the results of the competition analyses suggested that members of the ISRE-binding IRF family specifically bound the two ISRE probes both in the presence and absence of IFN-␥ stimulation.
To examine the identity of the proteins bound to the elements, we incubated the extracts with antibodies to different proteins associated with the IFN-␥ signaling pathway (Fig. 5,  lower panel). We tested the following transcription factors: Stat1, the first transcription factor of the IFN-␥ signaling pathway (21); IRF-1, the ISRE-binding transcription factor downstream of Stat1 in the IFN-␥-pathway (34); IRF-2, a constitutive factor suggested to compete with IRF-1 for binding ISRE (34); and IFN-stimulated gene factor-3␥ (p48), a subunit involved in the IFN-␣ pathway through complex formation with Stat1 and Stat2 (34,35) but recently also shown to bind ISRE in complex with Stat1 homodimer (36). To investigate the possible binding of other IRF family members to the elements, we included antibodies against IFN consensus sequence-binding protein, IFN consensus sequence-binding protein in adult Tcell leukemia cell lines or activated T cells, IRF-3, and IRF-7 (34).
The anti-IRF-1 antibodies were the only antibodies giving supershift with the extract of IFN-␥-stimulated ECV304 cells (Fig. 5, lower panel, left and right, and results not shown). For both ISRE1 and ISRE2, the anti-IRF-1 antibodies supershifted the lower complex shown by competitors to be ISRE-specific. This indicated that the IFN-␥-inducible transcription factor IRF-1 mediated the observed IFN-␥ induction of the MECL1 promoter, by binding to both ISREs.
With nuclear extract from unstimulated ECV304 cells, no supershift could be seen for the ISRE1 probe by the antibodies tested (Fig. 5, lower panel, left, and results not shown). Interestingly, the anti-IRF-2 antibodies gave a reproducible supershift of the lower complex for ISRE2, which was not seen for any of the other antibodies tested (lower panel, right, and results not shown). The transfection experiments revealed an important role of ISRE2 in constitutive expression. Thus, the results suggested that IRF-2, or an IRF-2-like protein crossreacting with the antibodies, might be involved in the regulation of the basal activity of the MECL1 promoter. In response to IFN-␥, this protein might be replaced by IRF-1, increasing the transcriptional activity of the MECL1 promoter.
To investigate whether the induction of the ISRE-binding protein precedes induction of the MECL1 promoter, nuclear extracts were prepared from cells stimulated with IFN-␥ for different time periods. Shorter stimulation periods resulted in formation of the same complexes as observed for 24 h stimulation (Fig. 6). For both the ISRE1 probe and the ISRE2 probe the amount of the lower complex increased after 2 h of stimulation (Fig. 6), thus preceding the MECL1 promoter induction, which showed a lag period of 3 h (see Fig. 2). To examine whether the IFN-␥-induction of the complexes were dependent on new protein synthesis, cells were treated with cycloheximide prior to stimulation. For both ISRE1 and ISRE2, the IFN-␥-induced increase of the lower complex was inhibited (Fig. 6, lanes 8c), showing that IFN-␥-induced protein binding of the two MECL1 ISRE sequences is dependent on newly synthesized IRF-1. Cycloheximide also inhibited the induction of MECL1 mRNA (results not shown).

IRF-2 Is the Major ISRE-binding Protein in B Cell
Extracts-We have previously shown that MECL1 is expressed at a high constitutive level in B cell lines and not significantly further induced by IFN-␥ (27). The high basal level of MECL1 mRNA and protein might be caused by constitutively active Stat1 observed in some B cell lines (37). It was therefore interesting to investigate whether IRF-1 was present in extracts from unstimulated B cells and could shift the ISRE1 and ISRE2 probes. The two B cell lines IM9 and Reh were used, and both were shown to have a high basal level of MECL1 mRNA (27). Nuclear extracts from both stimulated and unstimulated cells were analyzed by EMSA in the same manner as for the ECV304 cells. Interestingly, the band pattern found for the two B cell lines (Fig. 7) was very similar to the pattern observed for ECV304 cells (Fig. 5).
Moreover, the competition pattern demonstrated that the ISRE1 and ISRE2 probes were bound by ISRE-specific proteins in the B cell extracts (Fig. 7, A and B, upper and middle panels; for details of interpretation, see Fig. 5). However, in contrast to the ECV304 nuclear extracts, the intensity of the complexes did not increase upon IFN-␥ stimulation (Fig. 7, A and B, middle versus upper panels). The most striking finding was that the ISRE-specific complex of both oligo A (ISRE1) and oligo E (ISRE2) was supershifted with anti-IRF-2 antibodies rather than with anti-IRF-1 antibodies, both in stimulated and unstimulated cells. The ISRE-specific complex was not supershifted with any of the other antibodies tested (Fig. 7 and results not shown). Therefore, rather than constitutively activated IRF-1, the B cell extracts contained constitutively present IRF-2, or an IRF-2-like protein, binding the ISRE sequence of both probes, similar to the ISRE2-binding protein observed in nuclear extracts from unstimulated ECV304 cells.
To compare the amount of IRF-1 and IRF-2 in the cell lines, nuclear extracts from IFN-␥-stimulated and unstimulated ECV304 and IM9 cells were analyzed for IRF-1 and IRF-2 by immunoblotting (Fig. 8). Whereas the amount and induction of IRF-1 were similar in the two cell types, IM9 nuclear extract contained much more IRF-2 than ECV304 extract. The higher level of IRF-2 in the IM9 nuclear extracts may explain why IRF-2 occupies the ISRE probes in these extracts also after IFN-␥ stimulation of the cells. 9). Co-transfecting the 2051 construct with IRF-1 demonstrated clearly that IRF-1 induces the MECL1 promoter in the absence of IFN-␥ (Fig. 9, top left). The level of induction was comparable to the level achieved by IFN-␥ itself. Similar results were obtained for the other promoter constructs containing functional ISRE2 element (Fig. 9, left and middle). Mutation of ISRE1 in the 2051 m1 construct reduced the response with/without IFN-␥ to about 50 -60% (Fig. 9, top middle), and mutation of ISRE2 abolished the response completely ( Fig. 9, top right), confirming that ISRE2 is essential for induction of the MECL1 promoter. When cells co-transfected with IRF-1 were treated with IFN-␥, the promoter activity of the inducible constructs increased somewhat further, suggesting that maximum induction was not reached under either of these conditions. In these series of transfections, the level of induction was lower than observed previously (Figs. 2-4). There was a generally lower level of reporter gene activity in the present experiments, resulting in a high relative background level (Fig. 9, right graphs).
Co-transfection of IRF-2 expression vector with the promoter constructs resulted in a decrease of promoter activity in unstimulated cells and no significant change in activity of IFN-␥-stimulated cells (Fig. 9). When the expression vectors for IRF-1 and IRF-2 were combined, IRF-2 did not inhibit the effect of IRF-1. Thus, in ECV304 cells, IRF-1 could mimic the effect of IFN-␥ on MECL1 promoter activity, and the promoter induction was not affected by overexpressed IRF-2.

DISCUSSION
The results presented here suggest that IRF-1 mediates the IFN-␥ induction of MECL1 in the ECV304 endothelial cell line by binding the two ISREs of the promoter. Five lines of evidence support this conclusion. First, the lag period in the IFN-␥ induction seen for both the reporter gene, for MECL1 mRNA, and the EMSA complexes pointed to an inducible factor downstream of the rapidly activated Stat1 in the IFN-␥ signaling pathway. Second, the two ISREs were required for IFN-␥ induction, as demonstrated in transfection experiments both by sequentially deleting the elements and by mutating the ISREbases generally known to be required for IRF-1 binding (23,38). The importance of correct ISRE sequence for binding nuclear proteins was also confirmed in EMSA, in which mutated oligonucleotides were unable to compete for binding. Third, the amount of the ISRE-specific protein binding the ISRE1 and ISRE2 probes in EMSA increased upon stimulation with IFN-␥, and the increase was inhibited by cycloheximide, consistent with a role for the IFN-␥-inducible IRF-1. Fourth, the anti-IRF-1 antibodies supershifted the ISRE-specific complex for both ISRE probes in nuclear extract from IFN-␥-stimulated endothelial cells. Fifth, co-transfection with IRF-1 mimicked the IFN-␥ induction of the MECL1 promoter.
The ISRE2 element seemed to have an additional essential role for constitutive transcription, because mutation of ISRE2 abolished both constitutive and IFN-␥-induced expression. The supershift results suggested that ISRE2 was bound by IRF-2 in unstimulated endothelial cells. The same was observed for the B cells, where an abundant anti-IRF-2-reacting protein was binding both ISRE1 and ISRE2. The supershift with anti-IRF-2 antibodies was surprising because IRF-2 is commonly regarded as being a repressor (34). Interestingly, IRF-2 also has a latent activation domain (39) and has been reported to function as a transcriptional activator (40,41). However, overexpressed IRF-2 decreased the basal promoter activity but did not essentially affect the level of induction. It may therefore be that IRF-2 in endothelial cells confers a low basal transcriptional activity to the promoter, which is increased upon IFN-␥ stimulation, leading to the replacement of IRF-2 by IRF-1.
The human MECL1 promoter was studied as a genomic fragment cloned into a reporter vector. Therefore, the results obtained do not necessarily reflect the regulation of MECL1 in vivo. We have previously shown that MECL1 protein is expressed at a low constitutive level in the ECV304 cells, highly induced by IFN-␥ stimulation (27), and in the present report, we also demonstrate the time course of induction for MECL1 mRNA. The transfection results presented are consistent with the observed MECL1 expression pattern, suggesting that the regulation of MECL1 can be explained by the elements confined to the selected promoter region.
The two ISREs are conserved in the mouse MECL1 promoter (26,31). Our results are consistent with the results reported for the murine MECL1 promoter (31) in that a promoter construct devoid of functional ISREs shows neither IFN-␥-induced nor constitutive transcriptional activity. However, the particular dependence on ISRE2 reported here for the human promoter was not observed in mouse (31). Rather, the two ISREs in mouse seemed to have similar roles in that each could be deleted without blocking either the constitutive or the IFN-␥induced transcriptional activity of the other (31). The need for two ISREs for full IFN-␥ induction of MECL1, as observed for both murine (31) and human MECL1 promoter constructs, is supported by the fact that both ISRE sequences have been conserved through evolution. In contrast to MECL1, the two other IFN-␥-inducible proteasome subunits, LMP2 and LMP7, have single ISRE sequences in their promoters (24). The ISRE sequence in the LMP2 promoter has been demonstrated to bind IRF-1 and to be essential for IFN-␥ induction (42).
In conclusion, our data indicate that the IFN-␥ induction of human MECL1 is mediated by IRF-1 and that IRF-2 plays a role in constitutive transcription of MECL1. The involvement of IRF-1 in the induction of both LMP2 and MECL1 suggests a concerted regulation of these immunoproteasome subunits at the transcriptional level.