Constitutive and Cytokine-induced Expression of the Melanoma Growth Stimulatory Activity/GROα Gene Requires Both NF-κB and Novel Constitutive Factors

Melanoma growth stimulatory activity (MGSA)/growth regulated (GRO) and interleukin-8 (IL-8) are highly related chemokines that have a causal role in melanoma progression. Expression of these chemokines is similar in that both require the NF-κB element and additional regions such as the CAAT/enhancer binding protein (C/EBP) element of the IL-8 promoter. The constitutive and cytokine IL-1-induced promoter activity of the chemokine MGSA/GROα in normal retinal pigment epithelial and the Hs294T melanoma cells is partially regulated through the NF-κB element, which binds both NF-κB p50 and RelA (NF-κB p65) homodimers and heterodimers. Mutational analysis of the MGSA/GROα promoter reveals that, in addition to the NF-κB element, the immediate upstream region (IUR) is necessary for basal expression in retinal pigment epithelial and Hs294T cells. Gel mobility shift and UV cross-linking analyses demonstrate that several constitutive DNA binding proteins interact with the IUR. Although this region has sequence similarity to the several transcription factor elements including C/EBP, the IUR includes sequences that have no similarity to previously identified enhancer regions. Furthermore, RelA transactivates through either the NF-κB element or the IUR, suggesting a putative interaction between NF-κB and this novel complex.


Melanoma growth stimulatory activity (MGSA)/ growth regulated (GRO) and interleukin-8 (IL-8) are highly related chemokines that have a causal role in melanoma progression. Expression of these chemokines is similar in that both require the NF-B element and additional regions such as the CAAT/enhancer binding protein (C/EBP) element of the IL-8 promoter. The constitutive and cytokine IL-1-induced promoter activity of the chemokine MGSA/GRO␣ in normal retinal pigment epithelial and the Hs294T melanoma cells is partially regulated through the NF-B element, which binds both NF-B p50 and RelA (NF-B p65) homodimers and heterodimers. Mutational analysis of the MGSA/GRO␣ promoter reveals that, in addition to the NF-B element, the immediate upstream region (IUR) is necessary for basal expression in retinal pigment epithelial and Hs294T cells. Gel mobility shift and UV cross-linking analyses
demonstrate that several constitutive DNA binding proteins interact with the IUR. Although this region has sequence similarity to the several transcription factor elements including C/EBP, the IUR includes sequences that have no similarity to previously identified enhancer regions. Furthermore, RelA transactivates through either the NF-B element or the IUR, suggesting a putative interaction between NF-B and this novel complex.
The human IL-8 and MGSA/GRO genes contain an NF-B element within their enhancer regions that has been shown to be necessary for transcriptional activation of these chemokines (35)(36)(37). Furthermore, IL-8 gene regulation also requires the C/EBP element adjacent to the NF-B element (36). The C/EBP and NF-B complexes directly interact or cross-couple to further enhance IL-8 gene transcription (38 -40). Although the chemokine MGSA/GRO has an essential role in inflammation and tumor progression, regulation of MGSA/GRO gene expression is not as well understood as the closely related IL-8 gene.
In this work, we demonstrate that in addition to the NF-B element, the immediate upstream region (IUR) is necessary for basal and cytokine induced expression of MGSA/GRO␣ in RPE cells. Likewise, basal MGSA/GRO␣ promoter activity within the Hs294T melanoma cells also requires this region. However, unlike the IL-8 promoter, neither C/EBP␣ nor C/EBP␤ recognize the similar region in the MGSA/GRO␣ promoter. Furthermore, RelA (NF-B p65) transactivates MGSA/GRO␣ transcription either directly through the NF-B element or indirectly through the adjacent IUR. We have identified a novel complex that is constitutively bound to the IUR in both normal RPE and melanoma cells. We propose that transcriptional regulation of the MGSA/GRO␣ gene involves multiple factors that recognize the NF-B and immediate surrounding regions.

MATERIALS AND METHODS
Northern Blot Analysis-Hs294T and RPE cells were cultured as described by Shattuck et al. (35). Total RNA was purified as described previously (35). Random primed 700-base pair EcoRI fragment of MGSA/GRO␣ cDNA (16) and a 400-base pair EcoRI fragment of IL-8 cDNA (41) were used as probes. Hybridization of cyclophilin (1B15) was used as a standard for quantitation.
CAT Reporter Gene Plasmid Construction-MGSA␣350/CAT and mutant NF-B MGSA␣350/CAT constructs were described earlier (35). Mutation of the IUR within the context of the MGSA␣350/CAT constructs was achieved by recombinant polymerase chain reaction using polymerase chain reaction primers containing the mutations underlined 5Ј-GGGATCGACCTGGGTCTCCG-3Ј. Site-directed mutagenesis was utilized to generate an additional mutation within the IUR element (IUR-B) (see Table I) using the Altered Sites mutagenesis system (Promega). The expected point mutations were confirmed by restriction enzyme digestion and sequencing. MGSA␣(Ϫ97/Ϫ62)TATA/CAT, MGSA␣ 2xIUR, and MGSA␣ m.2xIUR were generated by ligating the respective annealed oligonucleotides (see Table I) into the cloning region in the parental TATA/CAT vector (39).
Expression Vector Constructs and Recombinant Protein-The RelA expression vector was as described previously (42) and was the gift of Warner Greene (University of California, San Francisco). NF-IL6 cDNA was as described previously (43) and was the gift of Tadamitsu Kishimoto (Osaka University, Osaka, Japan). Recombinant C/EBP␤ and NF-IL6 were produced utilizing pRSET vectors as described previously (44) and were the generous gifts of Linda Sealy (Vanderbilt University).
Transfection and CAT Reporter Assay-Either RPE or Hs294T cells were co-transfected with 10 g of the indicated MGSA/CAT fusion genes (for point mutations see Table I) and 2 g of pCMVhGH (obtained from Dr. Lynn Matrisian, Vanderbilt University), which allowed for normalization of transfection efficiency by measuring growth hormone secretion utilizing the immunoassay (Nichols Institute). Transfections were performed by the calcium phosphate coprecipitation method (45). CAT enzymatic activity was assayed as described previously (46). The percent [ 14 C]chloramphenicol converted to acetylated forms was determined by PhosphorImage analysis (Molecular Dynamics).
Radiolabeled and Competitor DNA-Oligonucleotides were synthesized on a Milligen 7500 DNA synthesizer (Diabetes Research DNA Core, Vanderbilt University). Equal amounts of each oligonucleotide and its complement were annealed in STE (10 mM Tris, pH 7.8, 1 mM EDTA, pH 8.0, 200 mM NaCl) by boiling the oligonucleotides in a water bath that was slowly cooled to room temperature (approximately 4 h). The oligonucleotides (coding strand) are shown in Table I. Probes for gel mobility shift analyses were prepared by radiolabeling 100 ng of annealed oligonucleotides with T4 polynucleotide kinase.
Nuclear Extracts and DNA Binding Assay-Nuclear extracts were prepared from Hs294T and RPE cells as described previously (35) with the exception that cell lysis was performed by vortexing vigorously in the presence of buffer A with 1% Nonidet P-40. DNA binding reactions with recombinant C/EBP␤ and NF-IL6 proteins were performed by combining the indicated amounts of recombinant proteins in 20 l of 19 mM Hepes, pH 8.0, 1 mM Tris, pH 8.0, 50 mM NaCl, 10 mM KCl, 0.18 mM EDTA, pH 8.0, 1 mM spermidine, 10% glycerol, 0.5% Triton X-100, 0.1% Nonidet P-40, 0.4 g of BSA, 10.2 mM dithiothreitol, and 1 g poly(dI-dC)⅐poly(dI-dC) for 20 min at 37°C prior to probe addition (40,000 counts/reaction) for 20 min at room temperature. The binding reaction was then analyzed by electrophoresis in a nondenaturing 6% polyacrylamide gel in 0.25 ϫ TBE (22 mM Tris, pH 8.0, 22 mM boric acid, 0.5 mM EDTA, pH 8.0). DNA binding reactions with Hs294T and RPE nuclear extracts were performed by incubating 5-g nuclear extracts in 10 mM Tris, pH 8.0, 50 mM NaCl, 5% glycerol, and 1 g of poly(dI-dC)⅐poly(dI-dC) for 15 min prior to probe addition (20,000 counts) for 20 min at room temperature. The resulting protein-DNA complexes were separated on 6% polyacrylamide gel in 0.5 ϫ TBE (45 mM Tris, pH 8.0, 45 mM boric acid, 1 mM EDTA, pH 8.0). Jurkat T lymphocytes were transfected with a RelA expression vector or parental pCMV4 vector alone by electroporation as described previously (47). Whole cell extracts were prepared from Jurkat T-cell transfectants by high salt extraction as described previously (47). Gel shift assays were performed as above with 10 g of Jurkat extract.
UV Cross-linking-Nuclear extracts (5 g) incubated with the labeled MGSA␣ 2xIUR oligonucleotide (20 l total reaction volume) were exposed to short wave UV radiation for 15 min (Stratalinker). Half of the reaction (10 l) was separated on a 6% nondenaturing gel as described for the DNA binding assay. The remaining 10 l was heated (95°C) for 5 min in SDS loading buffer (50 mM Tris, pH 6.8, 2% SDS, 10% glycerol, 1% ␤-mercaptoethanol, 0.1% bromphenol blue). Labeled proteins were then separated by electrophoresis on a 9% SDS-polyacrylamide gel and compared with molecular weight standards.
Antibodies-Antibodies to NF-B p50, RelA and preimmune antisera (42,48) were the kind gift of Warner Greene. In addition, antisera to RelA, NF-B p50, NF-B p52, and C/EBP␤ were obtained from Santa Cruz. Additional antibodies directed against C/EBP␣, C/EBP␤, and C/EBP␦ were the generous gift of Stephen McKnight and were as described previously (49). Anti-leader binding protein (LBP) antisera was generously provided by Robert Roeder (50).

Regulation of MGSA/GRO and IL-8 mRNA Synthesis
by IL-1-MGSA/GRO and IL-8 mRNA levels increased rapidly in normal RPE cells stimulated with IL-1 ( Fig. 1 and Ref. 35). This induction of MGSA/GRO was primarily due to increased transcription (35). The Hs294T melanoma cells have a constitutive level of both MGSA/GRO and IL-8 gene expression (Fig.  1). The basal level of activity of both MGSA/GRO and IL-8 in the Hs294T cells was equal to or greater than the IL-1-induced level in the RPE cells. We had previously demonstrated by promoter deletion studies that the MGSA/GRO␣ region between Ϫ100 and Ϫ43 from the transcription start site were necessary for basal expression in Hs294T melanoma cells and basal and cytokine-induced expression in RPE cells (35). Further analysis demonstrated that the NF-B element within this region was necessary for this activation.
Mukaida et al. (36) had shown that both the NF-B and the adjacent C/EBP binding elements were required for IL-1 and tumor necrosis factor ␣ activation of IL-8 in a fibrosarcoma cell line. Further work demonstrated that C/EBP proteins bound to the IL-8 promoter and that NF-B directly interacted with the

Sequences of oligonucleotides used in gel mobility shift assays
The NF-IL-6 oligonucleotide represents Ϫ168 to Ϫ137 from IL-6 promoter, which contains the NF-IL-6 element (43). Wild-type IL-8 and MGSA/GRO␣ sequences are shown. For each mutated oligonucleotide, nucleotides similar to MGSA/GRO␣ are indicated by dots, and mutant nucleotides are shown. Mutations in the MGSA/GRO␣ NF-B element include mNF-B and mIUR ϩ mB. Mutations in the MGSA/GRO␣ IUR include mIUR, mIUR-B, and mIUR-C. Wild-type MGSA␣ 2ϫ IUR and respective mutant (MT 1-4) oligonucleotides contain two repeats of the underlined region in MGSA/GRO␣Ϫ97/Ϫ62. Expression of the MGSA/GRO␣ Gene C/EBP complexes (38,39). Sequence analysis between the IL-8 and MGSA/GRO␣ promoter region indicated an almost identical NF-B element and adjacent C/EBP-like region (Fig. 2). The MGSA/GRO␣ nucleotide sequence adjacent to the NF-B element contains several nucleotides that are conserved between previously identified C/EBP enhancers and are essential for C/EBP binding and subsequent transactivation for the IL-8, IL-6, albumin gene (DE1), and the serum amyloid A genes (36,43,51,52). The MGSA/GRO region also has close homology with the human immunodeficiency virus type I (HIV-1) LBP-1 binding site, which is necessary in addition to the NF-B and Sp1 for full transcriptional activation of the HIV-1 long terminal repeat (53-56) (Fig. 2). This sequence was of interest since we have recently demonstrated that basal MGSA/GRO␣ promoter activity required both NF-Band Sp1-related complexes bound within the immediate promoter (57). Immediate Upstream Region Is Required for MGSA/GRO␣ Promoter Activity-To determine if the IUR located adjacent to the NF-B element was involved in MGSA/GRO␣ regulation as demonstrated for IL-8 gene regulation (36,38,58), we studied the activities of MGSA␣350/CAT constructs with point mutations in both the NF-B and upstream region in transiently transfected RPE cells. Mutations in either site resulted in a substantial loss of basal CAT activity (Ͼ50%) and subsequent loss of IL-1 induction in RPE cells (Fig. 3A). Mutations made in both regions resulted in a complete loss of CAT activity comparable with the activity obtained from the parental vector pPLFCAT, indicating a complete loss of activation through the MGSA/GRO␣ promoter region (Fig. 3A). Likewise, within the Hs294T melanoma cell line, mutation of either the NF-B or the IUR resulted in a substantial loss in the base-line CAT activity (Fig. 3B). These results indicated that, in addition to the consensus NF-B element, the IUR element plays an important role in MGSA/GRO␣ gene regulation in both RPE and Hs294T cells.

5Ј
Characterization of the Nuclear Proteins That Bind to the MGSA/GRO␣-97/-62 Region-Gel mobility shift analyses were performed to characterize the nuclear factors that bind to the IUR and NF-B region. Nuclear extracts from RPE cells either untreated or IL-1 treated were incubated with either the wild type (WT), mutant IUR (m.IUR), mutant NF-B (m.B) or double mutant (m.IURϩm.B) MGSA␣Ϫ97/Ϫ62 probe (Fig. 4). IL-1 induction resulted in the appearance of two shifted nuclear complexes bound to the wild-type MGSA␣Ϫ97/Ϫ62 oligonucleotide (lanes 1 and 2). Addition of various NF-B antisera indicated that the lower complex consisted of the NF-B p50/ RelA heterodimer, while the upper complex contained RelA (presumably homodimers) (lanes 4 and 5). Addition of preimmune, NF-B p52 and C/EBP␤ antisera had no effect on the shifted complexes (lanes 3, 5, and 7). Gel mobility shift analyses with labeled mutant IUR (m.IUR) MGSA␣Ϫ97/Ϫ62 oligonucleotide demonstrated the identical pattern of shifted complexes as wild-type oligonucleotide (lanes 8 and 9). Labeled mutant NF-B (m.B) or double mutant (m.IURϩm.B) MGSA␣Ϫ97/Ϫ62 oligonucleotides did not retard nuclear complexes (lanes 10 -13). These data demonstrated that the nuclear complexes bound to the MGSA/GRO␣ Ϫ97/Ϫ62 region are specific for the NF-B element and not the IUR.
Characterization of MGSA/GRO␣ C/EBP-like Sequence as a Potential C/EBP Binding Element-Several groups have demonstrated cooperative binding of C/EBP family members with RelA to the IL-8 C/EBP enhancer region (38,39). However, IL-8 has a weak binding affinity for C/EBP proteins. Using the IL-8 or MGSA/GRO␣ C/EBP-like regions as probes, we have not been able to identify a C/EBP-related complex from RPE or Hs294T nuclear extracts, although C/EBP proteins were present in these extracts based on immunoblot and gel shift analysis with the NF-IL6 consensus oligonucleotide as a probe (data not shown). Purified C/EBP␤ or NF-IL6 did not bind to the C/EBP-like region in the MGSA/GRO␣ promoter, although it did bind weakly to the IL-8 C/EBP enhancer region (data not shown). We addressed the possibility that the affinity of C/EBP␤ for the MGSA/GRO␣ enhancer region was too weak to detect an interaction after electrophoresis during gel shift analysis. A converse approach was to analyze the ability of the IL-8 and MGSA/GRO␣ regions to compete for C/EBP␤ protein bound to the labeled NF-IL6 consensus element (Fig. 5). Unlabeled NF-IL6 oligonucleotide at 5-fold excess to radiolabeled NF-IL6 probe reduced C/EBP␤ binding by Ͼ90% (Fig. 5). With 50-fold excess NF-IL6 oligonucleotide, the C/EBP␤ complex was completely removed. Unlabeled IL-8Ϫ101/Ϫ63 oligonucleotide reduced C/EBP␤ binding to the NF-IL6 oligonucleotide by Ͼ70% at 200-fold excess, while addition of higher concentrations of this oligonucleotide completely removed C/EBP␤ binding. In contrast, addition of unlabeled MGSA␣Ϫ97/Ϫ62 did not reduce C/EBP␤ binding even at 2000-fold excess (Fig. 5C). Our data agreed with earlier studies in which purified C/EBP␤ bound weakly to the IL-8 enhancer region as compared to the NF-IL-6 element (38,39). Moreover, these results indicated that although there were nucleotide similarities between the MGSA␣Ϫ97/Ϫ62 region and previously identified C/EBP consensus elements, the MGSA/GRO␣ region adjacent to the NF-B element did not bind C/EBP␤ proteins.
Identification of IUR-bound Complexes by Gel Mobility Shift Analysis-Since the majority of the nuclear complexes that recognized the MGSA/GRO␣ Ϫ97/Ϫ62 region were NF-B-related, a 42-base pair probe was prepared that contained two copies of the IUR without the NF-B element present (designated MGSA␣ 2xIUR). When this probe was used in gel mobility shift assays with nuclear extracts from either RPE or Hs294T cells, two complexes bound (IUR-F) (Fig. 6, A and B). Cytoplasmic extracts demonstrated the presence of a slower migrating complex that appeared to be nonspecific in that all MGSA/GRO␣ oligonucleotides tested including wild-type and mutants removed this complex in competition analysis. Unlabeled competitor DNAs were used to test the specificity of the nuclear IUR-F complexes bound to the labeled MGSA␣ 2xIUR. An oligonucleotide from the IL-8 promoter containing the C/EBP and NF-B elements did not compete (Fig. 6A, lane 6), complexes were identical to those that demonstrated a loss of basal and cytokine-induced MGSA/GRO␣ promoter activity when placed in the MGSA␣350/CAT constructs (Fig. 3). Although the IUR sequences were similar to several consensus DNA binding elements including C/EBP and LBP sites, antisera to these transcription factors had no effect on the bound complexes (Fig. 6B, lanes 1-5). In addition, the consensus elements for LBP and NF-IL6 did not remove the bound complexes (Fig. 6B, lanes 11 and 12).
Further analysis of the IUR sequence was performed by creating additional point mutations in the MGSA/GRO␣ Ϫ93 to Ϫ77 region present in the oligonucleotide MGSA␣ 2xIUR. Several mutant IUR oligonucleotides (MT 1, MT 2, MT 3) (lanes 7-9) did not compete, while the WT and MT 4 effectively re-moved the bound complexes (lanes 6 and 10). Collectively, these data indicated that the IUR included the sequence TCGAT located at position Ϫ97 to Ϫ93.
Activation of MGSA/GRO␣ Promoter by RelA-Previous work has demonstrated a cross-coupling of NF-B and C/EBP family members (39,59). In particular, these investigators demonstrated that regulation of the chemokine IL-8 gene expression relied on the ratio of NF-B and C/EBP complexes in that the C/EBP␤ protein had an inhibitory effect through the adjacent NF-B element, while RelA enhanced transactivation through the C/EBP element (39). We were interested in determining if RelA regulated MGSA/GRO␣ gene expression similarly through the NF-B and adjacent regions in a manner similar to that demonstrated for the IL-8 gene (38)(39)(40). RPE cells were transiently co-transfected with a RelA expression vector, and the MGSA␣350/CAT reporter constructs with point mutations in the IUR and NF-B element. Overexpression of RelA increased activity through both the MGSA␣350/CAT and the mutant IUR MGSA␣350/CAT (Fig. 7). Moreover, RelA also increased transactivation through the mutant NF-B MGSA␣350/CAT. The transactivation seen with the mutant NF-B construct was less than that observed with the mutant IUR construct. Furthermore, a second set of mutations in the NF-B element (GAAAATTTGGC) within the context of MGSA␣350/CAT indicated that RelA still significantly increased promoter activity (data not shown). RelA expression did not transactivate through either the double mutant MGSA␣350/CAT or parental pPLFCAT vectors, suggesting that the RelA transactivation observed with the mutant NF-B element construct was through the adjacent IUR region (Fig. 7).
To address whether RelA directly bound to the MGSA/GRO␣ IUR, RelA produced by transfected Jurkat T-cells was used in gel mobility shift analysis. RelA specifically recognized the MGSA/GRO␣ and IL-8 NF-B elements, although not the mutated MGSA/GRO␣ NF-B element. Furthermore, RelA does not bind to the MGSA/GRO␣ IUR nor enhance the binding ability of the IUR bound complexes present in nuclear extracts (data not shown). These data suggest that RelA indirectly effects transactivation through the MGSA/GRO␣ IUR.
IUR Contributes to Basal MGSA/GRO␣ Expression-Gel  Table I for sequences). B, nuclear extracts from IL-1-stimulated RPE cells were preincubated with preimmune antisera (PI) (lane 2) or antisera to RelA, C/EBP␤, and LBP (lanes 3-5) 20 min prior to labeled MGSA␣ 2xIUR addition. In addition, competitor DNAs including wild-type and mutant MGSA␣ 2xIUR, LBP, and NF-IL6 oligonucleotides were incubated with nuclear extracts prior to probe addition (lanes 6 -12). The resulting protein-DNA complexes were analyzed on 0.5 ϫ TBE polyacrylamide gels; the specific IUR bound factors (IUR-F) and nonspecific complexes (ns) are designated. The arrow indicates the nonspecific complex present in cytoplasmic extracts. shift analysis indicated that several key nucleotides in the IUR were essential for recognition by the constitutive complexes in both RPE and Hs294T melanoma cells. Furthermore, these point mutations strongly effected basal and cytokine-induced MGSA␣350/CAT activity, suggesting that the bound IUR complexes, in addition to the NF-B complexes, contributed to the transcriptional regulation of MGSA/GRO␣. Minimal promoter contructs containing a single copy of the IUR plus the NF-B element (MGSA␣(Ϫ97/Ϫ62)TATA) or two copies of the wild type (MGSA␣ 2xIUR/TATA) or mutated IUR (MGSA␣ m.2xIUR/TATA) were generated. Transient transfections in RPE cells demonstrated that both the wild-type MGSA␣(Ϫ97/ Ϫ62) and MGSA␣ 2xIUR/TATA constructs had a higher level of basal promoter activity as compared with the parental TATA/ CAT (Fig. 8). Furthermore, point mutations in the IUR, which resulted in loss of the constitutively bound IUR complexes effectively eliminated all basal promoter activity (Fig. 8). Together with the endogenous MGSA/GRO␣ promoter analyses, these data indicated that the IUR-bound complexes significantly contributed to MGSA/GRO␣ basal promoter activity.
Identification of IUR-bound Complexes by UV Cross-linking-To further characterize the complexes bound to the IUR, UV cross-linking studies were performed. Nuclear proteins bound to the labeled MGSA␣ 2xIUR in the gel shift reactions were either left untreated or exposed to short-wave UV irradiation for 15 min. Present in the reaction was 50-fold excess of either the wild-type or mutant MGSA␣ 2xIUR oligonucleotides. Half of the reaction was separated on a native 6% polyacrylamide gel (Fig. 9A). The wild-type MGSA␣ 2xIUR oligonucleotide (WT) effectively competed the upper bound complexes, while the mutant MGSA␣ 2xIUR oligonucleotide (MT) did not. UV radiation did not affect the bound complexes or the competition analysis (Fig. 9A, compare lanes 1-3 to lanes 4 -6). The remaining half of the binding reaction was separated on a 9% SDS-polyacrylamide gel (Fig. 9B). Two complexes were crosslinked to the labeled probe and appeared to be specific in that the wild-type oligonucleotide significantly lessened their presence, while mutant oligonucleotide did not (Fig. 9B, lanes 13). These complexes were not observed when the binding reactions were not exposed to UV radiation (Fig. 9B, lanes 4 -6). In these cross-linking studies, there were not specific complexes detected below the 40-kDa marker on the SDS-polyacrylamide gel (data not shown). The cross-linked complexes observed were approximately 68 and 50 kDa; however, the 42-base pair oligo-nucleotide radiolabeled probe present in the cross-linked complexes contributed approximately 28 kDa. Therefore, the two IUR bound complexes are within the 22-40 kDa range. These data indicated that multiple DNA binding proteins bind to the MGSA/GRO␣ IUR.  (15,(35)(36)(37). Furthermore, IL-8 gene regulation requires the C/EBP site adjacent to the NF-B site for complete cytokine induction (36,38,40). Similar to IL-8, the MGSA/ GRO␣ promoter contains a region adjacent to the NF-B element that is necessary for basal activity in both RPE and the Hs294T melanoma cells. Loss of either the IUR or the NF-B element eliminates most of the endogenous promoter activity. Cytokine-induced MGSA/GRO␣ promoter activity in RPE cells also requires both the NF-B element and the IUR. However, we demonstrate here that the regions adjacent to NF-B for the IL-8 and MGSA/GRO␣ chemokines differ markedly in their capacity to bind transactivating factors. For IL-8, the C/EBPlike consensus sequence binds several C/EBP proteins, although with weaker affinity than established C/EBP sites. In contrast, although there is sequence similarity to a C/EBP enhancer, C/EBP proteins do not bind the IUR in the MGSA/ GRO␣ promoter. Moreover, several of the essential nucleotides in the IUR are located upstream from the C/EBP-like region.
Gel shift analyses demonstrate that it is difficult to detect a factor specific for the IUR using nuclear extraction procedures that give optimal NF-B binding. This may in part be due to the observations that the IUR complexes require a more vigorous extraction procedure, which negatively affects NF-B binding, and that the IUR complexes are labile after nuclear extract collection (data not shown). Alternatively, the failure to detect  9. UV cross-linking analysis of IUR bound complexes. Nuclear extracts (5 g) from IL-1 stimulated RPE cells were incubated in 20 l of total reaction volume with radiolabeled MGSA␣ 2xIUR oligonucleotide in the presence or absence of WT or mutant (MT) MGSA␣ 2xIUR for 20 min at room temperature. The binding reaction was either left at room temperature (0Ј) or exposed to UV radiation for 15 min (15Ј). A, half of the binding reaction (10 l) was separated on a 6% 0.5 ϫ TBE native polyacrylamide gel. The specific IUR (IUR-F) and nonspecific (ns) complexes are indicated. B, the remaining binding reaction volume was separated on a 9% SDS-polyacrylamide gel. Molecular weight standards are indicated as are two specific complexes (approximately 68 and 50 kDa) cross-linked to the labeled MGSA␣ 2xIUR oligonucleotide (designated I and II). specific IUR complexes from nuclear extracts may be due to an unstable interaction with the labeled promoter regions. Specific MGSA/GRO␣ IUR-bound complexes are readily observed in vitro from both RPE and Hs294T nuclear extracts when two copies of the IUR are present, suggesting an increased affinity with duplicate copies. The observations that a single IUR element can compete for complex binding yet not detectably bind nuclear factors in gel shift assays suggest that the protein complexes bound to a single IUR may be unstable during electrophoresis.
The IUR bound complexes are effectively competed by oligonucleotides containing a single copy of the IUR without the NF-B element, suggesting the IUR complexes bind the DNA independent of the NF-B element. Furthermore, a series of point mutations narrows the IUR to a span of nucleotides containing the sequence TCGAT. This sequence does not have any similarities to known transcription factor sequences, suggesting that the factors that bind to the IUR are novel. UV cross-linking indicates the presence of at least two proteins of approximately 22 and 40 kDa bound to the MGSA/ GRO␣ IUR. These IUR-bound complexes are constitutively present, and IL-1 induction does not further increase the binding activity.
In addition to the NF-B element, RelA is able to transactivate through the IUR region within the MGSA/GRO␣ promoter. These observations are similar to those observed with IL-6 and IL-8 gene regulation in that both the adjacent C/EBP-like and NF-B elements are essential for cytokine induction and RelA transactivation (40). However, our results differ in that RelA alone is able to transactivate the MGSA/GRO␣ promoter and does not require an additional factor to be co-transfected.
The RelA induction through the IUR alone does not dictate that RelA cross-couples with the IUR-bound complexes as demonstrated for the IL-8 gene. RelA antisera has no effect on the IUR-bound complexes nor does RelA enhance the binding of the IUR complexes; therefore, a direct interaction between NF-B and IUR-bound complexes is not currently supported by our data. The IUR complexes do significantly contribute to MGSA/ GRO␣ basal promoter activity in that loss of the IUR complexes bound to the endogenous MGSA/GRO␣ promoter, or to a minimal promoter containing only the IUR sequences, dramatically decrease the amount of activity obtained from the native promoter. RelA may act indirectly to either induce expression of the IUR binding proteins or stabilize their interaction with the basal transcription machinery.
In summary, our results indicate that, as with other genes encoding proteins involved in the inflammatory response including IL-6 (40), IL-8 (36,38,39), serum amyloid A genes (52,60), and angiotensinogen (61), MGSA/GRO␣ transcriptional regulation requires multiple factors recognizing the NF-B and adjacent DNA binding elements. However, the IUR adjacent to the NF-B element in the MGSA/GRO␣ promoter appears to be unique. In addition, RelA has a dual role in MGSA/GRO␣ activation in that it is able to transactivate through both the NF-B element and the adjacent IUR. Since NF-B is ubiquitous and a multitude of genes contain NF-B elements, the regulation by RelA through a separate enhancer region may allow a tighter and more specific level of gene regulation. Future studies are needed to determine the mechanism by which this region interacts with the adjacent NF-B element to regulate MGSA/GRO transcription in normal and transformed cells.