The Role of Enhancer A in the Locus-Specific Transactivation of Classical and Nonclassical HLA Class I Genes by Nuclear Factor κB¹

The human classical MHC class I genes HLA-A, HLA-B, and HLA-C and the nonclassical MHC class I genes HLA-E, HLA-F and HLA-G differ in their level of expression in different cell types (reviewed in Refs. 1 and 2). The classical MHC class I molecules are crucial for immune recognition, as they present antigenic peptides to cytotoxic T lymphocytes (3). Nonclassical MHC class I molecules, such as HLA-E and HLA-G, are also able to present peptides (reviewed in 4 , but their exact role in the T lymphocyte-mediated immune response remains to be elucidated. However, there is increasing evidence that nonclassical HLA class I molecules are important in protection against NK cell responses (reviewed in 4 . Classical MHC class I molecules are ubiquitously expressed in adult tissues (reviewed in Refs. 1 and 2). In embryonic tissue, there is a general lack of classical MHC class I expression, a characteristic thought to be important for evading a potential maternal immune reaction against the developing fetus (5). In contrast, the nonclassical MHC class I genes HLA-E and HLA-F are expressed in many fetal and in some adult tissue (reviewed in Refs. 2 and 4). HLA-G has the most restricted expression pattern and is only expressed on the cytotrophoblasts at the materno-fetal interface, where it is likely to be important for the protection of the fetus against NK cytolysis (5, 6).

The regulation of MHC class I gene expression is mediated by several conserved cis-acting regulatory elements clustered in the promoter region of MHC class I genes. These include the enhancer A element, the IFN-stimulated response element (ISRE) and site α (reviewed in Refs. 2, 7, and 8–10).

Enhancer A is the target binding site for transcription factors of the NF-κB/Rel family and is thought to be essential for constitutive and cytokine-induced gene expression. The κB motif GGGGATTCCCC in enhancer A of the HLA class I gene promoters is highly conserved, particularly in the HLA-A and HLA-B loci (2). This is a symmetrical variant of the more divergent κB site in the promoter of the Ig κ-light chain gene (GGGACTTCC (11)). Although the κB motif is the principal target sequence for proteins of the NF-κB/Rel family, it is also bound by several other DNA-binding proteins, such as the high mobility group protein I(Y) (HMG I(Y))³ and proteins that belong to the leucine zipper family of transcription factors (reviewed in Refs. 12 and 13). The NF-κB/Rel family of transcription factors comprises at least five members, p50, p65 (also termed RelA), p52, c-Rel, and RelB, that bind DNA either as homo- or heterodimers (reviewed in Refs. 12 and 13). These dimers posess different binding affinities for κB sites and their half-sites, as well as different transcriptional properties (12, 13, 14, 15, 16).

The p50-p65 heterodimer, termed NF-κB, is present in virtually all differentiated cells and is the most abundant of the NF-κB/Rel dimers (12, 13). NF-κB is normally retained in the cytoplasm complexed to the inhibitory protein IκB (reviewed in Refs. 16–18). Following activation by stimuli such as cytokines and phorbol esters, IκB is inactivated and degraded, after which NF-κB is released and subsequently translocated to the nucleus (reviewed in Refs. 12 and 13). Interaction of NF-κB with the κB site results in transactivation of MHC class I and a variety of other genes, including those encoding cytokines and viral genes.

The level of gene transcription of the various MHC class I loci is determined by tissue-specific levels of expression of the NF-κB/Rel family proteins, their differential binding affinity for a particular κB site, and the transactivation capacities of the different dimers (reviewed in Refs. 12 and 13). The NF-κB subunits p65 and c-Rel both have a transactivation domain, although p65 is the more potent of the two. The NF-κB subunit p50, lacking such transactivation domain, is thought to have no transactivating capacity as a homodimer but rather to fulfill an ancillary function for the p65 and c-Rel subunits. However, since in cell-free assays the p50 homodimer can transactivate MHC class I (19), its exact role in transactivation is still unclear.

The human classical and nonclassical MHC class I genes differ in their level of constitutive expression in different cell types and also in their cytokine-induced expression patterns (reviewed in 2 . Locus-specific variation in the two putative κB sites of the enhancer A elements (2, 20, 21) determines their binding capacity and functioning. Locus-specific contribution of the enhancer A element to transactivation may be an important mechanism in the differential regulation of the constitutive and cytokine-induced levels of classical and nonclassical HLA class I expression in various cell types.

In this study, we assessed the capacity of the two putative κB sites within the enhancer A region in the promoter of the classical and nonclassical HLA class I genes to bind proteins of the NF-κB/Rel family of transcription factors and to mediate transactivation by NF-κB. The NF-κB-mediated transactivation of HLA class I genes is restricted to the HLA-A and HLA-B loci and seems to require binding not only of NF-κB to the κB1 site, but also of either NF-κB or Sp1 to the κB2 site.

The cells used were the teratocarcinoma cell line Tera-2 and the B lymphoblastoid cell line CCRF-SB (American Type Culture Collection, Manassas, VA). The cells were grown in Iscove’s modified DMEM supplemented with 10% (v/v) heat-inactivated FCS (Life Technologies, Paisley, Scotland), penicillin (100 IU/ml), and streptomycin (100 μg/ml).

Reporter constructs pGL3-A230 and pGL3-A140 were generated by cloning a 228-bp BglI-AhaII HLA-A*0201promoter fragment or a 143-bp PpuMI-AhaII HLA-A*0201 promoter fragment, respectively, upstream of the firefly luciferase gene in pGL3-Basic (Promega, Madison, WI). Enhancer A-containing reporter constructs were generated by cloning ds-oligonucleotides of the enhancer A sequence (containing both the κB1 and κB2 motifs) from the various HLA class I genes upstream of the 143-bp HLA-A*0201 promoter fragment in pGL3-A140.

The expression vector constructs of pRSV-p50 and pRSV-p65 (a kind gift of Dr. L. Struyk, Academic Medical Centre, Amsterdam, The Netherlands) contain the 1.3-kb cDNA and 2.6-kb cDNA fragments encoding NF-κB-p50 and NF-κB-p65, respectively, and were generated by inserting the cDNAs into pRSVNeo (22) from which the Neo^R gene has been deleted.

Cells were transfected by the calcium phosphate coprecipitation method of Chen and Okayama (23). In each of four wells of a six-well plate, 0.2 × 10⁶ Tera-2 cells were transfected with a DNA mix containing 2.5 μg of pGL3 reporter plasmid, 2.5 μg of pRSV-lacZ plasmid, and 5 μg of pRSV (control) or 2.5 μg of pRSV-p50 and pRSV-p65 (NF-κB) each and harvested 2 days after transfection. Luciferase activity was determined using a luminometer (Tropix, Badford, MA) and corrected for transfection efficiency by measuring β-galactosidase activity. Transfection of pGL3-Basic was used as a reference to correct for the possible effects of cotransfection with the expression vectors. The relative luciferase values are given as mean relative light units ± SD (n = 4).

Nuclear extracts were prepared from 10 × 10⁶ cells. The cells were harvested, washed with PBS, taken up in 300 μl (three cell volumes) of hypotonic solution (20 mM HEPES, pH 8.0, 10 mM KCl, 0.15 mM EGTA, 0.15 mM EDTA, 1 mM DTT, 0.5 mM 4-(2-aminoethyl)-benzenesulfonylfluoride (AEBSF)), and then left on ice for 15 min. The cells were lysed with Nonidet P-40 (final concentration of 0.1% for CCRF-SB cells; 0.2% for Tera-2) for 3 to 5 min. Then, 80 μl (80% of the cell volume) of a sucrose solution (50 mM HEPES, pH 8.0, 10 mM KCl, 0.25 mM EDTA, 1 mM DTT, 0.5 mM AEBSF, 70% (w/v) sucrose) was added, and the nuclei were centrifuged at 5000 rpm for 5 min at 4°C. The supernatant was discarded, and the pellet was gently taken up in 300 μl (three cell volumes) of solution B (10 mM HEPES, pH 8.0, 100 mM NaCl, 0.1 mM EDTA, 1 mM DTT, 0.5 mM AEBSF, 25% (v/v) glycerol) and centrifuged at 5000 rpm for 5 min at 4°C. After the supernatant was discarded, the cell pellet was taken up in 200 μl (2 cell volumes) of extraction solution (10 mM HEPES, pH 8.0, 400 mM NaCl, 0.1 mM EDTA, 1 mM DTT, 0.5 mM AEBSF, 25% glycerol) and left on ice for 30 min with intermittent vortexing. The extracted nuclei were centrifuged at 14,000 rpm for 5 min at 4°C, and the supernatant was aliquoted and stored at −80°C. The total amount of protein was determined using the BCA (bicinchoninic acid) Protein Assay Reagent kit (Pierce, Rockford, IL) according to the manufacturer’s instructions.

Nuclear extracts (∼5 μg protein) were incubated in DNA/protein-binding buffer (20 mM HEPES, pH 7.9, 50 mM KCl, 10% (v/v) glycerol, 0.5 mM DTT, 0.1 mM EDTA), with 250 ng poly(dI · dC), 100 ng sonicated herring sperm ssDNA, and 1 ng ³²P-radiolabeled probe for 15 min at 4°C. The samples were run on a 6% nondenaturing polyacrylamide gel in 0.25× TBE (0.89 M Tris-borate, 0.89 M boric acid, and 0.02 M EDTA) at 200 V for 2 h. The gels were fixed with a 10% methanol, 10% acetic acid solution, dried onto Whatmann 3M paper, and exposed to an x-ray film. The ds-oligonucleotides containing the individual κB1 and κB2 sites from the various HLA class I genes were used as probes (see Table I).

For the supershift assays, 1 μg of each Ab specifically directed against a member of the NF-κB/Rel family of transcription factors was added to the nuclear extract and probe mixture and incubated for 1 h at 4°C. The Abs used were anti-p50 Ab (sc-114), anti-p65 Ab (sc-109), anti-c-Rel Ab (sc-71), anti-Sp1 Ab (sc-59), anti-Fra1 Ab (sc-605), anti-Fra2 Ab (sc-171), anti-Fos Ab (sc-413), anti-Jun Ab (sc-44), and anti-IRF-1 Ab (sc-640), all from Santa Cruz Biotechnology (Santa Cruz, CA), and an anti-HMG I(Y) antiserum (24) (kindly provided by Dr. Dimitris Thanos, Columbia University, New York, NY).

The promoter regions of classical and nonclassical HLA class I genes contain two putative κB sites, κB1 and κB2, in their enhancer A elements. The κB1 site in enhancer A of HLA-A(GGGGATTCCCC; Table I) is the most conserved throughout the MHC class I loci and was used as reference to test for binding of proteins of the NF-κB/Rel family in EMSAs. Using nuclear extracts from the B lymphoblastoid line CCRF-SB, which constitutively expresses proteins of the NF-κB/Rel family, this κB motif was found to bind several protein complexes: an upper complex consisting of two closely migrating bands, designated complex 1 and complex 2; and a lower complex, designated complex 3 (Fig. 1,A). Employing antisera specific for p50, p65, and c-Rel (25, 26), the proteins contained in these complexes could be identified (Fig. 1, A and B). The uppermost band in the duplex (complex 1) was shown to contain p65. This protein/DNA complex probably represents the p50-p65 heterodimer, although the presence of p65-p65 homodimers cannot be excluded. The lower band of the two closely migrating protein/DNA complexes (complex 2) was shown to contain c-Rel. Complex 2 could be either the p50-c-Rel or the p65-c-Rel heterodimer. Finally, the lower band (complex 3), shown to contain the p50 subunit, is most probably the p50-p50 homodimer (Fig. 1, A and B). No supershift was obtained with an irrelevant Ab (Fig. 1 C).

Next, a panel of probes representing the individual κB motifs of all HLA class I loci (Table I) was used to perform a complete analysis of their binding properties for proteins of the NF-κB/Rel family using nuclear extracts from B cells. In Figure 2, A and B, it is shown that, similar to the conserved κB1 site of the HLA-A locus, the κB1 site of the HLA-B locus (GGGGATTCCCC), was bound by the complexes containing the p50, p65, and c-Rel subunits. The κB2 site in HLA-A (GGGGAGTCCCA) was also bound by these complexes, but with an apparently weaker binding affinity (Fig. 2, A and B). No binding of proteins of the NF-κB/Rel family was detected for the κB2 sites of HLA-B(GGGGAGGCGCA) and the κB1 and κB2 sites of HLA-C(GAGGATTCTCC and GGGGAGGCGCC, respectively) under these experimental conditions (Fig. 2,A). EMSAs with κB1 and κB2 probes of the nonclassical HLA class I genes revealed that the κB1 site in HLA-F (GAGAATTCCCC) was bound by the p50, p65, and c-Rel subunits (Fig. 2,A), while both the κB1 and κB2 sites in HLA-G (GGGGATTCTCT and GGGGAGGCCCC, respectively) were bound only by the p50-containing complex (complex 3; Fig. 2, A and B); binding of the p65-containing complex was hardly detectable and only after long exposure times (data not shown). No NF-κB complex could be detected binding to the κB2 site of HLA-F (TTGGAAGGCTC) nor to the κB1 and κB2 sites of HLA-E (GGGAAACTCTG and AAGAACTGCTG, respectively) (Fig. 2 A).

Several κB site probes were seen to bind proteins not belonging to the NF-κB family of transcription factors for which we were testing. For example, a slowly migrating complex (referred to as complex A) was found to bind the κB2 sites of HLA-B, HLA-C, and HLA-G, which all share the GGGGAGGC^G/_CC sequence (Table I). The specificity of the binding was tested by competition with cold κB probes. Cold κB2 probe of HLA-C or of HLA-G competed with the κB2 site of HLA-C for binding of complex A, whereas cold κB1 probe of HLA-A did not compete (Fig. 3,A). Conversely, binding of complex A to the κB2 site of HLA-G was inhibited when using cold κB2 probe of HLA-C or of HLA-G, but not when using the κB1 probe of HLA-A (data not shown). Since the sequence of these κB2 sites is homologous to an Sp1 binding site (27, 28, 29), we tested for binding of Sp1 to the κB2 site of HLA-C. Using an anti-Sp1-specific Ab in supershift assays, complex A was shown to contain Sp1 (Fig. 3,B). The complex could not be supershifted with an irrelevant Ab (Fig. 3 B). Similarly, Sp1 binding could also be demonstrated to the κB2 sites of HLA-B and HLA-G (data not shown).

A fast migrating complex (complex B) was found to bind specifically to the κB1 sites of HLA-C and HLA-G (Fig. 2, A and B), which share the GGATTCTC sequence. The specificity of the binding was also tested by competition with cold κB probes, and binding of complex B was competitive when using the cold κB2 probe of HLA-C or of HLA-G (containing the homologous sequence), but only weakly competitive with the divergent κB1 probe of HLA-A (data not shown). These characteristics suggest that complex B contains the high mobility group I(Y) protein (HMG I(Y)). Using an HMG I(Y)-specific Ab (24), a supershift was obtained (data not shown), strongly suggesting that this complex contains HMG I(Y) protein.

Recently, the Fos/Jun family protein Fra-2 has been shown to bind the κB binding site of porcine MHC class I (30). Since this κB binding site is homologous to the κB2 site of HLA-A, we tested for binding of members of the Fos/Jun family of transcription factors to the κB1 and κB2 sites of HLA-A. However, we were unable to detect Fra-2 or other factors of the Fos/Jun family in the complex binding to the κB1 site or κB2 site of HLA-A, using nuclear extracts of CCRF-SB or EBV-transformed B cell nuclear extracts (data not shown).

Having determined the binding properties of NF-κB/Rel proteins to the κB1 and κB2 sites of enhancer A, we evaluated the contribution of the enhancer A elements of the various HLA class I loci to transactivation in transient cotransfection experiments. For these experiments, reporter constructs were generated containing the enhancer A sequence of the various HLA class I genes in front of a minimal promoter fragment of HLA-A2.1 (Fig. 4 A).

First, we determined the transactivation capacity of the NF-κB subunits using pGL3-A230, containing the 228-bp HLA-A2.1 promoter fragment. As shown in Figure 4,B, cotransfection with the p50 subunit alone did not result in an increase of HLA-A2.1-driven transcription in Tera-2 cells, whereas cotransfection with p65 resulted in a 6-fold induction. Together, p50 and p65 were able to induce HLA-A2.1-driven transactivation to an even higher level (14-fold), which supports the suggestion that they act synergistically (Fig. 4 B).

Subsequently, the reporter constructs containing the enhancer A sequence of the various HLA class I genes in front of a 143-bp HLA-A2.1 promoter fragment were transiently cotransfected with expression vectors of p50 and p65 in Tera-2 cells. Figure 4,C shows a typical transient cotransfection experiment. Enhancer A of HLA-A, containing two κB sites able to bind NF-κB, gave rise to a significant transcriptional activity when cotransfected with p50 and p65 (15-fold induction), whereas transactivation of enhancer A of HLA-B (containing only one κB binding sequence) was relatively low (3-fold induction; Fig. 4,C). The enhancer A of HLA-C did not give rise to any significant NF-κB-induced transactivation activity, which is in accordance with the inability of the κB1 and κB2 motifs of HLA-C to bind NF-κB subunits. None of the enhancer A elements of the nonclassical HLA class I genes HLA-E, HLA-F, or HLA-G gave rise to any significant NF-κB-induced transactivation activity (Fig. 4 C). Despite the binding activity of the p50 and p65 to the κB1 of HLA-F, no transactivation of HLA-F was observed; this could be attributed to the lack of binding to the κB2 site. As expected, binding activity of p50 to both the κB1 and κB2 of HLA-G did not result in transactivation. Similar results were obtained in HeLa cells (data not shown).

Variation in the nucleotide sequence of the κB binding sites determines both the binding affinity of the various NF-κB dimers and conformational changes in the nucleoprotein complex (19, 31, 32, 33, 34). Since the various NF-κB complexes differ in their transactivation capacities (12, 13), this could result in differences in activation of gene transcription. Therefore, locus-specific variations in the nucleotide sequences of the κB binding sites in HLA class I genes could be at the heart of their differential NF-κB-mediated regulation via enhancer A. In this study, the κB sites in enhancer A of the promoter region of classical and nonclassical HLA class I genes were investigated for their capacity to bind proteins of the NF-κB/Rel family of transcription factors and their contribution to transactivation.

The enhancer A element in HLA class I contains a palindromic κB site, GGGGATTCCCC, referred to as κB1. This site is conserved in the HLA-A and HLA-B loci but is also found, albeit with nucleotide alterations, in the promoters of the other HLA class I loci. In addition, upstream of the κB1 site is the κB2 site GGGGAGTCCC, which is found in the HLA-A alleles and also, with minor to major nucleotide variations, in the promoters of the other HLA class I loci. Both the κB1 and κB2 sites of the enhancer A region of HLA-A have been shown to be occupied, as determined by in vitro footprinting (35). In this study, the κB1 site of HLA-A and HLA-B (GGGGATTCCCC) was found to bind three complexes containing members of the NF-κB/Rel family of transcription factors, which is in general agreement with findings by others (21, 29, 36). Supershift assays indicated that the three complexes that were found to bind the κB1 site contained p50, p65, and c-Rel subunits. These complexes could represent the p50-p65 heterodimer, the p50-c-Rel or p65-c-Rel heterodimer, and the p50-p50 homodimer. Although in HLA-A the κB2 site (GGGGAGTCCCA) differs by only two nucleotides (underlined) from the κB1 site, it displayed a reduced binding affinity for all three complexes under these experimental conditions. The central nucleotides (ATT) in a κB site function as the binding site for HMG I(Y) proteins, which play a role in the DNA bending and the ultimate structure of the nucleoprotein complex and also act as coactivators in NF-κB-mediated transactivation (24, 37). These central nucleotides are rarely guanines, and it is therefore likely that the central guanine in the κB2 site of HLA-Ais responsible for this weaker binding affinity. No binding of any NF-κB complexes could be detected to the κB2 site of HLA-B or to either of the κB sites of HLA-C.

Two of the nonclassical HLA class I genes were bound by NF-κB complexes. The κB1 site of HLA-F (GAGAATTCCCC) was bound by the three p50-, p65-, and c-Rel-containing complexes, despite two nucleotide differences from the consensus sequence. The κB1 site of HLA-G(GGGATTCTCT) showed binding of only p50, whereas binding of the other NF-κB subunits (p65 and c-Rel) was barely detectable. The specificity of the κB2 site of HLA-Gfor p50 is in agreement with in vitro DNA binding studies in which it was shown that the otherwise perfect κB motif GGGGAGGCCCC is bound only by p50 (33). No binding of any κB-specific complexes could be detected to both κB sites of HLA-E or to the κB2 site of HLA-F. Together, these data argue that only the κB2 site of HLA-A is a true NF-κB binding site and that the putative κB2 site in the promoter of the other HLA class I loci cannot be termed a κB site.

Two other complexes, designated complex A and B, were found to bind to some of the κB motifs of HLA class I loci; these complexes did not contain any of the κB/Rel family proteins for which we were testing. The slower migrating complex A had an apparent binding specificity for the GGGGAGGC^G/_CC sequence. This sequence has been proposed as a binding site for Sp1 in HLA-B(28, 29). Furthermore, both half-sites of this sequence are homologous to zinc finger binding sites: the 5′ half-site is homologous to the binding site for the myeloid zinc finger protein MZF1 (38), and the 3′ half site is homologous to the binding site for the zinc finger protein ZFX, identified in the promoters for HLA-A genes (39, 40). In this study, we demonstrated that complex A contained Sp1; but since the supershift was not complete with the Ab used, the possibility cannot be excluded that other (zinc finger) proteins are contained in the complex. The faster migrating complex B had an apparent binding specificity for the GGATTCTC sequence. These characteristics, and the fact that a supershift could be obtained using an anti-HMG I(Y) antiserum (24), strongly suggest that complex B consists of HMG I(Y) proteins (24, 37). It is possible that in the absence of NF-κB binding, such as to the κB1 sites of HLA-C and HLA-G, formation of a complex with HMG I(Y) proteins is more prominent.

Transactivation by NF-κB dimers is dependent on the subunits they contain. Both the p65 and c-Rel subunits possess transactivation domains, but p65 has been shown to be the more potent transactivator of the two (12, 13). Experiments in mice have shown that a transgene driven by a promoter containing multiple κB sites was not transcribed in organs in which only the p50 subunit was expressed; the transgene was found to be transcribed only in those organs in which both p65 and p50 were expressed (41). Moreover, p50 alone does not activate transcription in transient transfection assays (this study and Refs. 29, 42, and 43). The p50 subunit has therefore been attributed an ancillary function in DNA binding of the dimer and in transactivation by the p65 or c-Rel subunits.

The enhancer A elements of classical HLA class I genes HLA-A and HLA-B were, in accordance with their ability to bind the p50-p65 heterodimer, mediators of NF-κB-induced transcription. Enhancer A of HLA-A, which contains two κB sites binding the different NF-κB dimers, displayed the strongest transactivation capacity of all of the HLA class I loci. Multiple κB binding sites that are occupied by NF-κB, as is the case for HLA-A, can lead to a synergistic transcriptional effect (44) and are thought to be required for transcriptional induction (21, 45). Less marked was the transactivation through the enhancer A region of HLA-B, which contains only one NF-κB binding site (κB1). In this locus, the κB2 site is bound by Sp1 (this study and Refs. 28 and 29), a transcription factor that is shown able to interact with NF-κB (46, 47). Sp1 binding to the κB2 site may assist in the NF-κB-mediated transactivation through the κB1 site by allowing protein/protein interactions, leading to transactivation through the enhancer A of HLA-B (28, 48). This would explain why there is still transactivation through the single NF-κB-occupied κB1 site in this locus (this study and Refs. 28 and 29). In support of this hypothesis is the lack of NF-κB-mediated transcriptional activity through the κB1 site in the enhancer A region of HLA-F, because unlike the HLA-B gene, HLA-F does not bind transcription factors to the upstream κB2 site. These findings are corroborated by the observation that the single κB site in the bidirectional promoter of the TAP1 and LMP2 genes has been found to depend on the flanking Sp1 site for TNFα-induced transcription by NF-κB (49, 50). The other HLA class I loci were not found to be regulated by NF-κB, which is in accordance with the lack of NF-κB binding to the κB1 site (HLA-C, HLA-E) or to the κB2 site (HLA-F), as explained above. Finally, the lack of NF-κB-induced transactivation of HLA-G can simply be explained by a lack of transactivating activity by the p50-p50 homodimer (this study and Refs. 29 and 43).

In this study, it is shown that two classical, but none of the nonclassical, HLA class I genes are transactivated by NF-κB. Transactivation was most important in the HLA-A locus, which contains two NF-κB binding sites in the enhancer A region. Although the enhancer A of HLA-B contains only one NF-κB binding site (κB1), there was still a moderate transactivation by NF-κB. Binding of Sp1 to the upstream κB2 site seems to contribute to this transactivation through the κB1 motif in HLA-B, because HLA-F, which also has one NF-κB binding site but lacks an NF-κB or Sp1 binding site flanking the κB1 site, was not transactivated by NF-κB. Thus, NF-κB-mediated transactivation of the HLA class I genes seems to require the binding not only of the κB1 site by NF-κB, but also of the κB2 by either NF-κB or Sp1. Since transcriptional regulation of HLA class I genes by NF-κB is restricted to HLA-A and HLA-B genes, the expression and activation of NF-κB contribute only to the developmental and tissue-specific expression patterns of these classical HLA class I genes.

We thank Drs. C. S. Bigland, F. H. J. Claas, M. J. Giphart, A. Peijnenburg, and C. Verweij for critically reading the manuscript; and Drs. J. Arts, L. Struyk, and D. Thanos for the kind gift of the anti-Sp1 Ab, NF-κB expression vectors, and anti-HMG I(Y) antiserum, respectively.