Promoter Region Architecture and Transcriptional Regulation of the Genes for the MHC Class I-Related Chain A and B Ligands of NKG2D¹

Diverse MHC class I-like molecules that have no role in Ag presentation and limited tissue distributions serve as ligands for the NKG2D-DAP10 receptor complex, which activates NK cells and costimulates effector T cell subsets (1, 2). In humans, these ligands include the closely related MHC class I-related chains (MIC)³ A and B (MICA and MICB) transmembrane glycoproteins, which are encoded near HLA-B in the MHC and are represented by orthologous sequences in most mammals, except rodents (3). Expression of MIC is mostly restricted to intestinal mucosa, but can be induced by CMV infection in fibroblasts and endothelial cells, and by mycobacterial infection in dendritic and epithelial cells (4, 5, 6). Moreover, MIC are frequently associated with epithelial tumors of diverse tissue origins and are aberrantly expressed in rheumatoid arthritis synoviocytes and celiac disease intestinal epithelial cells (7, 8, 9). Thus, NKG2D triggering by MIC delivers immunostimulatory signals that can be beneficial under adverse conditions, such as infections and malignancies, but may exacerbate autoimmune disease progression.

Despite the immunological significance of MIC, molecular mechanisms controlling gene regulation are poorly defined, and it is unknown whether and how different cellular and environmental stimuli converge to induce gene expression. Recent evidence has indicated that activation of DNA damage control pathways results in induction of NKG2D ligands, including UL16-binding protein family members and possibly MICA (10). Moreover, the 5′-end flanking regions of MICA and MICB contain putative heat shock elements (HSE), which are prototypic transcription inducer sites in heat shock protein 70 (HSP70) genes that bind activated trimeric heat shock factor 1 (HSF1) (4, 11, 12, 13). With cell lines, MIC mRNA and protein expression are mostly limited to proliferating epithelial cells. Quiescent epithelial cells grown for extended time at high confluence display relatively small amounts of MIC mRNA and surface proteins that are sharply increased upon exposure to heat shock (14). Oxidative stress has also been found to induce MIC gene expression in colon carcinoma cells, although increased cell surface expression was not observed (15). The present study has used molecular and functional approaches to define the architecture of MIC gene promoter regions and the significance of transcriptional control elements for cell stress-induced, proliferation-associated, and CMV-mediated transcriptional activation.

Cell lines were from the American Type Culture Collection. Primary human fibroblasts (passages 4–6) and Hela S3 cells were grown in Waymouth’s and MEM-Joklik medium (Invitrogen Life Technologies) supplemented with 10% FBS (HyClone), glutamine, and antibiotics. Surface expression of MICA, or of MICA and MICB, was monitored by flow cytometry using mAbs 2C10 and 6G6, respectively (4, 14). For heat shock, culture plates with adherent HCT116 cells grown for 7 days at high confluence were sealed with parafilm and floated for 1 h on a 42.5°C water bath. Suspension Hela S3 cells were pelleted, and resuspended and maintained for 40 min in 43°C medium.

EMSA were performed using human rHSF1 (StressGen Biotechnologies) and whole cell extracts of heat shock-treated Hela S3 cells (16, 17). Annealed oligonucleotides (Fig. 3 A) were labeled with [γ-P³²]ATP using T4 polynucleotide kinase (New England Biolabs), run in polyacrylamide gels, and purified using NENSorb columns (NEN-DuPont). DNA probes (20,000 cpm, ∼1 ng) were mixed with 4 μg of cell extract or 0.2 μg of HSF1 in 20 μl of binding buffer, incubated, and subjected to electrophoresis, as described (18). For supershifts, 1 μl of a 1/50 dilution of a rabbit anti-human HSF1 serum (StressGen Biotechnologies) was added to the HSF1-binding reaction after 10 min at 22°C (18). For competition experiments, 100-fold excess unlabeled oligonucleotides were added 5 min before the labeled probes. Probes for DNase I footprinting were made by PCR amplification of a fragment, including the MICB 5′-flanking region from cosmid R5A (3) using primers (both 5′-3′) ACAGGGTCCAGGTCGTGCTCATA (–334/–312) and GTGCAAAAGGGAAGGCGACG (+52/+33) after 5′-end labeling of one of the primers. PCR products were isolated by gel electrophoresis and spin column purified (Qiagen). Binding of HSF1 was tested by incubation of 2 μg of recombinant protein with DNA probe (25,000 cpm, ∼15 fmol) in reaction buffer for 10 min at 22°C before addition of DNase I (0.1 U) for 2 min (19). After addition of stop buffer, reaction products were purified and resolved in 6% sequencing gels (19).

MICA and MICB transcription start regions were localized by RT-PCR using sets of forward primers extending into the 5′-flanking regions. The reverse primers (all 5′-3′) were TGCCAGCCAGAAGCAGGAAGACC (MICA +37/+15) and AGGCGACGGCCAGAAACAGCAG (MICB +41/+20). The forward primers (all 5′-3′) were CCACTGCTTGAGCCGCTGAGAG (–40/–19), CAGTTTCATTGGATGAGCGGTCG (–109/ –87), GTTCCGGGCCCCAGTTTCATTGGATG (–120/–95), CGTGGCCCCGCCCTCTCCGCTC (–156/–135), and CTTCTAAATCTCCCCAGGTCTCCAG (–205/–181) for MICA; and CCACTGCTGAGCAGCTGAGAAG (–39/–18), CAGTTTTCACTGGATAAGCGGTCG (–124/–101), GTTCCGGGCACAGTTTTCACTGGAT (–135/–110), CTCCACTCATGATTGGCCCTA (–156/–135), and CGTGGCCCCGCCCTCTCCACTC (–170/ –149) for MICB.

Nested sets of MICA (*004) and MICB (*004) 5′-end flanking regions were PCR amplified from cosmids M32A (gi:3451361) and R9A (gi:3924652), respectively (3). The 3′-end primers (same as the reverse primers above) were downstream of the translational start codons, which are within NcoI restriction sites that were used for insertion into pGL3-basic (Promega). The 5′-end primers included BamHI restriction sites and were positioned as illustrated in Fig. 2. MICB promoter element mutations were made by megaprimer PCR (20) using the flanking primers of the B-470 regulatory region and the following mutagenic primers (all 5′-3′; nucleotide substitutions shown in bold, wild type in parentheses): HSE* CCAGGCCGCTTC(AG)AATTTTG(C)TCTTCTGAACGTGGCC (–198/–164); Sp1* GGCCCCGAGT(CCC)TCTCCACTCATGATTGGCC (–167/–139); CBE* CCACTCATGATAC(TG)GCCCTAAGTTCCGGGCACAG (–154/–122); and TATA* TCCCGCCTTCGC(TA)AATTCCCCAGGTCTCCAG(–225/–195). The corresponding MICA promoter element mutations were generated similarly within the A-455 region. The 14-bp deletions (Δ) in the MICB (–41 to −54) and insertions (▿) in the MICA (between −41 and −42) promoter regions were generated using double-stranded primers (5′-3′) CCACGGGTTCTTCTCACCG(Δ)GCCACTGCTGAGCAGCTGAG and CCGCGGCGCCTTCTCCCCGGTTTCATTCAGTTGGCCACTGCTTGAGCCGCTGAGG, respectively, and QuikChange-II site-directed mutagenesis methodology (Stratagene). All DNA fragments in pGL3-basic were verified by sequencing.

Promoter region activities in transiently transfected cell lines were determined by dual luciferase reporter assays. HCT116 cells were seeded at 2 × 10⁵/well in 48-well plates, grown to 50% confluence, and transfected with reporter plasmid DNAs (∼0.2 μg, adjusted for m.w. equivalence) together with internal control phRL-TK (∼8 ng; Promega) using Fugene 6 reagent (Roche). For assays of transcriptional activities in proliferating, and heat shock-treated or untreated HCT116 cells, these were grown for 48 h or 7 days, respectively. After heat shock, cells were cultured for a 6-h recovery period.

Human primary dermal fibroblasts were grown in six-well plates and transfected with reporter plasmids and pEQ276 driving expression of the CMV immediate-early (IE) 1 and 2 genes or mock control pEQ336 containing the IE2 promoter/enhancer, but not the coding region using Lipofectamine 2000 (Invitrogen Life Technologies). To test for induction of MIC surface expression, human primary dermal fibroblasts were also transfected with pEQ274 and pEQ326 encoding IE1 and IE2, respectively, and with pEQ876 encoding the irrelevant CMV tegument protein (21) (plasmids were a gift from A. Geballe, Fred Hutchinson Cancer Research Center, Seattle, WA). Cells were examined by flow cytometry 72 h after transfection. For the CMV activation experiments, fibroblasts were infected with viral strain AD169 (American Type Culture Collection; multiplicity of infection of 5) 24 h after transfection with MIC gene promoter region reporter constructs. Cells were harvested and assayed 40 h after infection. For all reporter enzyme assays, washed cells from individual wells were lysed in 100 μl of passive lysis buffer (Promega), and firefly and Renilla luciferase activities were measured in 10 μl of cell extracts in 96-well white luminometer plates (Fisher Scientific) using an EG&G Berthold luminometer with automated dual injectors and time delay. Promoter construct activities were determined by normalizing firefly luciferase activity for transfection efficiency, as measured by Renilla luciferase activity.

Heat shock-treated or untreated HCT116 cells were processed for ChIP assays, following the kit protocol from Upstate Biotechnology. Immunoprecipitations used mixed rabbit and mouse IgG, anti-HSF1 and anti-CCAAT box factor (CBF)-B mAbs, and a polyclonal rabbit anti-Sp1 antiserum (1 μg/ml; Santa Cruz Biotechnology). Immunocomplexes were collected with protein A/G agarose beads and sequentially washed, as detailed in the Upstate Biotechnology protocol. Chromatin was eluted with 1% SDS/0.1 M NaHCO₃ and digested with proteinase K, and cross-linking was reversed for 5 h in the presence of 200 mM NaCl at 65°C. DNA was purified using Qiaquick Gel Extraction kit (Qiagen) and dissolved in 100 μl of Tris-EDTA. Fractions of precleared chromatin were processed similar to immunoprecipitated chromatin for input DNA controls. The forward and reverse primers for PCR amplification (all 5′-3′) were GGGCCCTGGCCGTGCTTATGAAGTTGG (–317/–291) and CGCCACCCTCTCAGCGGCTCAAGC (–12/–35) for MICA; GcGA CAGGGTCCAGGTCGTGCTC (–337/–315) and CCCTACGTCGCCACCTTCTCAG CT (–4/–27) for MICB; and GTCCCTGTCCCCTCCAGTGAAT (–439/–418) and GAACACTGGATCCGCGAGAAGA (–53/–74) for HSP70.

Aligned cosmid-derived regions 1.5 kb upstream of the translation start codon (ATG) in MICA and MICB share ∼90% sequence identity and include within a 50-bp region proximal sequence motifs for binding of HSF1 (HSE; consensus GAANNTTC) and Sp1 (consensus GGG(A/C/T)GGG), and an inverted CCAAT box-like element (ICE) (Fig. 1 A) (22, 23, 24). Additional conserved Sp1 sites are located ∼160 and 190 bp upstream of the HSE. A potential AP-1 element is located downstream. Although canonical TATA boxes are absent, aligned TATA-like sequences are present in MICA (TAAGTTT and TAAATCT, at –67 and –200, respectively) and MICB (TAAATTT at –80 and –215). A sequence of 14 bp in MICB (GTTTCATTCAGTTG, from –41 to –54) includes two overlapping motifs matching the consensus for an initiator (Inr) element (YYCANYY) (25). This sequence is absent in MICA. However, potential Inr sequences occur in MICA at −112, –185, and −231; moreover, there is an additional unique Inr in MICB at –155 and a conserved Inr in both genes (at –281 and –295, respectively).

Transcriptional initiation was investigated with colon carcinoma HCT116 cells. Similar to other epithelial tumor lines, highly confluent HCT116 cells grown to quiescence contain small amounts of MICA and MICB mRNAs and display low MIC surface protein expression. However, in rapidly proliferating or heat-shocked cells, MICA and MICB mRNAs and surface proteins are strongly induced (Fig. 1,C and see figure 3 in Ref. 14). We failed to obtain informative transcription initiation data using primer extension and RNase protection assays. Therefore, MICA and MICB 5′-end flanking regions were amplified by RT-PCR using nested sets of primer pairs of which the reverse primers were invariably positioned downstream of the respective ATG. For each gene, the transcript data were similar irrespective of whether proliferating or heat shock-treated HCT116 cells were examined (Fig. 1,B). The majority of MICA transcripts initiated between about –40 and –100 upstream of the initiation codon, and a minor fraction extended further upstream. These results were corroborated by comparison with 23 MICA sequences in GenBank expressed sequence tag database, which localized a major transcription initiation region ∼30 bp downstream of the proximal TATA-like motif. Among these sequences, 15 initiate between –32 and –42, including 9 sequences that start at −38; 5 sequences initiate between –48 and –55, 1 at –82, and only 2 sequences initiate far upstream (–2880 and −2932; gi:14057251 and gi:82384203, respectively). These sequences were derived from normal human pericardium and an adenocarcinoma cell line. The immediate upstream genomic sequence includes distal TATA-like, Sp1, and CAAT box motifs, but no potential HSE. Their existence tentatively suggests the presence of a second upstream promoter in MICA, but supportive functional evidence is lacking, and seeking such evidence was outside the scope of our study. In contrast to MICA, transcriptional initiation of MICB was localized predominantly between –124 and –156, upstream of the proximal TATA-like sequence and downstream of the conserved SP1 site (Fig. 1 B).

For functional studies of MICA and MICB gene promoter regions, nested sets of 5′-end flanking sequences immediately upstream of the translation start codons were inserted into pGL3-basic and tested for their ability to drive firefly luciferase reporter gene expression in transiently transfected proliferating HCT116 cells. Transfection efficiencies were normalized by cotransfection of phRL-TK directing expression of Renilla luciferase. No significant proliferation-associated activities (as little as with the promoterless pGL3-basic) were detected with the shortest MICB −132 and −192 fragments, of which the latter includes the conserved HSE, Sp1, and ICE sequence motifs. Activities were increased with fragments −225 and −360 and reached a 5-fold higher near maximum strength with the −470 fragment, which was ∼60% of the level recorded with the pGL3-SV40 control (Fig. 2). These results localized MICB core and fully functional promoter regions within −225 and −470 bp, which corresponded to MICA upstream regions of −211 and −455 bp, respectively. However, MICA displayed substantially stronger promoter activities under this study’s examined proliferation-induced condition.

To investigate promoter region requirements for heat shock-induced transcription, HCT116 cells cotransfected with reporter constructs and phRL-TK were grown to high confluence past quiescence, heat shock treated or untreated, and tested for luciferase activity after 6 h of recovery. Parallel to functional testing, heat shock-induced MIC surface protein expression was routinely tested by flow cytometry after 12–16 h and was consistently similar to the data shown in Fig. 1,C. The recovery time period allowed for de novo synthesis of the reporter enzymes that were inactivated by heat shock. However, even after recovery, heat shock reduced pGL3-SV40-driven luciferase activity by ∼70% (Fig. 2). Therefore, to minimize bias in data representation, MIC promoter region activities in both heat shock-treated and untreated cells were normalized against SV40 promoter activities in untreated cells. This approach enabled accurate relative comparisons of different MIC promoter region activities, but underestimated absolute heat shock-induced promoter region strength. As with proliferating cells, the MICB –470 fragment conferred maximum heat shock-induced transcriptional activation, which was ∼8-fold stronger. Notably, fragment –225 also showed near maximum activity, which was opposed to its poor activity in proliferating cells (Fig. 2). Thus, sequences required for strong heat shock-induced activation extended ∼40 bp upstream of the HSE, whereas full proliferation-associated activation was dependent on an additional region up to 240 bp further upstream.

With MICA, heat shock-induced promoter region activities were substantially lower than those of equivalent MICB constructs, which was consistent with lesser heat shock inducibility of MICA mRNA (see figure 3 in Ref. 14). Only fragments –211 and –346 showed significant activation, whereas the –455 promoter region, although comparably active as –346, was of insufficient strength to overcompensate heat shock-induced loss of luciferase activity during the recovery period. Nevertheless, the –455 fragment displayed much stronger induction than the SV40 promoter (Fig. 2). Thus, these results supported heat shock-induced transcriptional activation also of MICA, which was similar to MICB in its dependence on the core promoter, including a short region upstream of the HSE.

As with HSP70, the aligned MICA and MICB core promoter region sequences include inverted repeat 5′-NGAAN-3′ pentamers that are characteristic of HSE (Fig. 3,A) (22). EMSAs using labeled MICB and control HSP70 oligonucleotide probes and purified recombinant protein confirmed in vitro binding of HSF1. This was indicated by the formation of a predominant single or two separate species of DNA-protein complexes, which were supershifted in the presence of anti-HSF1 Ab. Labeled DNA probes were displaced by excess unlabeled MICA or HSP70 HSE oligonucleotides, but not when these were altered by core consensus mutations (Fig. 3,A). Equivalent results were obtained for MICA (data not shown) and by using whole cell extract from heat shock-treated Hela S3 cells (Fig. 3,B). A lower m.w. complex presumably corresponds to a constitutive binding factor that is commonly observed with cell extracts (26). Binding of purified HSF1 to an appropriate location within the MICB promoter region was visualized by DNase I protection assay, which revealed a footprint located approximately between nt positions –174 and –200 (Fig. 3 C).

In vivo binding of HSF1, Sp1, and CBF complex to the core promoter regions of MICA and MICB was investigated by ChIP assay using heat shock-treated or untreated HCT116 cells. Promoter region fragments of ∼300 bp were amplified from cross-linked and sonicated chromatin preparations after immunoprecipitation with Abs specific for the transcription factors, reverse cross-linking, and DNA purification. As with HSP70, heat shock induced rapid HSF1 binding to the core promoter regions of MICA and MICB. With MICB, strong binding persisted during 60 min of heat shock as well as after an additional 90 min of recovery (Fig. 4). With MICA, HSF1 binding appeared diminished after 60 min of heat shock, but increased during the recovery period. This discrepancy was most likely due to sample loss in this particular experiment, as it was not a reproducible finding. Sp1 and CBF bound constitutively under all conditions, but Sp1 binding appeared much weaker to the MICA than to the MICB promoter region (Fig. 4). Altogether, these results demonstrated inducible and constitutive transcription factor occupancies of predicted binding elements in the core promoter regions of MICA and MICB. The apparent weaker binding of HSF1 and Sp1 to the core promoter region of MICA might be related to its less inducible expression (see figure 3 in Ref. 14).

The functional significance of consensus elements and transcription factor binding was studied by mutational analysis and luciferase reporter gene assays. Constructs with site-directed mutations in the HSE, Sp1, ICE, or TATA-like elements within the MICA –455 and MICB –470 regions were analyzed for promoter activities in proliferating and quiescent heat shock-treated HCT116 cells. Under these two conditions, mutation of the Sp1 element decreased luciferase activity by 20 and 80% (MICA), and by 50 and 85% (MICB), respectively. Similar reductions resulted from mutation of the MICA and MICB HSE (20 and 70%, and 15 and 90% with proliferating and heat-shocked cells, respectively). Thus, these results confirmed the significance of the in vivo transcription factor-binding data shown in Fig. 4 and demonstrated critical roles of Sp1 and HSF1 in heat shock-induced and to lesser extents in proliferation-associated transcriptional activation (Fig. 5). By contrast, mutation of the conserved ICE sequence produced opposite effects with MICA and MICB, significantly reducing and enhancing luciferase activities, respectively. Thus, in the context of the MICB promoter, the CBF complex may function as a negative regulator. These discordant effects might be related to the presence of an additional potential ICE in the MICA promoter region 26 bp downstream (Fig. 1,A). Proximal to the conserved ICE is a TATA-like element (at –67 in MICA and –80 in MICB) that was unlikely to have functional significance because the short B-132 and B-192 promoter region fragments showed no or little activity in the reporter gene assays (Fig. 2). However, the unusually positioned TATA-like element 28 bp upstream of the HSE was included in the core promoter region constructs from both genes. Mutation of this element substantially lowered luciferase activities, with reductions of ∼45% (MICA) and 75% (MICB) in proliferating cells, and 45% (MICA and MICB) in heat shock-treated HCT116 cells (Fig. 5). Thus, as with the conserved HSE, Sp1, and ICE elements, this distal TATA-like sequence was essential for optimal MIC gene transcription under both experimental conditions.

The most conspicuous difference between the MIC gene promoters is the 14-bp sequence in MICB (–41 to –54) that includes two overlapping Inr-like elements and is absent in MICA (Fig. 1). Reciprocal insertion (▿) and deletion (Δ) of this sequence in the A-455 and B-470 fully functional, and in the B-225 core promoter constructs, abrogated or severely reduced proliferation-associated and heat shock-induced transcriptional activation. By contrast, insertion of this sequence in the A-211 core promoter construct had a moderately enhancing effect in proliferating cells and resulted in substantially increased activity after heat shock induction (Figs. 2 and 5). Thus, these results revealed profound differences in promoter context-dependent regulation, indicating the presence of MICA and MICB gene-specific sequences that may be associated with negative or positive modulation.

Exposure of HCT116 cells to hydrogen peroxide after transfection with promoter region reporter constructs revealed activation patterns that were similar to those caused by heat shock treatment, although some variances occurred. Moreover, the dependencies on transcription factor-binding elements were comparable (Fig. 6,A). Thus, oxidative stress, which could be associated with MIC expression in tumors and the intestinal mucosa, was a condition that strongly activated MIC gene transcription. Physiological significance was supported by marked increases of MIC surface expression on near confluent HCT116 cells after 72 h of exposure to 0.1 or 1 mM hydrogen peroxide (Fig. 6 B).

CMV infection of human fibroblasts or endothelial cells results in up to 10-fold increases of cell surface MIC and is associated with induced HSP70 expression (see figure 1 in Ref. 6) (27). We therefore anticipated an involvement of HSF1 in CMV-mediated MIC gene activation. Reporter gene assays were carried out with primary human fibroblasts infected or mock infected with CMV strain AD169. Infection had no noticeable effect on the synthesis of control Renilla luciferase used for normalization. The core (A-211 and B-225) and fully functional (A-455 and B-470) promoter regions of MICA and MICB displayed potently induced activities, with ∼20- and 40-, and 10- and 20-fold increases, respectively, over the mock infection expression levels (Fig. 7,A). Strong activation (∼25-fold increase) was also conferred by the MICB –192 region, which produced no or minimal activity in proliferating or quiescent heat shock-treated HCT116 cells (Fig. 2). Moreover, the mutations of the HSE, Sp1, or ICE elements in the fully functional promoter constructs failed to severely suppress activation. Thus, MIC gene induction by CMV was not critically dependent on HSF1 and had no discernible promoter sequence requirements (Fig. 7,A). Cotransfection of reporter constructs with pEQ276 directing the expression of the CMV IE1 and IE2 gene products resulted in 20- and 10-, and 10- and 5-fold inductions of the MICA and MICB core and fully functional promoter regions, respectively, as compared with transfection of negative control pEQ336, which contains the IE2 promoter/enhancer, but no coding region (Fig. 7,B). Moreover, transient expression of IE1 and IE2 together or individually induced cell surface MIC expression, thus indicating that viral trans activation was largely, if not completely, mediated by these CMV gene products (Fig. 7 C).

The present results provide a basic analysis of the architecture and function of the promoter/regulatory regions of the MICA and MICB genes, and of the involvement of transcription factor binding sites under various activating conditions. Similar to HSP70 genes, the conserved MIC gene HSE inducibly binds HSF1 and is critical for activation by heat shock and oxidative stress. Constitutively occupied Sp1 sites and TATA-like elements, which are located unusually far upstream, moderately or profoundly affect stress-induced and proliferation-associated induction. Typically, Sp1 recruits basal transcription machinery in TATA-less CpG island promoters of housekeeping genes (28, 29). As with MIC, Sp1 is required for constitutive and inducible expression of human HSP70 (30). A common ICE appears to be a negative regulator of MICB, but not of MICA. The interacting CBF complex contributes significantly to basal and stress-induced expression of HSP70 and activates genes involved in diverse cellular processes (25, 31); however, there is precedence for negative regulation of a number of genes (32). Although the MICA and MICB promoters are very similar, reflecting the close evolutionary relationship of the two genes, our results provide evidence for differential regulation, which is best exemplified by the context-dependent enhancing function of the 14-bp sequence in MICB. Generally, in accord with mRNA data (see figure 3 in Ref. 14), the baseline transcriptional activities of MICA were higher than those of MICB, which was more stringently regulated and displayed stronger inductions. With regard to transcriptional initiation, our data suggest that the majority of MICA transcripts initiate downstream of the conserved AP-1-TATA-like motifs, a region that lacks a recognizable Inr. By contrast, most MICB transcripts initiate further upstream, proximal of the HSE-Sp1-ICE elements. The Inr at −231 may represent a minor initiation site in MICA and could be driven by potential Sp1 binding sites located at −364 upstream. Although these elements are conserved in MICB, the corresponding Inr is missing because of a critical cytosine for adenosine substitution (Fig. 1). There was no evidence for an involvement of separate promoters during proliferation or heat shock-associated gene activation.

Not surprisingly, MIC lack the prototypic MHC class I gene regulatory elements located between about –95 to −220 upstream of the translation initiation codons. These are constituted by the S-X-Y module, which binds RFX, CREB, and CBF complexes representing the MHC enhanceosome, and by the IFN-stimulated response element and enhancer A, which interact with IFN-regulated factor and NF-κB, respectively (33). However, MHC class I gene core promoters include TATA- and Inr-like motifs and a CA/GT-rich region that binds Sp1, but these elements are differently organized. By contrast, considerable symmetry is shared between MIC and HSP70 promoters in the arrangement of the AP-1/TATA-like (TATA in HSP70), CBE, Sp1, and HSE elements, although intervals vary (11, 12).

In the induction of MIC by CMV, there were only modest effects of mutations in the HSF1-, Sp1-, and CBF-binding elements, although combinatorial effects were not assessed, and activation was independent of sequences upstream in the fully functional promoter regions. Expression of CMV IE1 or IE2 alone was sufficient for trans activation. These nuclear phosphoproteins are promiscuous activators of many viral and cellular genes, such as of c-fos, c-myc, and HSP70, and interact with various transcription factors within very short basal promoter regions (34). The activation mode by IE1 and presumably IE2 involves chromatin remodeling through displacement of histone deacetylases (35). This mechanism may also operate in the trans activation of MIC gene expression in CD4 and CD8 T cells by the viral Tax regulator protein in human T cell leukemia virus-1-associated neurologic disease (36). Histone deacetylase inhibitors have been shown to induce MIC gene expression (37).

The similarities in regulation between HSP70 and MIC genes suggest that conditions in tumor environments that lead to HSP70 induction, such as oxidative stress, hypoxia, and hypoglycemia, may also favor MIC expression. Moreover, oxidative stress could be a critical condition underlying the characteristic MIC expression in normal intestinal epithelium (4). Our results provide no insights into the regulation of MIC in autoimmune lesions. However, in rheumatoid arthritis, proliferating, but not quiescent synoviocytes express MIC, resulting in costimulation of autoreactive CD28⁻NKG2D⁺ CD4 T cells, and thus presumably in exacerbation of disease progression (8). In this regard, the present results demonstrate the requirement of MIC fully functional promoter region sequences and of defined core promoter elements for maximum proliferation-associated transcriptional induction. It remains unknown, however, whether chromatin modifications may also be involved. Altogether, our results provide a molecular framework of MIC gene regulation that may be applicable to future studies of gene expression in malignancies, infections, and autoimmune disease.

We thank Adam Geballe for the CMV IE gene constructs and advice, and Kimberly Smythe for technical help.

The authors have no financial conflict of interest.